Wiley Wastewater Quality Monitoring and Treatment_15 pptx

JWBK117-4.1 JWBK117-Quevauviller October 10, 2006 20:31 Char Count= 0 256 State Estimation for Wastewater Treatment Processes structure to design observers that are independent of the uncertain reaction rates, via a linear change of variables. As we shall see, the main condition to design a so-called asymptotic observer is that enough variables are measured. Note also that asymptotic observers fall into the broad class of observers with unknown inputs (Kudva et al., 1980; Hou and M¨uller, 1991; Darouach et al., 1994), whose principle relies on cancellation of the unknown part via a change of variables. 4.1.4.1 Preliminaries Throughout this section, we consider mass-balance models for WWTPs of the fol- lowing form (Bastin and Dochain, 1990): dx(t) dt = Kr(x) − D(t)x + D(t)x in (t) − Q(x) (5) where x, x in ∈  n represent the concentrations in the reactor and the influent, respectively, D ∈  n×n , the dilution rate matrix, Q(x) ∈  n , the gaseous ex- change between the reaction medium and the environment, r(x) ∈  p , the reac- tion rates and K ∈  n×p , a constant pseudo-stoichiometric coefficient matrix. In this representation, Kr(x) stands for the biological and biochemical conversions in the reactor (per unit of time) according to the underlying macroscopic reaction network. Suppose that the set of available measurements y corresponds to is partitioned into y 1 and y 2 such that: r y 1 is a set of q ≤ n measured state variables; without loss of generality, we assume that y 1 corresponds to the first q components of x, y 1 = [x 1 , ,x q ] T . r y 2 consists of the measured gaseous flow rates, y 2 = [q 1 (x), ,q n (x)] T . The measurements y 1 induce a partition of the state variables x = [x 1 , x 2 ] with x 1 = y 1 . Accordingly, Equation (5) can be rephrased as: dx 1 dt = K 1 r(x) − Dx 1 + Dx 1 in − Q 1 (x) dx 2 dt = K 2 r(x) − Dx 2 + Dx 2 in − Q 2 (x) (6) where matrices K 1 and K 2 , vectors x 1 in , x 2 in , q 1 and q 2 are such that K =  K 1 K 2  , x in =  x 1 in x 2 in  , q =  q 1 q 2  JWBK117-4.1 JWBK117-Quevauviller October 10, 2006 20:31 Char Count= 0 Observers for Mass-balance-based System 257 4.1.4.2 Asymptotic Observers Constructing an asymptotic observers for system (5) requires the following two technical assumptions to hold: (1) There are more measured quantities than reactions, i.e. q ≥ p. (2) Matrix K 1 has full rank. These assumptions guarantee that a nonzero r cannot cancel the term K 1 r and the q × p matrix K 1 has a left inverse. Accordingly, there exists a p ×q matrix G such that: GK 1 = I p Let us denote A =−K 2 G and M = (AI n−p ). An observer for subsystem (6) can be obtained as indicated in the following. Property 2. (Bastin and Dochain, 1990). If D is positive definite, the solution ˆ x 2 of the auxiliary system ⎧ ⎨ ⎩ d ˆ ζ 2 dt =−D  ˆ ζ 2 − Mx in  − My 2 ˆ x 2 = ˆ ξ 2 − Ay 1 converges to the solution x 2 of subsystem (6), asymptotically. 4.1.4.3 Application to an Anaerobic Digester We consider a very simple model of the anaerobic digestion process that accounts for a single ‘global’ degradation step of the soluble COD (S t ) by the biomass X t (Andrews, 1968): k t S t μ t (.)X t −−−−→ X t + k m CH 4 where k t and k m are the yield coefficients associated with COD degradation and methane production, respectively; μ t (·) stands for bacterial growth rate. The corre- sponding mass-balance model reads: dS t dt =−k t μ t (·)X t − D(S t − S in ) (7) dX t dt = μ t (·)X t − α DX t q CH 4 = k m μ t (·)X t (8) JWBK117-4.1 JWBK117-Quevauviller October 10, 2006 20:31 Char Count= 0 258 State Estimation for Wastewater Treatment Processes where q CH 4 stands for the methane outflow rate, D is the dilution rate and α is the fraction of bacteria not attached onto a support (i.e. being affected by the dilution rate D in the reactor). The objective is to design an asymptotic observer for the biomass X t , based on COD measurements, without knowing the reaction rates μ t (·). In this case, we have x 1 = y 1 = S t , x 2 = X t and Q(x) = 0. Moreover, K 1 =−k t and K 2 = 1, hence a possible choice for matrices A and G is: G =−1/k t , A = 1/k t . Finally, property 1 provides the following auxiliary differential system: d ˆ ζ 2 dt =−αD  ˆ ζ 2 − S in αk t  − (1 − α)D y 1 k t (9) ˆ X t = ˆ ζ 2 − y 1 k t (10) Equations (9) and (10) are an observer for the biomass concentration in the digester. However, this observer may provide poor estimates in practice, because of the large errors made on S in and k t . A more appropriate solution is then to use an interval observer for coping with uncertainty, as discussed in the following section. 