Volume 04 - Heat Treating Part 10 pdf

If the tools used for process control in heat treating today are viewed against this backdrop, our current position on the evolutionary ladder can be pinpointed: • Step 1: Rudimentary pr

case-hardening steels, increases warpage on quenching (Ref 80)

• Tight (that is, thin and highly adherent) scale and decarburization, at least in certain areas Tight scale is usually a problem encountered in forgings hardened from direct-fired gas furnaces having high-pressure burners Quenching in areas with tight scale is extremely retarded compared to the areas where the scale comes off This produces soft spots, and, in some cases, severe unpredicted distortion Some heat treaters coat the components with a scale-loosening chemical prior to their entry into the furnace (Ref 79) Similarly, the areas beneath the decarburized surface do not harden as completely as the areas below the nondecarburized surface The decarburized layer also varies in depth and produces an inconsistent softer region as compared to the region with full carbon All these factors can cause a condition of unbalanced stresses with resultant distortion (Ref 79)

• Long parts with small cross sections (>L = 5d for water quenching, >L = 8d for oil quenching, and >L = 10d for austempering, where L is the length of the part, and d is its diameter or thickness)

• Thin parts with larger areas (>A = 50t, where A is the area of the part, and t is its thickness)

• Unevenness of, or greater variation in, section

Examples of Distortion

Ring Die. Quenching of ring die through the bore produces the reduction in bore diameter as a result of formation of martensite, associated with the increased volume In other words, metal in the bore is upset by shrinkage of the surrounding metal and is short when it cools (Ref 24) However, allover quenching causes the outside diameter to increase and the bore diameter to increase or decrease, depending upon precise dimensions of the part When the outside diameter of the steel part is induction- or flame-hardened (with water quench), it causes the part to shrink in outer diameter (Ref 63) These are the examples of the effect of mode of quenching on distortion (Ref 81)

Thin die (with respect to wall thickness) is likely to increase in bore diameter, decrease in outside diameter, and decrease

in thickness when the faces are hardened If the die has a very small hole, insufficient quenching of the bore may enlarge the hole diameter because the body of die moves with the outside hardened portion

Bore of Finished Gear. Similarly, the bore of a finished gear might turn oval or change to such an extent that the shaft cannot be fitted by the allowances that have been provided Even a simple shape such as a diaphragm or orifice plate may, after heat treatment, lose its flatness in such a way that it may become unusable

Production of Long Pins. In the case of the production of long pins (250 mm long × 6 mm diameter, or 10 × 1

4in.) made from medium-alloy steel, it was found, after conventional hardening, that when mounted between centers, the maximum swing was over 5 mm (0.20 in.) However, the camber could be reduced to within acceptable limits by martempering, intense or press quenching

Hardening and Annealing of Long Bar. When a 1% carbon steel bar, 300 mm long (or more) × 25 mm diameter (12 in long, or more, × 1 in diameter), is water quenched vertically from 780 °C (1435 °F), the bar increases both in diameter and volume but decreases in length When such bars are annealed or austenitized, they will sag badly between the widely spaced supports Hence, they should be supported along their entire length in order to avoid distortion

Hardening of Half-Round Files. Files are usually made from hypereutectoid steel containing 0.5% chromium Files are heated to 760 °C (1400 °F) in an electric furnace after being surface coated with powdered wheat, charcoal, and ferrocyanide to prevent decarburization They are then quenched vertically in a water tank On their removal from the tank, the files appear like the proverbial dog's tail The flat side has curved down, the camber becomes excessive, and the files can no longer be used in service One practical solution is to give the files a reverse camber prior to quenching The dead flat files could, however, be made possible, and the judgment with regard to the actual camber needed depends upon the length and the slenderness of the recut files (Ref 82)

Similarly, when a long slender shear knife is heat treated, it tends to curve like a dog's tail, unless special precautions are taken

Hardening of Chisels (Ref 63) Chisels about 460 mm (18 in.) long and made from 13 mm (0.5 in.) AISI 6150 bar steel are austenitized at 900 °C (1650 °F) for 1.5 h and quenched in oil at 180 °C (360 °F) by standing in the vertical position with chisel point down in special baskets that allow stacking of two 13 mm (0.5 in.) round chisels per 650 mm2

(1 in.) hole Subsequently, hardened chisels are tempered between 205 and 215 °C (400 and 420 °F) for 1.5 h These heat-treated parts show 55 to 57 HRC hardness but are warped The reasons for this distortion are:

• The portion of the bar that touches the basket cools slowly, producing uneven contraction and thermal stress

• The martensite formation is delayed on the inner or abutting side of the bar, causing unequal expansion during transformation This distortion can be eliminated or minimized by loading the parts in the screen- basket in such a way that stacking arrangement permits sufficient space between each part and by slightly decreasing the austenitizing temperature (Ref 62) Distortion can also be minimized by austempering the part, provided that the carbon content is on the high side of specification to produce the lower bainitic structure of 55 to 57 HRC If higher yield stress is not warranted, only chisel ends need hardening and subsequent tempering (Ref 63)

Hardening of a Two-Pounder Shot. The hardness of a two-pounder shot was specified at 60 HRC on the nose and

35 HRC at the base A differential hardening technique was performed on the shot made of a Ni-Cr-Mo steel This technique consisted of quenching the shot in the ice-cold water by its immersion in a tank up to the shoulder, followed by drawing out the water from the tank at a stipulated rate until the water line reached the base of the nose The final step involved withdrawing the shot from the tank when completely cold The back end was then softened by heating in a lead bath after initial tempering The first few shots hardened in this way were observed to split vertically across the nose The failure was, however, avoided by withdrawal of the shot before attaining ice-cold temperature and its subsequent immersion in warm water (Ref 82)

Hardening of a Burnishing Wheel. In the manufacture of railway axles, the gearing surface on which the axle rests

in the housing has to be given a high burnishing polish employing a circular pressure tool that is made of 1.2C-1.5Cr steel For satisfactory results, the hardness of the tool surface should be about 60 HRC It has been found that the tool usually cracks before its withdrawal from the cold-water quenching bath This problem may, however, be avoided by quenching the tool in water for 10 s prior to transferring it to an oil bath for finish quenching Time quenching can be judiciously applied for many heat treatment problems of distortion or cracking Stress-relieving treatment after the use of the tool for some time may also enhance its performance life As indicated above, martempering is also one of the solutions for this problem (Ref 81)

Hardening of Case-Carburized Mild Steel. If oil-hardening steels are not available for making a component, mild steel parts are carburized and water quenched to obtain the desired hardness, possibly resulting in excessive distortion, which is very difficult to straighten without cracking

Hardening of Carburized Low-Carbon Steel Rollers. The best course of quenching carburized En32 steel rollers (25 mm diam × ≥600 mm long, or 1 in diam × ≥2 ft long), employed in textile printing, is to roll them down skids into water-quenching tanks because this produces less warpage than when quenched slowly with the bar either in vertical, horizontal, or inclined positions These are the procedures adopted for hardening of cylinders with length considerably greater than the diameter

Hardening of Helix Gears. The distortion of the helical gears made of IS 20MnCr1 grade steel (similar to AISI 5120) used as the third speed gear in the gear box of Tata trucks is an unavoidable natural consequence of the hardening process after carburizing This type of distortion is linked with increased length and decreased diameter and occasionally increased helical angle (Ref 83) If the extent of distortion can be controlled, a constant correction to the helix angle can

be imparted in the soft-stage manufacturing (machining) prior to heat treatment so that this correction can compensate for the distorted angle and may result in a gear with desired helix angle Thus a constant magnitude of distortion without minimization is assured in every job of every batch of production in commercial manufacturing However, the residual stress system and metallurgical properties such as core strength, case depth, surface hardness, proper microhardness in the surface regions, and so forth, are assured (Ref 84) Similarly, when heavy-duty tooth gear is gas carburized and quenched

to harden the surface layer, the diameter and tooth span increase and tapering and bending also occur

Nitriding of Screw. A rolling mill screw, after liquid nitriding, may also show a small decrease in length, which causes pitch errors in the screws (Ref 83)

Induction and Flame Hardening of Spur Gears. Spur gears, after induction and flame hardening, exhibit increased circular pitch, the error being maximum for the tooth groove quenched first Similarly, in line-heating process, the thin plate undergoes convex bending and the thick plate concave bending (Ref 83)

Precautions

Inadequate support during the heat-treatment cycle, poorly designed jigs and quenching fixtures, or incorrect loading

of the parts may cause distortion (Ref 73) In general, plain-carbon and low-alloy steels have such a low yield strength at the hardening temperature that the parts are capable of distorting under their own weight Every care, therefore, must be taken to ensure that parts are carefully supported or suspended during heating; long parts are preferably heated in a vertical furnace or with the length in the vertical plane (Ref 85) They should be quenched in the vertical position with vertical agitation of the quenchants Also, it must be remembered that many tool steels are spoiled by failure to provide enough support when they are taken out from the furnace for quenching Thus, every precaution is taken to ensure that parts are adequately supported during entire heat treatment by employing well-designed jigs, fixtures, and so on

Other precautions to minimize distortion include:

• Tool steels should be heated to hardening temperature slowly, or in steps, and uniformly Hot salt baths are used to render fast, uniform heat input

• It is best to heat small sections to the lower region of the recommended hardening temperature range and to heat large sections at the higher temperature range Overheating by employing too high a temperature or too long a heating time must be avoided

• It is a good practice to protect the surface of the component from decarburization (by packing it in cast iron chips or using a vacuum furnace, for example) If a separate preheating furnace is not available, the part can be put in a cold furnace, after which the temperature is raised to proper preheating temperature and kept at that temperature to attain uniform heating throughout, prior to proceeding to the hardening temperature (Ref 86)

• With the slower cooling rate, which is consistent with good hardening practice, a lower thermal gradient will be developed, thereby producing less distortion

• Thus rapid heating and cooling rates of irregularly shaped parts must be avoided

