Designing Capable and Reliable Products Episode 1 Part 5 potx

this expertise should be used to augment the analysis. Because the assembly variability knowledge collated is a mixture of quantitative and qualitative data, validation of the indices used in the charts is more dicult than the manufacturing variability risks analysis. Experts from the assembly industry were therefore used in their ®nal assessment and allocation. However, despite the fact that the validation of the assembly capability measures are dicult to determine empirically, they are found to be of great bene®t when comparing and evaluating a number of design principles. Notes ¬ Also includes high ®nish.  Chemical contamination to operator/machine or product. ® If component is not sensitive to the characteristics listed, h p  1:0. ¯ A component is said to be thin if one of its major dimensions is 0.25 mm thick. ° Includes mechanical and air assisted mechanisms. Figure 2.17 Handling process risk, h p Component Assembly Variability Risks Analysis 65 Figure 2.18 Fitting process risk, f p 66 Designing capable components and assemblies 2.5 The effects of non-conformance 2.5.1 Design acceptability FMEA can be used to provide a quantitative measure of the risk for a design. Because it can be applied hierarchically from system through subassembly and component levels down to individual dimensions and characteristics, it follows the progress of the design into detail. FMEA also lists potential failure modes and rates their Severity (S), Occurrence (O) and Detectability (D). It therefore provides a possible means for linking potential variability risks with consequent design acceptability and associated costs. Note that the ratings of Occurrence and Detectability are equated to prob- ability levels. To determine the costs of failing to design into a product the levels of process capability proper to the severity of consequential failure, it has been assumed that a fault is undetected and always results in a failure. Also, the estimate is not sensitive to changes in the product cost, volume and rework costs. The cost of customer dissatisfaction, over and above the accounted cost of rework or a returned product, is very dicult to assess. The model which follows is based on research with Lucas Automotive Electronics (Lucas PLC, 1994) and utilizes their speci®c ratings for FMEA Occurrence (O)andSeverity(S) (see Figures 2.19 and 2.20 respectively). The analysis of failure consequence is a specialist area within each industry, as is the de®ni- tion of the probability of occurrence (Smith, 1993). The alignment of the ratings chosen here with other existing FMEA ratings, for example those in Chrysler Corporation et al. (1995), may be considered when using CA with other business practices. `Failure' in the context here means that product performance does not meet requirements and is related back in the design FMEA to some component/character- istic being out of speci®ed limits ± a fault. The probability of occurrence of failure (O) caused by a fault can be expressed as: O  P f Á d H Á P of 2:11 where: P f  probability of a fault in the component/characteristic d H  probability that the fault will escape detection P of  conditional probability that, given there is a fault, the failure will actually occur: Equation 2.11 recognizes that for a failure to occur, there must be fault, and that other events may need to combine with the fault to bring about a failure. The equa- tion states that the probabilities of each of these factors occurring must be multiplied together to calculate the probability of failure. In the special case when d H  1 and P of  1, that is, when if a fault occurs it will not be detected and that failure is certain, occurrence equates directly to the probability of a fault. The probability of a fault in turn depends on the capability of the process used for the component/characteristic. The effects of non-conformance 67 A standard for the minimum acceptable process capability index for any component/characteristic is normally set at C pk  1:33, and this standard will be used later to align costs of failure estimates. If the characteristics follow a Normal distribution, C pk  1:33 corresponds to a fault probability of: P f  30 ppm  30  10 ÿ6 2:12 While 30 ppm may be acceptable as a maximum probability of occurrence for a failure of low severity, it is not acceptable as severity increases. An example table of FMEA Severity Ratings was shown in Figure 2.20. In the `de®nite return to manufacturer' (a warranty return) or `violation of statutory requirement' region (S  5orS  6), the designer would seek ways to enhance the process capability or else utilize some inspec- tion or test process. Reducing d H will reduce occurrence, as indicated by equation 2.11, but inspection or test is of limited eciency. If severity increases to, say, `complete failure with probable severe injury and/or loss of life' (S  9), the designer should reduce occurrence below the level aorded by two independent characteristics protecting against the fault. If each is designed to, then this implies a failure rate of: P f 30  10 ÿ6  2  9  10 ÿ10 2:13 Again, this standard aligns with the costs of failure analysis below. Figure 2.19 Typical FMEA Occurrence Ratings (O) 68 Designing capable components and assemblies Figure 2.21 is a graph of Occurrence against Severity showing a boundary of the acceptable design based on these criteria for the case when d H  1andP of  1. The graph is scaled in terms of FMEA ratings for Occurrence, equivalent probabilities and parts-per-million (ppm). In general, just as reducing d H reduces the probability of failure occurrence, so redu- cing the conditional probability of failure P of moves the component/characteristic towards the acceptable design zone. However, this must be applied with considerable caution. The derivation