4.1.5 INTERVAL OBSERVERS Usual observers rely on the implicit assumption that the process model is a good approximation of the real plant. Nevertheless, we have seen that WWTP models are often corrupted. In such situations where large modelling and measurement errors prevail, one can no longer construct an exact observer (i.e. with the guarantee that the observation error converges to zero asymptotically and that the convergence rate can be tuned). Instead, the observation principle must be weakened. In this section, we explain how to derive rigorous bounds enclosing the estimated states by accounting for the uncertainty in the process model (Rapaport and Gouz´e, 1999; Gouz´e et al., 2000). 4.1.5.1 Principle Interval observers require knowledge of time-varying bounds enclosing the uncer- tainty. These bounds are used to calculate time-varying bounds enclosing the state variable to be estimated. Consider the following general system: ⎧ ⎨ ⎩ dx(t) dt = f [x(t), u(t),ω(t)]; x(t 0 ) = x 0 y(t) = h[x(t),υ(t)] (S I ) JWBK117-4.1 JWBK117-Quevauviller October 10, 2006 20:31 Char Count= 0 Interval Observers 259 where known lower and upper bounds are available for the uncertain quantities ω ∈  r and υ ∈  s : ω − (t) ≤ ω(t) ≤ ω + (t) ∀t ≥ t 0 υ − (t) ≤ υ(t) ≤ υ + (t) ∀t ≥ t 0 Based on the fixed model structure (S I ) and on the set of measured data, an auxiliary dynamic system (O I ) can be designed such that it provides a lower bound and an upper bound for the state variables: ⎧ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎨ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎩ dz − dt = f − (z − , z + , u, y,ω − ,ω + ,υ − ,υ + ); z − (t 0 ) = g −  x − 0 , x + 0  dz + dt = f + (z − , z + , u, y,ω − ,ω + ,υ − ,υ + ); z + (t 0 ) = g +  x − 0 , x + 0  x − = h − (z − , z + , u, y,ω − ,ω + ,υ − ,υ + ) x + = h + (z − , z + , u, y,ω − ,ω + ,υ − ,υ + ). (O I ) Definition 2 (interval estimator). System (O I ) is said to be an interval estimator of system (S I ) if for any pair of initial conditions x − (t 0 ) ≤ x(t 0 ) ≤ x + (t 0 ), there exists bounds z − (t 0 ), z + (t 0 ) such that the solutions of the coupled system (S I , O I ) verify: x − (t) ≤ x(t) ≤ x + (t) ∀t ≥ t 0 Interval estimators result from the coupling of two estimators which provide both an underestimate x − (t) andan overestimate x + (t)ofx(t). Of course, such bounds can be very large, thusmaking interval observerspracticallyuseless in somecases. However, for particular classes of systems (e.g. linear systems up to an output injection), theoretical guarantees can be given that the time-varying intervals [x − (t), x + (t)] converge to a ‘limit’ interval of finite magnitude (Gouz´e et al., 2000). Moreover, the convergence rate towards this limit interval can be tuned if certain properties hold (Rapaport and Gouz´e, 2003). Note that these ideas find their origin in the theory of positive systems (Smith, 1995). More recently, probabilistic observers have been formulated for a class of uncertain biological processes (Chachuat and Bernard, 2006); these observers take advantage of the knowledge of probability density functions (PDFs) for the uncertain parameters to calculate the PDFs of the unmeasured state variables. An application of interval observers to an anaerobic WWTP is presented next; another application to an activated sludge process can be found in Hadj-Sadok and Gouz´e (Hadj-Sadok and Gouz´e, 2001). JWBK117-4.1 JWBK117-Quevauviller October 10, 2006 20:31 Char Count= 0 260 State Estimation for Wastewater Treatment Processes 4.1.5.2 Application to an Anaerobic Digester We consider the same reduced model of an anaerobic digestion plant as in Sec- tion 4.1.4.3. The objective is to design an interval observer that