• Proper selection of quenchant with desirable quenching properties and adequate agitation during hardening must be provided

Methods of Preventing Distortion (Ref 82, 87)

Straightening is one method to remove or minimize distortion Since straightening (after hardening) can largely relieve the desirable residual compressive stresses (in plain-carbon and low-alloy steels) that may cause breakage, it would be better to accomplish this before the steel cools below the Ms temperature, that is, when the steel is in the metastable austenitic state (Ref 35) This temperature is above 260 °C (500 °F) for most tool steels and is preferably about 400 °C (750 °F) for long shear knives, which are usually made of 2C-12Cr steel Warping on parts such as shafts and spindles can be corrected by straightening during or after hardening, followed by grinding to size (Ref 84) Mostly high-alloy steels are straightened after hardening due to the higher percentage of retained austenite and their comparatively low yield stress Straightening also can be accomplished during the tempering process (Ref 35) However, straightening of hardened parts with higher strength will cause a loss of fatigue properties and possibly initiation of cracks at the surface Hence, straightening after the hardening treatment must be very carefully controlled and should be followed by a low-temperature tempering treatment

The case-hardened (for example, nitrided, carburized) parts can be straightened to a very large extent as a result of their lower core hardness Nitrided parts may be straightened at 400 °C (750 °F) (Ref 35)

Support and Restraint Fixtures. Fixtures for holding finished parts or assemblies during heat treatment may be either support or restraint type For alloys that are subjected to very rapid cooling from the solution-treatment temperature, it is common practice to use minimum fixturing during solution treatment and to control dimensional relations by using restraining fixture during aging Support fixtures are used when restraint type is not needed or when the part itself renders adequate self restraint Long narrow parts are very easily fixtured by hanging vertically Asymmetrical parts may be supported by placing on a tray of sand or a ceramic casting formed to the shape of the part (Ref 64) Restraint fixtures may require machined grooves, plugs, or clamps Some straightening of parts can be accomplished in

aging fixtures by forcing and clamping slightly distorted parts into the fixture The threaded fasteners for clamping should not be used because they are difficult to remove after heat treatment It is preferable to use a slotted bar held in place by a wedge (Ref 64) The bore of a hub, the most important dimension in the hardening of thin spur gears, can be mechanically plugged to prevent the reduction of the bore and keep the out-of-roundness close to tolerance limits When hardening large hollows, either restraining bands on the outside during tempering or articulated fillers serve the same purpose

Quenching Fixtures. When water quenching or oil quenching is essential, distortion can be minimal by employing properly designed quenching fixtures that forcibly prevent the steel from distorting (Ref 88) Figure 14 shows a typical impingement-type quenching fixture The requirements essential for the better design of this type of fixture are as follows (Ref 79):

• There must be an accurate positioning of the part in the fixture Whenever possible, round bars should

be rotated during quenching to level out variations in jet pressure around the part

• There should be an unhindered flow of quenchant through the sufficiently large holes (3.3 to 6.4 mm, or 0.13 to 0.25 in in diameter) Jets as large as 12.25 mm (0.50 in.) in diameter may be employed with furnace-heated heavy sections (for example, plates) A large portion of the excess quenchant with these large jets is for the removal of scale (Ref 89)

• Spacing between the holes should be reasonably wide (for example, 4d, where d is the hole diameter)

• For oil-quenching fixtures, the facility to submerge the part is required to reduce fumes and flashing

• There must be the provision for efficient cleaning of the holes

• A facility must be available to drain out the hot quenchant for effective quenching performance with cold quenchant

Fig 14 A typical impingement-type quenching fixture Source: Ref 80

Pressure quenching is the most efficient method of cooling parts from elevated temperature by using a combination

of high pressure (such as 5 MPa, or 5 atm) and turbulent gas flow throughout the entire surface area of the workload (Ref 90) This is economical and fast, provides even cooling, offers a unique design and minimum distortion and improved metallurgical qualities As a result of these beneficial effects this is suited to quench large-diameter tooling for the aluminum extrusion industry; quench larger-diameter carburized gear, larger fasteners, and precision gears to be jigged vertically; harden high-speed steel tools (such as saw blades, dies, and other parts with edge configuration) and 718 jet engine compressor blades (Ref 90) This is also employed to quench (vacuum processed) large sections of titanium alloy castings for aircraft applications (Ref 91) Figure 15 is a pressure-quench module that may be attached to vacuum-sealed quenched and continuous-vacuum furnace as a replacement for the oil-quench section

Fig 15 Pressure-quench module for attachment into standard vacuum-sealed quenched and continuous

vacuum furnaces Source: Ref 90

Press quenching is widely employed in preventing and controlling quench distortion in components where the geometry renders them particularly prone to distortion (Ref 92) For example, flat circular diaphragms of spring steel used

in the control or measurement of pressure are press quenched between two copper blocks, which cannot be accomplished

by direct quenching (Ref 80)

Rolling Die Quenching. A rolling die quench machine can provide uniform water quenching with minimal distortion for large-production runs When a heated part is placed on the rollers, the die closes and the rolls turn This removes any distortion incurred during heating According to manufacturers of rolling die quench machines, symmetrical parts with the following straightness can be achieved in production:

l TIR K

to stress relieve plain carbon or low-alloy steel parts at a temperature of 550 to 650 °C (1020 to 1200 °F) (for 1 to 2 h),

hot-worked and high-speed steels at 600 to 750 °C (1110 to 1380 °F), and the heavily machined or large parts at 650 °C (1200 °F) (for 4 h) prior to final machining and heat-treatment operations Subresonant stress relieving may also be employed to neutralize thermally induced stress without changing the mechanical properties or the shape of the component These components include: large workpieces, premachined or finish-machined structural or tubular, nonferrous, hardened, nonsymmetrical or varying section thickness, stationary, or assembled However, this does not work on copper-rich alloys and the edges of burned plates (Ref 93)

Control of Distortion

In order to remove or minimize distortion, the modern trend is to shift from water-quenching practice to milder quenching, for example, oil quenching, polymer quenching, martempering, austempering, or even air-hardening practice Milder quenchants produce slower and more uniform cooling of the parts, which drastically reduces the potential distortion Other strategies of controlling distortion for age-hardening aluminum, beryllium, and other alloys include: alloy and temper selection, fixturing, age-hardening temperatures, proper machining, and stamping operations (Ref 94) The fewer the number of reheats applied to components in case-hardening steels following carburizing, the smaller is the distortion on the finished part When top priority is given to minimum distortion, it is desirable to make the parts from oil-hardening steels with a controlled grain size and to harden them by martempering direct from carburizing Presently polyalkylene glycol-base quenchants, such as UCON quenchants HT and HT-NN, are variously used for direct quenching from the forging treatment, continuous cast quenching, and usual hardening of forged and cast steels and cast iron In this case boiling does not take place at the component surface but rather at the external surface of the deposited polymer film More uniform cooling occurs, and thermal stresses are released Because of the lower boiling point and high thermal conductivity, UCON quenchants act through the martensite zone more rapidly than oil (Ref 95)

Distortion during ferritic nitrocarburizing is minimal because of low treatment temperature and the absence of subsequent phase transformations (Ref 66) There are many methods of reducing distortion in induction-hardened components; these methods are usually found by experience with variables such as the hardening temperature and the type and temperature

of quenching medium employed Methods of reducing distortion in induction-hardened parts include: the hardening of small spindles held vertically in jigs; the plug-quenching of gears to prevent the bores from closing in; the flattening of cams by clamping them together during tempering; and the selective hardening of complex shapes (Ref 96)

As a replacement of medium- or slow-quenching oils, UCON quenchants E and E-NN can be readily used in induction- and flame-hardening operations, both in spray and immersion types, for high-carbon and most alloy steels and traditional hardening of cast iron and cast or forged steels of complex geometry with better distortion-reduction properties Agitation

of quenchant should be carried out by motor-driven stirrers to move the medium with respect to the part being quenched

or by pumps that force the medium through the appropriate orifice Alternatively, the parts are moved through the medium, and for some applications, spray quenchant is recommended Water additives are sometimes employed in salt baths to increase heat extraction (Ref 64)

Ultrasonic quenching is also effective in controlling distortion, which involves the introduction of ultrasonic energy (waves with a frequency of 25 kHz) in the quenching bath This breaks down the vapor film that surrounds the part in the initial stages of water or oil quenching (Ref 86)

Distortion after Heat Treatment

Straightening. When every possible case has been employed to minimize distortion, it may still be essential to

straighten after heat treatment, which has already been discussed

Grinding after Heat Treatment. In the case of carburized or nitrided parts, the metallurgist and designer, together with the production engineer, must collaborate regarding the amount to be removed by grinding after heat treatment This grinding allowance must be taken into account when determining the initial dimensions and also when specification for the case depth is to be applied

Distortion may also occur after heat treatment, with time, owing to the completion of any unfinished transformation or the effect of increased temperature during grinding For example, fully hardened components such as blade shears may be damaged by characteristic crazing pattern because of heavy and careless grinding Local overheating results in the transformation of undecomposed austenite, and the accompanying changes in volume produce sufficient stresses to cause cracking and developing of a crazing pattern

Dimensional Stability. To achieve dimensional stabilization or stability (that is, retention of their exact size and shape) over long periods, which is a vital requirement for gages and test blocks, the amount of retained austenite in heat-treated parts must be reduced because retained austenite slowly transforms and produces distortion when the material is kept at room temperature, heated, or subjected to stress Dimensional stabilization also reduces internal (residual) stress, which causes distortion in service

Stabilization can be obtained by multiple tempering (with prolonged tempering times); the first tempering reduces internal stress and facilitates its transformation to martensite on cooling The second and third retempering reduce the internal stress produced during the transformation of retained austenite