of the acceptable maximum for failure occurrence at S  9 and above is essentially an application of the conditional probability: the conditional probability of failure at one characteristic is the probability of failure at the other protecting characteristic. In this case, it would normally be more appropriate to design two characteristics than to rely on one very capable one. At such low prob- ability levels, the chance of random unforeseen phenomena throwing the process out of control is signi®cant. On the other hand, consider the case of a `secondary back-up' (external to the system) for a component/characteristic of severity S  7. If the designer assumes the back-up is designed for C pk  1:33, then the analysis would reduce that for the internal back-up S  9 case above. The acceptable design limits on Figure 2.21 Figure 2.20 Typical FMEA Severity Ratings (S) The effects of non-conformance 69 imply that it is prudent to allow a safety margin for external components not under the designer's control and scale up the conditional probability of failure. In summary, for a component/characteristic it is possible to de®ne an area of acceptable design on a graph of occurrence versus severity. The acceptability of the design can be enhanced somewhat by the addition of inspection and test operations. The requirements of process capability may be relaxed to a degree, as the conditional probability of failure reduces, but this should be subject to a generous safety margin. Figure 2.21 The limits of acceptable design 70 Designing capable components and assemblies It will now be demonstrated that the boundaries of acceptable design proposed above can be expressed in terms of the costs of failure. 2.5.2 Map of quality costs From the quality±cost arguments made in Section 1.2, it is possible to plot points on the graph of Occurrence versus Severity and construct lines of equal failure cost (% isocosts). Figure 2.22 shows this graph, called a Conformability Map. Because of uncertainty in the estimates, only a broad band has been de®ned. The boundary of acceptable design for a component or assembly characteristic in the zone S ! 6 corresponds fairly closely to a failure cost line equivalent to 0.01% of the unit cost. The region of unacceptable design is bounded by the inter- section of the horizontal line of C pk  1 and the 1% isocost. A process is not con- sidered capable unless C pk ! 1 and a failure cost of less than 1% is thought to be acceptable. Components/characteristics in the unacceptable design zone are virtually certain to cause expensive failures unless redesigned to an occurrence level acceptable for their failure severity rating (minimum C pk  1:33). In the intermediate zone, if acceptable design conformability cannot be achieved, then special control action will be required. If special action is needed then the component/characteristic is critical. However, it is the designer's responsibility to ensure that every eort is made to improve the design to eliminate the need for special control action. The 0.01% line on Figure 2.22, implies that even in a well-designed product, there is an incurred failure cost. Just 100 dimensional characteristics on the limit of acceptable design are likely to incur a 1% cost of failure. The quality cost rises dramatically where design to process capability is inadequate. Just ten product characteristics on the 1% isocost line would give likely failure costs of 10%. (The 1% and 10% isocost lines on the Conformability Map are extrapolated from the conditions for the 0.01% and 1% isocost lines.) Clearly, the designer has a signi®cant role in reducing the high cost of failure reported by many manufacturing organizations. Variation is typically measured using process capability indices, as well as ppm fail- ure, but it would be advantageous to scale the variability risks predicted by CA on the Conformability Map too. Scales showing C pk and approximate variability risk measures, q m and q a , from CA equivalents have been added to the Conformability Map for use in the CA methodology. An empirical relationship exists between C pk and q m that allows us to do this, as discussed earlier. Because the assembly variability risk, q a , is also scaled in the same way as q m , it is assumed that the level of variability in assembly can also be associated to a notional process capability index. The Con- formability Map, through the inclusion of q m , q a and FMEA Severity Rating (S) for a particular design, can be used to give the total risk potential in terms of the isocost at the design stage. The accumulated isocosts can then be translated into potential failure costs. In this way, a total cost of failure for a design can be evaluated. The use of the Conformability Map in determining the potential costs of non- conformance or failure should be associated with the application of evaluating and comparing dierent design schemes in practice. The estimated failure costs act as a measure of performance by which to make a justi®ed selection of a particular The effects of non-conformance 71 design scheme. Assuming that the costs are absolute values is not recommended due to the diculty in obtaining accurate data for the quality±cost model (see later for guidance on the use and calculation of potential costs of failure for a design). It has been cited that Æ5% is a good enough accuracy for the prediction of quality costs at the business assessment level (Bendell et al., 1993), but the achievement of this at the design stage is a very dicult task. Figure 2.22 The Conformability Map 72 Designing capable components and assemblies The Conformability Map enables appropriate C pk values to be selected and through the link with the component variability risks, q m and q a , it is possible to deter- mine if a product design has characteristics that are unacceptable, and if so what the cost consequences are likely to be. The two modes of application are highlighted on Figure 2.23. Mode A shows that the quality loss associated with a characteristic at C pk  1 and FMEA Severity S6 could potentially be 8% of the total product Figure 2.23 Conformability Map applications The effects of non-conformance 73 cost. Mode B indicates that a target C pk  1:8 is required for an acceptable design characteristic having an FMEA Severity Rating S8. 