It is the usual practice to carry out a single or repeated cold treatment after the initial tempering treatment In cold treatment, the part is cooled below the Mf, which will cause the retained austenite to transform to martensite; the extent of transformation depends on whether the tool part is untempered or first tempered Cold treatment is normally accomplished in a refrigerator at a temperature of -70 to -95 °C (-100 to -140 °F) Tools must be retempered immediately after return to room temperature following cold treatment in order to reduce internal stress and increase the toughness of the fresh martensite Finally, they are ground to size It may be pointed out that vibratory techniques are being used more frequently to achieve dimensional stability but do not offer any metallurgical benefits (Ref 80)

Distortion and Its Control in Heat-Treated Aluminum Alloys

The high levels of residual stress and distortion that are produced in the water-quenched aluminum extrusion and forgings (such as 2000, 6000, and 7000 series) and aluminum castings can be reduced 60 to 100% by using proper selection of polyalkylene glycol quenchant or polyvinyl pyrrolidone 90 concentration (for example, 25% solutions for wrought alloys,

20 to 30% UCON quenchant A for thicknesses up to 25 mm (1 in.), and 17 to 22% for larger than 25 mm (1 in.) section thicknesses in casting alloys) with sufficient agitation, lower bath temperature, proper fixture (throughout solutionizing, quenching, and age-hardening treatments), and straightening (in the as-quenched state after taking out from the fixture) procedure The initial cost of these polymer solutions as a replacement to the conventional hot-water quenching method is easily compensated for by other advantages such as reduced scrap, reduced machining (compared to two machining operations required one before and another after heat treatment in the conventional water-quenching method), and increased fatigue life as a result of reduced convective heat transfer or film coefficient between the part and the quenchant, more uniform quench, precise control of quench rates, and improved heat-transfer qualities from the deposition of liquid organic polymer on the surface of the part being quenched (Ref 97, 98, 99) This method costs less, therefore saves time and allows easy shaping, bending, and twisting of the parts without establishing residual stresses Such parts as leading edge wing skins, spars, and bulkheads are used in the aerospace industries (Ref 96)

References cited in this section

24 R.F Kern, Selecting Steels and Designing Parts for Heat Treatment, American Society for Metals, 1969

35 K.E Thelning, Steel and Its Heat Treatment, Butterworths, 1985

62 D.H Stone, in Proceedings of the 1988 ASME/IEEE Joint Railroad Conference, American Society of

Mechanical Engineers, 1988, p 43-53

63 C.E "Joe" Devis, Ask Joe, American Society for Metals, 1983

64 Chapter 8, in Troubleshooting Manufacturing Processes, 4th ed., L.K Gillespie, Ed., Society of

Manufacturing Engineers, 1988

66 G Wahl and I.V Etchells, in Proceedings of Heat Treatment '81, Metals Society, 1983, p 116-122

71 G.E Hollox and R.T Von Bergn, Heat Treat Met., No 2, 1978, p 27-31

72 T Bell, Survey of Heat Treatment of Engineering Components, Iron and Steel Institute, 1973, p 69-72

73 K.W Chambers, Heat Treatment of Metals, Iron and Steel Institute, 1966, p 94-95

74 R Wilson, Metallurgy and Heat Treatment of Tool Steels, McGraw-Hill, 1975, p 93-95

75 P.G Greenwood and R.H Johnson, Proc R Soc., Vol A283, 1965, p 403

76 B.L Josefson, Mater Sci Technol., Vol 1 (No 10), 1985, p 904-908

77 A Ferrante, Met Prog., Vol 87, 1965, p 87-90

78 B.R Wilding, Heat Treatment of Engineering Components, Iron and Steel Institute, 1970, p 20-25

79 R.F Kern, Heat Treat., Vol 17 (No 3), 1985, p 41-45

80 D.J Grieve, Metall Mater Technol., Vol 7 (No 8), 1975, p 397-403

81 F.D Waterfall, in Met Treat Drop Forg., April 1985, p 139-144

82 S Visvanathan, TISCO J., Vol 23 (No 4), 1976, p 199-204

83 Y Toshioka, Mater Sci Technol., Vol 1 (No 10), 1985, p 883-892

84 R Verma, V.A Swaroop, and A.K Roy, TISCO J., Oct 1977, p 157-160

85 Section 8 in Cassels Handbook, 9th ed., ICI Ltd., 1964

86 R.F Harvey, Met Prog, Vol 79 (No 6), 1961, p 73-75

87 A.K Sinha, Tool Alloy Steels, Aug 1980, p 219-224

88 G.F Melloy, Hardening of Steel, Lesson 5, in Heat Treatment of Steels, Metals Engineering Institute,

American Society for Metals, 1979, p 1-28

89 R.F Kern, Heat Treat., Vol 18 (No 9), 1986, p 19-23

90 Hayes, Inc., private communication, Oct 1989

91 J.M Neiderman and C.H Luiten, Proceedings of Heat Treatment '84, Metals Society, 2984, p 43.1-43.8

92 Met Mater., Vol 9, July/August 1975, p 52-53

93 T.E Hebel, Heat Treat., Vol 21 (No 9), 1989, p 29-31

94 F Dunlevey, Heat Treat., Vol 21 (No 2), 1989, p 34-35

95 "UCON Quenchants for Ferrous and Nonferrous Metals," Tenaxol, Inc., 1988

96 R Creal, Heat Treat., Vol 18 (No 12), 1986, p 27-29

97 C.E Bates, J Heat Treat., Vol 5 (No 1), 1987, p 27-40

98 "Information on Polymer Quenchants," Tenaxol, Inc., 1989

99 C.E Bates and G.E Totten, Heat Treat Met., No 4, 1988, p 89-97

Importance of Design

The wrong design of the tool material may result in the establishment of nonuniform heating and cooling of the components, which produces overload and/or internal stresses leading to distortion and failure during or after hardening Correct consideration at the design stage plays an important role in lessening the distortion and danger of cracking The basic principle of successful design is to plan shapes that will minimize the temperature gradient through the part during quenching Fundamental rules such as maintaining a simple, uniform, regular, and symmetrical section with comparatively few shape changes, ensuring small and smooth cross-sectional size changes, and using large radii are still too frequently overlooked at the design stage Thus, successful heat treatment demands a rational design that avoids sharp corners as well as sudden and undue changes of section

It is often possible for tool designers to compensate for size distortion For example, in preparing precision hobs for gear cutting, dimensional accuracy must be kept within very close tolerances On linear longitudinal growth, it is the general practice to go out-of-round in the following high-speed steel bars as much as 0.3% in M1 type, 0.2% in M2 type, and 0.15% in T1 type during heat treatment These data will alter slightly with changes in design of the hobs, but essentially the growth in tungsten-base high-speed steel is lower than that of the molybdenum-base high-speed steel (M1 and M2) This does not require any difficulty if the growth is compensated for and if the steel is consistent in its growth (Ref 87)

The distortion produced in the surface hardening of long shafts by the scanning method can be a great problem if the equipment is not in very good condition Due consideration must be given so that locating centers run concentrically, in line and at the appropriate speed; the coil must be accurately aligned, and the quench must be correctly designed with sufficient number of holes of suitable size and angle For long shafts with a relatively small diameter (for example, half-shafts, which are likely to distort), the use of hydraulically operated restraining rolls usually overcomes this (Ref 100) The designer should bear in mind the following rules while designing a die or machine part that is to be heat treated:

• Distribution of the material should be as uniform as possible

• Provide fillets (large radii) at the base of keyways, cutter teeth, and gear teeth to avoid stress concentration; semicircular keyways, which permit the use of round-cornered keyways, are the right

choices Ideally, drives using involute splines are preferred over keyways

• Avoid abrupt changes of section; in other words, provide smooth changes of section

• Large holes (such as drawing or cutting openings in die rings or plates) must be centrally located from the outer contour In some cases holes are drilled through the heaviest section of the tool in order to help fairly balance the weight of the section rather than to unbalance it (Ref 64) Deep blind holes should always be avoided because they cause nonuniform quenching If this is not possible, the hole can be ground in after hardening Drilled hole junctions in a steel part should be avoided because they enhance very high and undesirable cooling conditions The problem with these cross holes is to get sufficient quenchant into them The inside surface of the holes tends to be in a state of high tensile stress, usually leading to cracking, at least with water quenching As a minimum, the corner at the junction of the holes with outer diameter of the part should be given a generous radius to better distribute the tensile stress (Ref 90) Similarly, grooves and keyways in highly stressed areas should be avoided, or, if possible, they should be located in low-stressed areas of the part Alternatively, fixtures should be used that make

it possible for the hole or the inside of the groove to be quenched in the beginning or more rapidly than the rest of the part (Ref 24)

• Round off all the holes, corners, and outer edges

• If sharp corners are unavoidable, provide relief notches in place of sharp edges

• The insertion of identification marks on the hardened component is recommended, preferably after hardening with tools having well-rounded edges and minimum deformation (shallow penetration depth), and at positions far away from the high-stress concentration zones (reentrant angles, bends, and so on) (Ref 101)

• Large intricate dies should be made up in sections, which frequently simplifies heat treatment (Ref 64)

References cited in this section

24 R.F Kern, Selecting Steels and Designing Parts for Heat Treatment, American Society for Metals, 1969

64 Chapter 8, in Troubleshooting Manufacturing Processes, 4th ed., L.K Gillespie, Ed., Society of

Manufacturing Engineers, 1988

87 A.K Sinha, Tool Alloy Steels, Aug 1980, p 219-224

90 Hayes, Inc., private communication, Oct 1989

100 P.D Jenkins, Metallurgia, Vol 45 (No 4), 1978, p 196-199

101 F Strasser, Heat Treat Met., No 4, 1980, p 91-96

Statistical Process Control of Heat-Treating Operations

Jon L Dossett, Midland Metal Treating, Inc.; Gordon M Baker, New Age Industries, Inc.; Terrence D Brown, Lindberg Heat Treating Company; Daniel W McCurdy, Marathon Monitors, Inc

Introduction

AS DEMAND FOR INCREASED QUALITY and documentation is felt by heat treaters, the subject of automatic collection and use of process information in a statistical process control/statistical quality control (SPC/SQC) format becomes increasingly critical Data acquisition and documentation 10 years ago meant a chart recorder for temperature and a log sheet for the operator's dew-point readings Today, it more than likely means a computer system tied into key points on the heat-treating equipment with the objective of logging important information for later review or perhaps being taken into account in real time

Traditional versus Statistical Control

When man first heat treated a sword made from an iron carbon alloy, he learned that it was necessary to heat it until it glowed red and then plunge it into water He further learned that the resulting product was often very brittle but could be made more usable by heating it again, this time to a much lower temperature This small body of knowledge was enough process control to produce many useful products for many years

Initially, it was noticed that if certain temperature ranges were used in certain circumstances, products of better quality would result and the idea of controlling temperature was born Much later it was discovered that control of the carbon content of the material was important Modern science was now controlling the properties of the end product

If the tools used for process control in heat treating today are viewed against this backdrop, our current position on the evolutionary ladder can be pinpointed:

• Step 1: Rudimentary process knowledge from direct observation

• Step 2: Understanding of certain obvious influences from experimentation

• Step 3: Manual control of obvious influences like temperature

• Step 4: Automatic control of these obvious influences

• Step 5: Documentation of continued variation in process results, using statistical techniques to manually

identify special problems

• Step 6: Use results of statistics and enhanced process understanding to gain control of the less obvious

influences

• Step 7: Control the process from a theoretically complete model, taking into account all possible

influences to produce a near-perfect product every time

The state of the art is currently at Step 5, the application of statistics in the search for problems The jump to Step 6 is being made in some manufacturing disciplines using a new technique called "design of experiments." This is a complex statistical approach that may incorporate artificial learning into data regression-based computer programs A program of this type will direct the human operator to perform experiment after experiment with a process in order to gain insight into any possible effects, direct or synergistic, that an entire list of possible process parameters might have A process model initialized with known theory, but ultimately based on derived statistics, will emerge that can be used to indicate those parameters that will make significant contributions and should therefore be subject to automatic control

A model such as this could be used to bake a better cake, for example The model might direct the operator to make a cake with 1 egg the first time, 0.5 eggs the second, and 1.5 eggs the third It would then ask for quantified results as to the measured quality of each of the experimental cakes The model might conclude by saying that the optimum cake must be made with between 1.1 and 1.2 eggs, along with similar odd amounts of all the other applicable ingredients It might even conclude that no salt is necessary in a cake because the statistics did not bear out importance of this item

Basic SPC/SPQ Nomenclature

The purpose of this article is to provide a practical discussion of the application of SQC techniques to heat-treating operations and for that reason many of the applicable definitions and equations are not used but can be found by the reader in other reference materials on this subject However, it is felt that the following minimum basic definitions and equations are necessary to be presented here for a better understanding of the text

• Accuracy versus precision: Accuracy is measured by the extent to which the measured average of a

group of readings, regardless of how widely the readings are dispersed, agrees with the true value of the unit being measured Precision is the repeatability of the measurement (how much dispersion exists between readings) regardless of how close the readings are to the true value (or how accurate the readings are)

• Gage repeatability and reproducibility study (GR & R): A study conducted on measurement devices to

determine the precision and accuracy of the device Results are expressed as an R & R index

• Process: Any specific combination of machines, tools, methods, materials, and/or people employed to

attain specific output in a product or a service A change in any one of the constituents results in a new process

• Process capability: Refers to the reproducibility of a process over a long time period during which

normal changes in workers, material, and other conditions are encountered

• Quality: Product features which are free from deficiencies and thus meet the needs of the customer and

provide product satisfaction

• Statistical process control: The application of statistical techniques for measuring and analyzing the

variation in processes

• Statistical quality control: The application of statistical techniques for measuring and improving the

quality of processes and products Statistical quality control includes statistical process control, diagnostic tools, sampling plans, and other statistical techniques

• Cp index: Stands for capability of process, and is the ratio of the specification tolerance to six standard

deviations (6σ) Cp is a measure of the dispersion of data only

• Cpk index: A measure of both dispersion and centeredness of the data as follows:

where USL is the upper specification limit, and LSL is the lower specification limit

Use of Statistical Control in Heat Treating. In the last few years, it has become commonplace to see heat treaters tracking the results of their most critical processing with SPC techniques The most commonly examined parameters include hardness, case depth, and distortion, although many others are tracked as well Unfortunately, it has also become commonplace to observe these very same heat treaters failing to use these techniques for anything more than supplying required after-the-fact documentation on treated parts It seems that the promise of statistical process control and its roll in the revitalization of manufacturing quality in heat treatment is not being fulfilled

It is important to understand that any SPC program is a means, not an end Statistical process control is a tool to audit a process and help distinguish controllable variables from uncontrollable variables It also provides a means for quantitatively measuring the level quality of a process Statistical process control alone does nothing to improve the process Continual improvement of the heat-treating process is the real goal and this comes from actions taken by people Statistical process control is utilized as a tool to direct those actions

Simply installing an SPC system on a dilapidated furnace does not improve the performance of the furnace SPC alerts the furnace operator to the fact that, statistically, a problem does exist and requires investigators to determine which variables

are causing excessive variations to occur Statistical process control does not tell what is wrong, only that, statistically,

excessive variation is present and this condition should be investigated to determine the assignable cause

SPC versus SQC. It is necessary to make a distinction between statistical process control and its relative, statistical quality control The latter is what most heat treatment shops are really using when histograms, mean/range charts, and capability indices are calculated for the variation in attained results (for example, hardness, case depth, and so on) in the processing of a given component

Because statistical quality control is an after-the-fact tool, its best use is in the control of continuous processing where trends can be noticed and corrected before significant damage occurs Processes such as large volume induction hardening (see the article "Induction Heat Treating of Steel" in this Volume) and continuous carburizing have been substantially improved with SQC charting techniques

In batch processing, however, statistical quality control is of little value in preventing problems because at least one entire load of parts will be adversely affected before a problem can even be noticed Even if the problem is caught after one load, the proposed solution cannot be tested without committing yet another load Statistical quality control can be very helpful in batch or short run type (set-up dominated) processes by using it to analyze set-up variables If the process is then set up to optimal set-up parameters (as determined by experimentation or evaluation of part outputs), meeting parts specifications will necessarily result

Statistical Process Control. The idea behind true statistical process control is that the results of a process can be guaranteed if none of the relevant process parameters are allowed to stray outside of previously established control limits

The long standing problem in applying statistical process control to heat treatment has been finding methods to quantify and measure process parameters that are of known importance (outside of the obvious ones) Many SPC programs are based upon charting controlled parameters such as temperature, atmosphere carbon potential, quenchant temperature, and

so on While this approach is certainly not incorrect, it does often lead to a situation where a deviation in an SQC chart (results) commonly cannot be attributed to any special cause deviation in a corresponding SPC chart (processing parameters), because all the things being charted are controlled variables that by design will not normally change

Processing Methods Considerations

Repeatability is a key issue when considering how to improve a heat treatment process The more process variables that can be controlled within specific known limits from part to part, furnace load to furnace load, and day to day, the more repeatable the process results will be

Continuous Operations. The continuous types of heat treatment equipment (that is, rotary retort, pusher carburizers, belt furnaces, and so on) offer the most straightforward approach to applying SQC and SPC techniques to improve process performance

Because a high volume of work pieces is involved, there is adequate opportunity to perform in-process sampling of key product characteristics Negative outcomes can be predicted before they take their full course Also, special causes are often more identifiable because process variables are steadier in continuous processes than in batch-type processes

Single Part Treatments. With induction and flame heat treating, parts are typically processed one at a time Using part evaluation techniques to predict negative results becomes difficult and impractical Thus, the focus must shift to statistical process control and the identification, monitoring, and controlling of the process variables to ensure repeatability of the results

Electric power, flame temperature, scan speed, coil dimension, part positioning, and quenchant temperature are some variables that need to be considered Trending of process variables can be used to determine special causes

Batch Operations. Batch-type furnaces usually offer the opportunity to do a significant amount of sampling and analysis within a load However, all this does is develop a degree of confidence on the results of the entire load The process variables must be monitored and analyzed to ensure that the process is under control and there is repeatability from load to load and from day to day This is especially true when each load is different in terms of part geometry, material and/or specification, which is the norm in commercial heat treating

Process Deterioration

A fact of life in any heat-treating process is that the equipment gradually succumbs to the wear and tear of constant operation, thus the process inevitably gets worse with time The challenge is to counter this natural deterioration with corrective action before out-of-specification parts are produced

SPC techniques can be utilized to measure furnace performance and address process deterioration in heat treating By monitoring key process variables and/or key process outputs, preferably in on-line fashion, trends can be spotted and action taken before nonconforming product is produced

Key process variables may mean not only controllable variables, but also uncontrollable and secondary variables

Uncontrollable Process Variables. Examples of uncontrollable variables that are useful in the monitoring of a treating process are:

heat-• Quench transfer time: Although this is not a controllable variable in most furnace systems, it is often a

critical parameter in terms of producing good parts It may be beneficial to monitor and analyze transfer time in order to get an early warning of the deterioration of the transfer mechanism One method for verification of sufficiently fast transfer time is to compare the maximum allowable transfer time for a successful process to the actual transfer time which would trigger an alarm if the maximum were exceeded This automatic control method would then flag suspect loads or parts resulting from either a mechanical failure or equipment function deterioration

• Temperature recovery time: By measuring and analyzing the time it takes for a batch furnace to reach

setpoint temperature (with a standard load weight or empty), trends can be picked up that would indicate

a loss in furnace performance These out of control trends as plotted on an SPC chart would then trigger

an investigation to determine the cause of the condition (for example, damaged insulation, poor door seals, heating system malfunction, and so on)

• Quench temperature rise: Although a quench system may be controlled within a specific range (that is,