2.6 Objectives, application and guidance for an analysis A short review of the CA process is given before proceeding with the applications of the technique to several industrial case studies. The three key stages of CA are shown in Figure 2.24 within the simpli®ed process of assessing a design scheme. . Component Manufacturing Variability Risks Analysis ± As mentioned previously, the ®rst of the three key stages in CA is the Component Manufacturing Variability Risks Analysis. When detailing a design, certain characteristics can be considered Figure 2.24 Conformability Analysis methodology 74 Designing capable components and assemblies [...]... and assemblies Figure 2. 25 Conformability Analysis procedure ¯owchart Figure 2.26 Component Manufacturing Variability Risks Analysis of a cover support leg Objectives, application and guidance for an analysis 79 80 Designing capable components and assemblies 12 10 Frequency 8 µ = 49 .57 5 mm = 49 .5 mm σ = 0.8 61 mm = 41 Superimposed Normal distribution 6 4 2 0 47 47 .5 48 48 .5 49 49 .5 50 50 .5 51 51 . 5 52... 1. 0 H ˆ 1. 0 For the bolt ®tting operation: fp ˆ A  B  C  D  E  F  G  H ˆ 2 :11 The Component Assembly Variability Risk is given by: qa ˆ hp  f p qa ˆ 1: 0  2 :11 ˆ 2 :11 Therefore, qa 4 ˆ 2 :11 However, once the cover, tab washer and bolt are in place an additional process is carried out on the washer to bend the tab Thus from the additional assembly process 83 84 Designing capable components and. .. Therefore, from the handling table in Figure 2 .17 , hp ˆ 1: 0 Now assessing the ®tting to process risk, fp , from Figure 2 .18 : A (cannot be assembled the wrong way) A ˆ 1. 0 B (no positioning reliance to process) C (automatic screwing) B ˆ 1. 0 C ˆ 1. 6 D (straight line assembly from above) D ˆ 1. 2 E (single process ± one bolt in hole) F (no restricted access or vision) E ˆ 1. 1 F ˆ 1. 0 G (no alignment problems)... risk indices hp and fp For the handling process, assuming automatic assembly, the 81 Figure 2.28 Variability risks table for the cover support leg 82 Designing capable components and assemblies Objectives, application and guidance for an analysis Figure 2.29 Fixing bolt assembly and sequence of assembly ®xing bolt is supplied in a feeder and does not have characteristics which complicate handling Therefore,... speci®cation, assuming a Normal distribution Cpk ˆ j ÿ Ln j 3 …2 :14 † Objectives, application and guidance for an analysis where:  ˆ mean of distribution Ln ˆ nearest tolerance limit  ˆ standard deviation: Therefore: Cpk ˆ j49 :57 5 ÿ 49:5j ˆ 0:03 3  0:8 61 As a result of poor capability, a high level of machine downtime was experienced, and the company which manufactured the assembly machine were involved... which can be considerable The costs in the safety critical area have been more 75 76 Designing capable components and assemblies dicult to assess and have a greater margin of error In essence, as failures get more severe, they cost more, so the only approach available to a business is to reduce the probability of occurrence 2.6 .1 Objectives CA is primarily a team-based product design technique that, through... tolerance set at Æ0 :5 mm was just within acceptable limits at a nominal dimension of 50 mm However, the variability risks mp and gp decreased the probability of obtaining it substantially Completing a variability risks table A variability risk table (as shown in Figure 2.28 for the cover support leg analysis above) is a more ecient and traceable way of presenting the results of the ®rst part of the analysis... isocosts in the non-safety critical region (FMEA Severity Rating 5) come from a sample of businesses and assume levels of cost at internal failure, returns from customer inspection or test and warranty returns The isocosts in the safety critical region (FMEA Severity Rating >5) are based on allowances for failure investigations, legal actions and product recall, but do not include elements for loss of current... Frequency 8 µ = 49 .57 5 mm = 49 .5 mm σ = 0.8 61 mm = 41 Superimposed Normal distribution 6 4 2 0 47 47 .5 48 48 .5 49 49 .5 50 50 .5 51 51 . 5 52 52 .5 53 Figure 2.27 Statistical process data for dimension `A' of the cover support leg on the bracket, which must be 50 Æ 0 :5 mm to e€ectively operate in an automated assembly machine The assembly machine is used to produce the ®nal product, of which the support... product development team be familiar with the main capabilities and characteristics of the manufacturing and assembly methods selected, as well as the materials considered, in order to obtain the full bene®t from the methodology Technical and economic knowledge for some common manufacturing processes can also be found in Swift and Booker (19 97) The analysis requires the declaration of a sequence of . Normal distribution. C pk  j ÿ L n j 3 2 :14  µ = 49 .57 5 mm = 49 .5 mm σ = 0.8 61 mm = 41 Superimposed Normal distribution 47 47 .5 48 48 .5 49 49 .5 50 50 .5 51 51 . 5 52 52 .5 53 Frequency 12 10 8 6 4 2 0 Figure 2.27 Statistical. of: P f 30  10 ÿ6  2  9  10 10 2 :13  Again, this standard aligns with the costs of failure analysis below. Figure 2 .19 Typical FMEA Occurrence Ratings (O) 68 Designing capable components and assemblies Figure. Assembly Variability Risks Analysis 65 Figure 2 .18 Fitting process risk, f p 66 Designing capable components and assemblies 2 .5 The effects of non-conformance 2 .5 .1 Design acceptability FMEA can be