30 to 65 °C, or 90 to 150 °F), it may be important to know how the temperature cycles from quench to quench This would give a macroanalysis of the entire quenching system and give warnings of failed, or impaired, agitation and/or quenchant cooling Over-loading of the furnace could also be indicated

Secondary process variables are those that are caused by the deterioration of control loops Examples are of secondary variables:

• Fuel consumption: By monitoring gas or electric consumption for a standardized furnace cycle and

loading (could be empty), diminished performance in the heating system can be detected

• Additive atmosphere gas: By monitoring and trending the amount of natural gas (or propane) addition

required to control a specific carbon potential setpoint, the deterioration of furnace atmosphere integrity can be detected

Process and Product Capabilities

Capability studies are conducted on all types of manufacturing processes to determine the statistical variation of a product with respect to a measured characteristic For heat-treating processes, characteristics frequently measured are hardness and case depth Because these metallurgical characteristics are sometimes difficult to define, specifications may initially need to be clarified with regard to the exact test scales or test methods to be used and the critical locations where these tests are to be made before a capability study is conducted Process results for many metallurgical and heat-treating processes are dependent on material-related characteristics such as hardenability, material chemistry, and/or part geometry that also make the process test results sensitive to those variables

After the metallurgical requirements are clearly established, a basic process capability study may be conducted Care should be taken so that the parts tested are from the loading locations representing the extremes in process variability A good guideline for test sample locations is to use those loading locations prescribed for temperature uniformity surveys in specification MIL-H-6875

For continuous processes, it is important to collect the samples over a sufficiently long period of time in order to reflect process heating power fluctuations or other process abnormalities that could be time dependent

The use of normal probability paper for data representation and plotting is highly recommended If the data does not plot

as a straight line indicating a normal distribution, a metallurgical or process-related reason for this skewness should be apparent or be determined An example of a capability study of an atmosphere harden and temper operation for automotive seat belt parts made from SAE 4037 steel is shown in Fig 1 and Fig 2

Fig 1 Final hardness distribution analysis for a typical quench and temper operation

Fig 2 Normal probability plot of data from Fig 1 (a) Frequency distribution (b) Distribution analysis sheet

Specification, mean = 35 HRC; range = 7 HRC Results, mean = 35.7 HRC, 6σ = 5.5 HRC, Cp = 1.27 Action, adjust temper to adjust mean to 35 HRC

As can be seen in Fig 3, the overall process capabilities results are the result of many contributing factors:

• Base material contributions: Unique material characteristics, material defects, and hardenability

differences These can vary from lot to lot and also between materials

• Part-related contributions: Part geometry and section size variations

• Process-related contributions: Temperature uniformity as affected by process control and mass effects,

time control, atmosphere control, and cooling method (as determined by uniformity and average severity)

• Evaluation method contribution: Standards accuracy and testing method accuracy

Fig 3 Factors contributing to overall heat-treating process result variations

Thus, to successfully use the process capability study as a dynamic tool to refine and narrow process variability, the following three steps should be used in conjunction with process capability studies:

Step No 1:

• Identify critical control variables and their relative contributions to process attribute variations (this can

be done by process modeling techniques)

• Measure process inputs with corresponding process output results

• Document process control procedures

Step No 2:

• Modify control procedures, manufacturing procedures, or equipment in order to reduce process variability

Step No 3:

• Remeasure process capability (as in Step No 1 above) to ascertain the effectiveness of the changes

The overall heat-treating process variability result may be characterized as being comprised of the following factors (they may be classified into four categories) and the accompanying sources of the undesirable background signals (intrinsic or extrinsic noise):

• Base material related (intrinsic noise)

• Part configuration and manufacture related (intrinsic noise)

• Process related (extrinsic noise)

• Evaluation method related (extrinsic noise)

By using properly standardized test coupons as the basis of a process capability study, we can separate out the variability due to intrinsic noise factors and arrive at the inherent process capability However, in practice, we will still have these contributions in the process and this should be kept in mind Additionally, a GR & R study may be performed on the evaluation method to determine the contribution of these factors to variability

Base Material Considerations

Cast irons are probably the best example of a material where test results (that is, hardness) can be a function of the hardness testing scale used This sensitivity of hardness value to the testing method and the hardness scale used is because the different phases present in the workpiece vary significantly in hardness This same effect exists in other materials which are heat treated (see Fig 4)

Fig 4 Microstructure of hot-rolled AISI 1022 steel showing severe banding Bands of pearlite (dark) and ferrite

were caused by segregation of carbon and other elements during solidification and later decomposition of austenite Etched in nital 250×

Another type of problem that can influence testing results which are not the direct result of processing is "banding." Many steels, particularly a resulfurized one such as AISI 1100 or 1200 series, exhibit banding or microalloy segregation The bands exist prior to heat treatment and the ferrite-rich and pearlite-rich areas run in bands across the longitudinal rolling direction of the bar stock from which parts are made

It has been found that this condition can result in a 4 to 10 point of Rockwell C hardness variation after hardening between these bands of different chemical composition This problem is greatest when the bands are widest and the heat treatment times are very short, such as for induction hardening processes

Decarburization. Surface carbon reduction to a greater or lesser degree exists on most steels having more than 0.30%

C This defect results from basic steel manufacturing and if not removed in the part manufacturing process prior to heat treatment can influence the surface hardness of parts after induction, flame, or direct hardening processes that may not be capable of correcting the surface decarburization condition However, it should be recognized that many heat-treating processes can also cause this same problem It is thus important for one to have characterized the incoming product to be processed so that the controllable incoming material variability can be isolated and corrected independently from the product variations due to the process

Material Variations. Before applying statistical control techniques to monitor process or product uniformity, it is important to understand how the raw material uniformity is controlled prior to heat treat processing That is, whether or not the incoming material is identified and kept separate by heat numbers in the case of steel or by batch number in the case of cast materials

• Properly selected for size, shape, and material that can be directly correlated to the material and parts configuration being processed

• Prepared in sufficient quantity (same heat of steel) and quality to eliminate or minimize the material uniformity variable from the processing variation

By using statistical quality control with test coupons in conjunction with statistical quality control on heat treated parts, product variations attributed to process only variations may be identified and controlled

Example 1: Use of 10 000 Test Pins Measuring 64 mm (2 1

2 in.) Long by up to 17.8 mm (0.700 in.) OD Made from a Single Heat of 8620H Steel Used to Monitor the Carburizing and Hardening Operations of 5- to 8-Pitch Gears

Test pins were used to monitor carburizing and hardening processing for 5- to 8-pitch gears made mostly from 8620H

steel This procedure is used to monitor the process variation in carburizing of surface hardness, effective case depth, and

core hardness The diameter chosen for the test pin is based on the gear tooth thickness and the fact that the test pin center cooling rate would be on the steeper portion of the Jominy hardenability curve This means that monitoring the center core hardness on test pins is an indirect measurement of quench uniformity

Purchase and Processing of Test Pins. The minimum quantity of test pins purchased was 10 000 pieces from a single heat of 8620H The OD of these pins were ±0.13 mm (±0.005 in.) for a given lot of pins with the absolute size being 12.7 to 17.8 mm (0.500 to 0.700 in.) The length was 64 ± 1.6 mm (2 1

The test pins are evaluated as-quenched only No tempering is permitted Test pins evaluated for purposes of SQC control

were from pure cycles with no abnormal changes in times, temperatures, or quench procedure

Testing Procedures for Test Pins. File the surface of the pin to check for file hardness and to make a smooth surface Check three hardness readings on Rockwell C scale and record the average Do not use V anvils but use flat or spot anvil only

Cut parallel section 6.4 mm (1

4 in.) thick from the center of the test pin Set the diamond and anvil by checking at radius Check the center hardness by Rockwell well C scale and record

mid-Effective Case Depth. On the section cut from above, grind the surface to be checked on a 120-grit or finer paper Test

from the surface in to the point where the hardness is 85.5 HR15N (50 HRC) Measure from the surface to the center of that mark using a Brinell glass Record the reading as effective case depth in thousandths of an inch

The referee method for checking effective case depth is by 500-g microhardness to 50 HRC equivalent Therefore, at least one of every ten checks and/or any check of effective case depth not within the specified limits is to be verified by the microhardness method

The results from above are to be plotted by cycle and furnace on the form shown in Fig 5

Fig 5 Chart for plotting 8620 steel test pin variation by characteristic (a) Effective case depth characteristic

(b) Surface hardness characteristic (c) Center core hardness characteristic (d) Chart to plot data from (a), (b), and (c) by cycle and furnace

This method can be started and used on a monitoring basis only for a short time until mean values with upper and lower control limits can be established

Process-Related Contributions. This is the most important characteristic to identify and control to reduce variability

of the heat treated products Exactly how to accurately control process parameters is covered in the section "Design of

Experiments" in this article By using standardized test pins and modeling to separate processing parameters, the individual parameter contribution to a measured characteristic such as effective case depth can be shown as detailed in Table 1

Table 1 Contribution of selected parameters to variations in effective case depth for required 0.85 to 1.00% surface carbon level at 870 °C (1600 °F) processing temperature

Variation in case depth for selected parameters, % (a)

Carbon variation (∆ C)

Case depth

Temperature variation (∆T) Time variation (∆t)

Atmosphere Quench uniformity (b)

mm in 11 °C

(20 °F)

28 °C (50 °F)

56 °C (100 °F)

5 min 10 min 30 min 0.10% 0.15% 0.25% 0.05% 0.10% 0.20%

(b) Variation in case carbon level when quenched to 50 HRC

The most significant observation from Table 1 is that quench uniformity is an equally significant factor in the carburizing process with time, temperature, and atmosphere control as variables

hardness units, there is the likelihood that there would be other 0.4 or greater measurement variations When plotting -R control charts (see the article "Statistical Quality Design and Control" in Volume 17 of ASM Handbook, formerly 9th Edition Metals Handbook) where is the sample mean and R is the range, this greater test block variation over time

could erroneously show the hardness tester is not consistently accurate within tight limits The variation could lead to incorrect process adjustments or put into question the process capability The higher consistency, more stable test block would show a tighter band of tester accuracy and repeatability performance

In order to assist in reducing the variability in hardness testing, the following items should be utilized:

• Digital hardness tester with reduced operator influence features (that is, motor drive or automatic full load application and removal)

• Very stable and consistent hardness test blocks

• Certified diamond indenters as reference standards

• Good operator training program on hardness testers

• Proper clamping devices to support odd-shaped parts and/or fixtures for additional support

• Daily checks of hardness testing equipment using standardized test blocks in the range or ranges most commonly encountered on production parts tested with the machines

SPC Process Analysis

The concept of process analysis is a relatively new way to deal with a group of old ideas For example, what is the best temperature from which to harden a particular steel? How was this temperature determined? Is this temperature the one that takes the best advantage of the material's hardenability or is it the one that gives the lowest distortion or was it chosen for other reasons? What effect will raising or lowering this temperature have on the final variation in results?

Process analysis should follow a step-by-step characterization plan as follows:

1 Determine the process capability

2 Describe the process in operational terms

3 List, in order of importance, the output characteristics for each process element

4 Decide upon the measurement method and determine instrument variability (GR & R study)

5 Estimate the process variability

6 List all input variables (both controllable and uncontrollable) for the above output characteristics

7 Select the variables to be included in the process characterization

8 Decide the objective of the process characterization

9 Determine characterization methods and how data will be displayed

10 Construct a characterization implementation plan

11 Choose analysis method for process optimization (that is, evolutionary operations, EVOP, response surface methodology, RSM, and so on)

12 Determine effective process control plan

13 New process capability determination

For example, in the hardening of steel, the variables of quenchant selection and quenchant temperature are vitally imporant Common sense (as well as heat transfer] logic) says that the higher the quenchant temperature, the slower the quench Unfortunately, this rule may or may not be true depending on the quenchant and the temperature range

Many quench oils show peaks in their cooling characteristics that dictate a certain quenchant temperature is required for the fastest quench Going either above or below this temperature will slow it back down To complicate the matter, this performance peak may change depending upon aging of the oil and the agitation level of the oil For all quenchants, agitation is a very significant factor in cooling performance Thus, picking the optimum quench is not an easy task The category of methods developed for dealing with such complexity is known as the design of experiments These methods must not be confused with simple linear regression interpolation from empirical data

For example, consider the fictitious empirical data on the hardening of a high-carbon steel part in Table 2

Table 2 Determining optimum austenitizing-tempering cycle to obtain a 55 HRC hardness in a high-carbon steel part

370 °C (700 °F)

A design viewpoint on these data would help the selection of the best combination from a result variability viewpoint It can be seen that the selection of 900 °C (1650 °F) for austenitizing along with perhaps a 315 °C (600 °F) temper will center the results at around 55 HRC, just as an austenitizing-temper combination of 790 °C (1450 °F) and 260 °C (500 °F) would The choice of which sequence is best to reduce variability in the desired hardness might be governed by two observations:

• It would appear from the data that the higher the austenitizing temperature, the wider the range of the measured hardness results

• The effect on hardness of increases in tempering temperatures seems more dramatic from 260 to 370 °C (500 to 700 °F) than from 150 to 260 °C (300 to 500 °F)

Based on these observations, one might conclude that the 790 °C/260 °C (1450 °F/500 °F) combination is more desirable than the 900 °C/315 °C (1650 °F/600 °F) combination, as the results will likely fall into a tighter range, that is, resulting

in a greater process capability around 55 HRC

The key goals in a design of experiments exercise in heat treating are generally two-fold:

• Determine which variables (and combinations of variables) have significant, observable influence over the results based on a balance between economic and quality considerations

• Determine the optimum values for these variables in order to center the process at the desired point, reduce the variation in results due to possible or expected variation in these process parameters, and reduce the variation in results due to variation in the material

The following case history is presented to illustrate a design of experiments exercise

Example 2: Induction Hardening of an AISI 1040 Cold-Drawn Combination Yoke and Spline Shaft Bar

The bar had a 34.11 mm (1.343 in.) diameter and was 900.68 mm (35.460 in.) in length

These parts were being hardened on a 6-station scanner with a 10 kHz motor generator (800 V, 375 A, 300 kW) The induction coil size was 44.45 mm (1.75 in.)

The stated objective of the design exercise was to determine the relationship between various induction hardening parameters and the resultant metallurgical properties (including distortion) The ultimate goal of the project was to optimize the process to produce a part with acceptable metallurgical properties while exhibiting minimum distortion Table 3 summarizes the various characteristics desired in the final part

Table 3 Desired properties of a 34.11 mm (1.343 in.) OD cold drawn combination yoke-spline 1040 steel shaft

Optimum condition Characteristic Preferred dimension

Effective case depth Nominal = 4.83 mm (0.190 in.) 40 HRC

Surface hardness Maximum 58 HRC 1040 steel

Spline size change Minimum 0.0000 mm (0.0000 in.) Absolute value

Out-of-straightness (TIR) Minimum 0.000 mm (0.000 in.)

The factors affecting the process were split into two groups:

• Process factors: Those factors that could be changed (for example, speed and power)

• Noise factors: Those factors, such as the induction station number, that cannot be changed One very

important variable that could not be changed during the experiment was the analysis of the individual heats of steel involved

The practical limits on the process factors and noise factors (that is, factors that could not be controlled) were determined and cataloged in Table 4 and Table 5

Table 4 Process factors for three heats to induction harden a 34.11 mm (1.343 in.) yoke-spline 1040 steel shaft per Table 3 specifications

Process conditions Process factor

Heat 1 Heat 2 Heat 3

Scan speed, s/m (spf)(a) 110 (36) 120 (40)

Power supply, %(b) 98 94 90

Quench temperature, °C (°F) 25 (80) 40 (100) 50 (120)

Quench pressure, kPa (psi) 55 (8) 70 (10) 83 (12)

Speed of rotation, rev/min 10 36 60

(a) spf, seconds per foot

(b) At 100% set point, power output is 800 V

Table 5 Noise factors for three heats to induction harden a 34.11 mm (1.343 in.) yoke-spline 1040 steel shaft per Table 3 specifications

Composition, % Degree of hardenability Steel making process Heat No

C Mn Ni Cr Mo Si

Chemical ideal

diameter, DI(a)

low Ingot cast, aluminum fine grain M1 N26177 0.37 0.71 0.02 0.04 0.01 0.23 0.88

mid Billet cast, vanadium fine grain M2 B944212 0.39 0.85 0.05 0.07 0.01 0.27 1.16

(a) Calculated hardenability expressed in inches

The following series of experiments was designed to test the possible various variably combinations The individual experiments were referred to as A1, C6, and so on (see Table 6)

Table 6 Series of experiments run to test process factor variable combinations shown in Table 4

A scan speed

B Power

C Quench temperature

D Quench pressure Experimental level

spf (a) s/m Setting, % Voltage, V °C °F kPa psi

E Speed of rotation, rev/min

1 110 36 98 785 25 80 55 8 10

2 110 36 98 785 40 100 70 10 36

3 110 36 98 785 50 120 83 12 60

(a) spf, seconds per foot

The experiments were run and the results (characteristics) measured and documented The results were then analyzed mathematically using the concept of signal-to-noise ratio, or S/N

(This concept is somewhat difficult to grasp as it relates to process control but it is helpful to consider the S/N phenomenon in the case of an overseas telephone call There is almost always some hiss in the background of such a

connection The voices on the line are the signal, while the hissing sound is the noise If the signal is much louder than the noise, it is possible to understand what is being said This is a high S/N ratio On the other hand, if the noise is almost as loud as the signal, it is not possible to understand; this is designated as a low S/N ratio.)

S/N ratios are measured in decibels, or dB This is a logarithmic unit that allows very large ratios to be compressed into small units In the case of process control:

• Signal: Controllable input of the process

• Noise: Uncontrollable variables which cause variation of the process

There are many specific formulas for calculating S/N in process control situations Each is of the general form:

(Eq 1)

where σ is the standard deviation Note that:

• A higher S/N ratio indicates a variable (or factor) with a corresponding large effect on the output characteristic of the part

• S/N measures change in both mean and/or variability

• Every time the S/N gains 3 dB, the loss by the loss function halves; thus a 3-dB change is significant

Tables 7(a) and 7(b) show the results of the mathematics as related to the characteristic of case depth A depth of 4.83 mm (0.190 in.) was considered nominal

Table 7(a) Signal-to-noise ratio analysis used to determine optimum process and noise factors required to obtain 4.83 mm (0.190 in.) average case depth in yoke-spline shaft

Noise factor

Average case depth

Material hardenability

Station mm in

to- noise ratio (S/N),

Signal-dB

A1 110 36

4.83 0.190 131

A2 120 40

5.31 0.209 119

B1 98 785

5.49 0.216 79

B2 94 750

5.13 0.202 90

B3 90 720

4.57 0.180 81

C1 25 80

5.18 0.204 85

C2 40 100

5.11 0.201 92

C3 50 120

4.90 0.193 84

M2

Mid 5.66 0.223

M3

High 5.79 0.228

S1

No 1 4.95 0.195

S2

No 2 4.98 0.196

S3

No 3 5.11 0.201

S4

No 4 5.11 0.201

S5 No 5 5.11 0.201

S6

No 6 5.18 0.204

Table 7(b) Cumulative contributions to variability attributed to the individual characteristic and noise ratio values listed in Table 7(a)

signal-to-Characteristic Variability

attributed to individual characteristic, %

Contribution due to signal-to-noise ratio,

%

Process factors:

Scan speed 4.7 29.0

Power supply 11.2 40.0

Quench temperature Insignificant Insignificant

Quench pressure Insignificant Insignificant

Speed of rotation Insignificant Insignificant

Noise factors:

Material hardenability 72.0

Station Insignificant

The S/N ratios were used to pinpoint the optimum process factors to produce the desired 4.83 mm (0.190 in.) case depth

As listed in Table 6, these were (by experiment number):

Experiment No Process factor

A1 Scan speed: 110 s/m (36 spf)

B2 Power supply: 94% (750V)

C1 Quench temperature: 25 °C (80 °F)

D2 Quench pressure: 70 kPa (10 psi)

The following general guidelines (applicable for a 1040 steel having midrange hardenability only) for case depth were derived from the experiment

Process factors include:

• Scan speed: A 13 s/m (4 spf) speed change will change the average case depth by 0.48 mm (0.019 in.)

• Power: An 8% change in voltage will change the average case depth by 0.91 mm (0.036 in.)

• Quench temperature: A 25 °C (40 °F) temperature change will change the average case depth by 0.28

mm (0.011 in.)

• Quench pressure: Pressure in the range of 55 to 83 kPa (8 to 12 psi) has no effect on the case depth

• Rotation speed: No effect on case depth in the 10 to 60 rev/min range

Noise factors include:

• Material hardenability, the major factor (72%) affecting the case depth

• Case depth, which varied little from station to station

Additional Factors. Similar analyses were carried out for the other characteristics of surface hardness, distortion, and scale Some of the observations are:

• The severity of the thermal cycle to which the shaft was subjected greatly influenced the spline size change For example, shafts heated with the lowest heat 120 s/m × 90% V input (or 36 spf × 90% V) in combination with the slowest quench rate (50 °C × 55 kPa, or 120 °F × 8 psi), had significantly less spline distortion than the shafts heated with the maximum heat input, and quenched at the fastest rate

• The theory that a deeper case depth produces greater size change was not proven in this experiment (see

Table 8)

• The material contribution to the spline size change was 53%, possibly due to hardenability differences,

or to differing amounts of cold work in the various bars

• Analysis of total indicator runout (TIR) of the parts gave the same conclusions as those reached with spline size change

• The minimum scaling was achieved with the fastest scan times (less time exposed to oxygen at high temperature)

• Surface hardness was not significantly affected by the range of process factors allowed in the experiment

Table 8 Effect of average case depth on spline size distortion

Power supply Quench

temperature

Quench pressure

Effective case depth 110 36

C1 25 80

E2 30

(a) TIR, total indicator runout

(b) Underline denotes best signal-to-noise ratio for that variable

Table 9, along with human interpretation, provides the information required to make a decision as to the best overall choice of a parameter combination, as follows:

Experiment No Process factor

A1 Scan speed: 110 s/m (36 spf)

B2 Power supply: 94% (750 V)

C3 Quench temperature: 50 °C (120 °F)

D2 Quench pressure: 70 kPa (10 psi)

E2 Speed of rotation: 36 rev/min

It should be noted that a verification test is necessary to determine whether the predicted outcome will be attained If the outcome of the verification does not meet the expected improvement, then, it must be concluded that the experiment was unsuccessful in locating the optimum parameter settings or that the dominant parameters were not located

The final results after optimum parameter selection are shown in Table 10

Table 10 Optimum characteristic parameters as determined by signal-to-noise analysis of a yoke-spline shaft

Average 4.78 0.188 56.3 0.02 0.0008 2.97 0.117 648 0.0229

Goal 4.83 0.190 58 0.00 0.0000 0.000 0.000 0 0

(a) TIR, total indicator runout

Conclusions. The conclusions that were drawn at the end of the exercise were:

• The relationships between the induction hardening process parameters and the resultant characteristics can be successfully established by the design of experiments method

• Both metallurgical properties and dimensional changes can be controlled to some degree in the induction hardening process

• The amount of prior cold working and/or hardenability of the 1040 steel bar is the most significant factor affecting variations in case depth (72%), surface hardness (18%), spline size change (53%), and TIR (12%)

The following recommendations resulted from this SPC analysis:

• The hardenability of 1040 steel should be controlled to an ideal diameter range of 24.4 to 32.0 mm (0.96

to 1.26 in.)

• The cold working operators should review methods of controlling and reducing residual stresses

Example 2 shows the general application of design of experiments techniques to locate and tune the vital process parameters Other areas where this technique has been successfully utilized to improve process reliability include press quenching of gears, gas carburizing, and quenching of aluminum

Monitor/Control Decisions

There are many process parameters of interest in almost any heat-treating operation These fall into three categories:

• Those variables which must be closely controlled in order to have the process occur (for example,

temperature, time, and so on)

• Variables that are known to affect the process but are either ignored or simply monitored manually because of the complexity or cost involved

• Variables which cannot be controlled at all (for example, natural gas composition, available heat input, and so on)

On a batch atmosphere carburizing furnace, for example, the common variables that can affect the carburizing process and should be compiled are:

• Temperature

• Atmosphere carbon potential (see the article "Control of Surface Carbon Content in Heat Treating of Steel" in this Volume)

• Time

• Endothermic gas flow (see the article "Furnace Atmospheres" in this Volume)

• Enriching gas flow

• Atmosphere circulation

• Quenchant properties (temperature and agitation)

While the first two on the list, temperature and carbon potential, will commonly be automatically controlled and the third, time, will usually be held fairly closely by either automatic or manual means, the last four on the list often go without

notice until a problem develops, at which time they are examined

In keeping with the idea that statistical process control represents preventative action (in that out-of-control indications

should arise before out-of-tolerance conditions arise), these other critical variables should obviously be given more attention, possibly on an automatic, ongoing basis

The decision to control, monitor, or ignore a given variable is a decision that must be made by taking into account the importance of the variable and the ease with which it can be measured This decision is ultimately a balance of economic and quality considerations The diagram in Fig 6 shows a more complete picture of the number of possible variables involved

Fig 6 Identification of heat-treating variables for the neutral hardening process

The subject of sensors and sensor technology becomes important at this point

The focus of the data acquisition problem becomes one of finding sensors and sensor technology suitable for use in the heat-treating environment It turns out that this is indeed a significant problem area for a variety of reasons that will be outlined later In general, there are five critical areas of concern:

• Sensor selection

• Signal conditioning

• Control using the sensor

• Sensor maintenance

• Documentation and training

Sensor Selection. A sensor can be anything from a limit switch to a thermocouple to a pressure transducer to an electronic flowmeter

Proper sensor selection begins with the identification of a process variable worthy of measurement The cost of reliable measurement versus the importance of the process variable must be evaluated closely Temperature happens to be a process variable that is usually extremely important, and a significant investment is usually made to ensure the accuracy

of its measurement

Time is also usually a critical variable, but ironically, it is still common practice to see it measured by a wall clock subject

to an operator's convenience

The signal conditioning process encompasses all those things that go into making the raw sensor data into a valuable bit

of stored data For example, the tripping of a limit switch may mean nothing by itself, but the time between two consecutive trips might be an extremely valuable piece of data

Sensor maintenance has to be a major consideration It does no good to put sensors on a furnace only to find that false readings are constantly coming up because of dirt build-up or other problems Many times it is necessary to spend much more money on sensor insurance systems than on the sensors themselves For example, an oxygen probe that constantly accumulates excessive soot in a high-carbon atmosphere may need an elaborate carbon burn-off system added to it just to get consistent and reliable readings

Documentation and training are key elements because they determine the long-term success rate of sensor usage The operators and maintenance people must be aware of the theory of operation and the limitations of the equipment

Temperature Parameters. Temperature measurement and control is the most fundamental measurement in heat treatment and deserves separate consideration in this discussion

Temperature Measurement. Temperature in the ranges of interest in heat treatment are generally measured by one

of two methods: thermocouple or infrared pyrometer

Because of its low cost, simplicity of construction, and inherent reliability, the thermocouple has always been and continues to be by far the most useful sensor in most situations It is perfect for sensing gas temperatures, and even works well in vacuum furnaces by virtue of radiation They are not very useful, however, for measurement of the actual temperature of parts going through a furnace

Because thermocouples often fail slowly by losing accuracy and because two thermocouples inserted at the same time will often agree even as they are both failing, it is important to change thermocouples on a regular schedule and to alternate replacements

Aside from the obvious process-related temperatures that are measured, it is possible that the temperature of nonprocess related items like cooling water temperature to bearings and door seals might be of interest Thermocouples are most often used for these purposes

The science of infrared temperature measurement has come a long way in the last 10 years Most of this type of equipment is used in induction hardening but many have been applied with mixed results in enclosed furnaces In these situations, results are usually better if the furnace does not use a hydrocarbon atmosphere, as the presence of soot in the optical path presents a problem

Furnace temperature uniformity is always of tremendous concern, especially in vacuum applications where there is

no convection to help even things out While the process of making a furnace uniform with respect to temperature may be difficult, the actual uniformity results are easily documented with multiple thermocouples

It is not at all out of the question to monitor uniformity by permanently placing thermocouples in several furnace locations and automatically calculating and alarming against their average and spread This approach has been used for years with vacuum furnaces (where uniformity becomes more of a function of loading)

The observation of temperature measurements in a furnace under control gives some insight into how this might be handled from an SPC standpoint (see Fig 7), where a control thermocouple is being held very closely to a setpoint (for a

Cpk > 10), while a companion thermocouple in a different location is showing a Cpk = 1.2 The controlled thermocouple is

of very little interest, as it will probably not deviate from the setpoint unless there is some drastic change The subtle information from the monitoring thermocouple is of much more value in evaluating uniformity and the repeatability of uniformity which can be another problem

Fig 7 Comparison of instrument temperature variations (a) Controlling thermocouple, Cpk > 10 (b) Monitoring

thermocouple, Cpk = 1.2

Combustion Efficiency. Because gas-fired burners have historically been the least expensive device to heat most nonvacuum furnaces, there has always been interest in attempting to keep them firing at optimum ratio This was previously done using portable oxygen analyzers or with manual carbon dioxide checks Most recently, interest in using low-cost automotive-style oxygen sensors has become possible

Although there is nothing to prevent the relative measurement of burner efficiency from day to day with such techniques

in an inexpensive format, it is very difficult to measure on an absolute scale that the burner is actually always firing at optimum ratio This situation is the result of problems with sensor positioning and interpretation of results For example,

in a high-low fire system on a radiant tube in which the sensor is positioned opposite from the burner, the sensing system would have to ignore the first few seconds of high fire as the old exhaust gas is blown out If the high fire cycle was very short, the sensor readings would mean very little If the cycle is longer, the temperature of the sensor would change rapidly, making it necessary to provide a compensating thermocouple to make the sensor output meaningful This complication requires a relatively high-cost measurement system to utilize the low-cost sensor

Furnace Chamber Atmosphere. Measurement of various parameters associated with gases in a furnace chamber are

of particular importance for controlling many heat-treating processes

Gas Pressure Level. The manual measurement of pressure in a furnace atmosphere is usually done with a water

manometer, with levels generally in the range of 0 to 25 mm (0 to 1 in.) water column The reading is actually a differential reading between the inside and outside of the furnace Inexpensive pressure transducers that produce high-level signals (for example, 4 to 20 mA) for this range are readily available, but care must be used to ensure lone-term reliability of this signal due to plugging or partial closing of tubing to the transducer

Transducers which feature an analog display of the pressure reading as well as the retransmission signal are most useful when the signal must be trimmed for calibration The transducer should have its own zero and span adjustments for maximum flexibility

Furnaces in which the pressure is constantly varying over a wide range because of opening and closing doors or other upsets will require intelligent signal conditioning that ignores the peaks and valleys in the pressure

Atmosphere pressure is sometimes controlled in a closed loop with an actuator-driven atmosphere effluent damper This device also helps to keep the atmosphere pressure constant

Vacuum Level. Measurement of vacuum level has been done with a variety of methods, but two of the most common have been the thermocouple gage for lower vacuum levels and the cold cathode gage for higher vacuum ranges

Recently a fourth method has been added for measurement all the way down to 13 nPa (10-10 torr), based on a hot filament ionizing the residual gas All methods except capacitance manometer are subject to error of the composition if the residual gas is different than expected

A typical vacuum furnace will generally require at least two sensors to accurately cover the full range that the furnace is capable of operating within Also, it is often desirable to check vacuum levels in various spots, including the foreline and vacuum chamber itself

A microprocessor-based instrument that will automatically select the required sensor in the range of interest is required These newer instruments also provide computer communications capability that makes data acquisition easy

It is anticipated that a version of the zirconia oxygen sensor, discussed in the section "Analysis of Gas Composition" in this article may be applied successfully in vacuum furnaces in the future This would represent a major step forward in vacuum technology, as low oxygen levels are desirable when high vacuums are used The oxygen probe would then provide a direct measurement of this gas

Additional information is available in the article "Heat Treating in Vacuum Furnaces and Auxiliary Equipment" in this Volume

Process Gas Flows. Measurement of process gas flows in controlled atmosphere furnaces generated interest recently There are several methods available, each varying widely in cost The selection of a particular method must be made with extreme care because the cost of gas flow measurement must be kept in balance with the other data acquisition elements

in the overall instrumentation plan

The least expensive method to measure gas flows is with simple flow switch devices These devices almost always operate on the principle of a pressure drop (and pressure switch) across a fixed orifice in the gas stream Their limitation

is that they can only sense whether flow above a certain fixed amount is present or absent However, by combining two flow switches in the gas stream set at different points, it is possible to ascertain whether a gas flow is above the high limit,

in the desired range, or below the low limit

Flow switch arrangements are especially suited for gas flows that do not vary (for example, endothermic or exothermic generated gases and nitrogen) or even liquids like methanol

Typical plumbing component costs for a dual flow switch arrangement might be in the $100 to $200 range for a single gas line Control strategy using this technique is to have the data acquisition system sound an alarm whenever the gas flow has dropped out of the desired range, with human intervention required to correct the flow

If an exact value of gas flow is required, there are two techniques:

• Electronic true mass flowmeters

• More familiar rotameter types with electronic adaptors

Both of these may require external power supplies and care in system wiring

The true mass technique uses the principle of measuring the amount of heat that the gas stream can remove from a heated bulb of controlled temperature This technique has the advantage of being inherently accurate (if the flowmeter has been calibrated against a known flow) regardless of pressure fluctuations It has the disadvantage of being a blind technique, with the electronic signal (typically 4 to 20 mA) being the only indication of flow

The rotameter technique, while slightly more expensive than the true mass technique (approximately $1500.00 per meter), has the advantage of having two outputs: a visual one by virtue of the float scale (see Fig 8) and an electronic one by virtue of electronic position sensing of the same float assembly The disadvantages of the rotameter technique lie in resolution and accuracy; the measurement is sensitive to temperature and pressure variations (good upstream regulation is required), and the electronic signal may only resolve to ±2.5% of scale However, this technique is still preferred because

of long-term calibration considerations

Fig 8 Schematic showing key visual output components of a rotameter

Measurement of low liquid flows, such as in nitrogen-methanol systems, is best done with the rotameter system However, the viscosity of methanol changes quickly with temperature, and for maximum accuracy it is necessary to electronically compensate the temperature readings from a rotameter-type device accordingly

Control of gas flows is usually done with an adjustable port valve and motor actuator The most easily installed and maintained motor actuator is one that accepts an electronic signal (for example, 4 to 20 mA) and positions itself accordingly Motors that employ a slidewire feedback technique are not always desirable because of long-term maintenance problems

One problem with motor-actuated control valves is that the flow is not proportional to the value position For example, it

is possible to have signals of 0% open and 100% open do just what is expected, while finding that the 50% open signal gives a measured gas flow of 20% of scale This is due to a variety of factors, including linkage adjustment, pressure drops in the piping, and the inherent nonlinearity of valve ports

Corrosive gases such as ammonia require stainless steel construction that makes these valves more expensive

Another completely different technique of measuring and controlling gas flow that can be used in many cases is the pulse time-proportioning system For example, if a simple on-off valve is set up in such a way that when the valve is open there

is a fixed, known flow through it, then the average gas flow is always easily calculated by the formula:

(Eq 2)

If the ON and OFF time are automatically varied and kept short with respect to the furnace size, this type of arrangement will provide a most cost-effective, easily calibrated system for measuring and controlling gas flows The power to perform ON and OFF time cycle control is easily found in many of the most powerful process controllers on the market today In the case of two process gases that must stay in ratio with each other (for example, nitrogen-methanol), it is possible to design an electronic control system to measure both process gases and control one or both of them to maintain constant ratio, using the techniques outlined above The cost of such a system can be very high, but varying methanol flow is a common problem that can significantly impact process results

Analysis of Gas Composition. The following measurement techniques for the composition of furnace atmospheres have successfully been used:

• Oxygen probe: Good for measurement of oxygen levels below 0.01%, and with inference can be used to

calculate percent carbon potential in a known CO gas, or percent water in a known H2gas Advantage of in-situ measurement and good reliability Frequent calibration not required

• Infrared absorption: Good for measurement of CO, CO2, or CH4 concentration in ranges of interest in hydrocarbon atmospheres Same unit can be configured to measure all of the above gases simultaneously if desired, which allows accurate calculation of percent carbon potential Cost is high, with multiple furnaces usually routed to a single analyzer Frequent calibration is required Major disadvantage is that sample of gas must be transported to analyzer

• Dew point: If measured by variable pressure change/condensation method, results by inference can be

used to calculate percent carbon potential against known carbon and hydrogen gas levels Disadvantage

is that measurements are physically difficult and operator technique and interpretation may play a major role Sample must be transported to the analyzer Although there are automatic dew point-measuring systems, none has worked for extended periods in hydrocarbon atmospheres without a significant amount of preventative maintenance having to be done on the system

• Mass spectrometer: Can determine the composition of gas completely, except for inaccuracies at low

levels with some gases Calibration is difficult Other disadvantages are cost, required operator skill level, and sample transport Rarely used on-line in heat treatment

• Gas chromatography: See mass spectrometer

While complex gas analysis systems have not yet found their way into day-to-day process control operations in heat treating, they may be of value in an off-line mode for a situation as shown in Fig 9 Here the results of gas analysis, specifically heavy hydrocarbons, are plotted over the course of a year The trend shows the shifting gas supplies (presumably as a result of demand) An on-site gas analysis could provide valuable information when evaluating furnace atmosphere quality problems

Fig 9 Variation in natural gas composition monitored in the Canton, OH, area over nearly a 2 1

Quenchant Bulk Temperature. The measurement of quenchant bulk temperature is generally the only measurement

undertaken in most shops Often this is a monitor-only function, with no feedback control of system heating or cooling involved

From a data acquisition standpoint, it is interesting to look at the bulk temperature during the actual quenching operation, noting any changes Typically the starting temperature and the highest temperature reached are recorded

In low agitation quench tanks, care must be taken to ensure that the temperature measured is truly indicative of the bulk temperature

Quenchant Viscosity. The viscosity of liquid quenchants, especially polymers, is an important indication of composition In oils, it can be an indication of oil aging or contamination

Viscosity may be measured in-situ, but the results will vary with temperature, making it necessary to compensate for variations in temperature

Quenchant Media Composition. Measurement of composition (and contamination) of quenching media has become

an important area of concern in recent years

In the case of quench oils, the parameters of additive levels, alkalinity, oxidation or sludge content, and water content have come to be recognized as being of significant importance Because the costs of these products have increased significantly in the last 15 years, much effort has been expended learning how to make a tank of oil last longer through the use of additives and cleaning Unfortunately, none of the tests required for oil composition determination are suited to real-time data acquisition, except for water content analysis Unfortunately, many commercially-available water-in-oil