estimates the COD concentration S t from on-line methane measurements, with μ t (·) and α being un- known. Using Equation (8) of the methane flow rate, Equation (7) can be rewritten as: dS t dt =−γ q CH 4 − D  S t − S in t  where the uncertain parameter γ = k t /k m is such that γ − ≤ γ ≤ γ + ; the inlet COD concentration S in t (t) fluctuates between known bounds as S in− t (t) ≤ S in t (t) ≤ S in+ t (t), ∀t ≥ t 0 ; and the initial COD concentration is bounded as S − t (t 0 ) ≤ S t (t 0 ) ≤ S + t (t 0 ). Property 3. The auxiliary dynamic system ⎧ ⎪ ⎪ ⎨ ⎪ ⎪ ⎩ dS − t dt =−γ + q CH 4 − D  S − t − S in− t  dS + t dt =−γ − q CH 4 − D  S + t − S in+ t  is an interval observer for S t , i.e. guarantees that S − t (t) ≤ S t (t) ≤ S + t (t) at each t ≥ t 0 . Proof. It is easily verified that both lower bound e − = S t − S − t and upper bound e + = S + t − S t on the observation error remain positive [see Bernard and Gouz´e (Bernard and Gouz´e, 2004) for details]. In practice, it was found that the upper bound S + t (t) is weak when large uncer- tainties are considered for γ and S in t (t). These considerations motivate the following improvements. Improvements. Theideafor reducingpredictionintervals consistsofusing structured kinetic models for μ t (·), despite uncertainty. We suppose here that the process does not operate in a region where inhibition phenomena occur, and use a Monod kinetic model for obtaining an estimate of q CH 4 as: ˆ q CH 4 (S t ) = k m μ t S t S t + K S t ˆ X t JWBK117-4.1 JWBK117-Quevauviller October 10, 2006 20:31 Char Count= 0 Interval Observers 261 where ˆ X t = (S in t − S t )/αk t . Then, these estimates are used in the following robust observer (Bernard and Gouz´e, 2004): dS t dt =−γq CH 4 − D(S t − S in t ) + λ  q CH 4 − ˆ q CH 4 (S t )  (11) Finally, an interval observer is derived from Equation (11) as: ⎧ ⎪ ⎪ ⎨ ⎪ ⎪ ⎩ dS − t dt = φ ε  S − t  −γ + q CH 4 − D  S − t −S in− t  + λ  S in− t −S − t  q CH 4 − ˆ q CH 4  S − t  dS + t dt = φ ε  S + t  −γ − q CH 4 − D  S + t − S in+ t  + λ  S in+ t −S + t  q CH 4 − ˆ q CH 4  S + t  where the function φ ε (S − t ) = S − t /(S − t + ε) {with 0 <ε inf[S t (t)]} is used to enforce S t ≥ 0; and the bounds ˆ q − CH 4 and ˆ q + CH 4 on ˆ q CH 4 are calculated by consid- ering the bounds μ − t and μ + t on μ t . This observer is applied to a real process in the next paragraph. Application to real measurements. The interval observer given in Equation (11) was implemented on a pilot-scale fixed bed up-flow anaerobic digester used for wine wastewater processing. Both the dilution rate D and methane outflow rate q CH 4 were measured on-line at a high frequency. Besides on-line measurements, a COD sensor was also used to validate the observer predictions. Details on the plant configuration and the experiments can be found in Bernard et al. (Bernard et al., 2001) and Steyer et al. (Steyer et al., 2002). The predictions of the interval observer are presented in Figure 4.1.4, together with the off-line COD measurements. Figure 4.1.4 Methane flow rate (a) and interval observer (- -) for soluble COD (b) JWBK117-4.1 JWBK117-Quevauviller October 10, 2006 20:31 Char Count= 0 262 State Estimation for Wastewater Treatment Processes 4.1.6 CONCLUSIONS In this chapter, we have provided the key ideas on how to build observers for WWTPs. Depending on the reliability of the process model at hand, the available measure- ments and the level of uncertainty associated with the influent concentrations, dif- ferent classes of observers have been considered. In particular, a distinction has been made between those observers relying on a full model description (e.g. the extended Kalman filter), and those based on a mass-balance model wherein the biological kinetics are assimilated to unknown inputs (e.g. the asymptotic observer); moreover, if bounds are known for the uncertainties, then interval observers can be designed. Clearly, other techniques exist, and we did not pretend to be exhaustive. The observers presentedhereinassumed constant parametervaluesinthe models. Insome cases, however, the parameters can evolve during process operation, and specific algorithms must be used to estimate these parameters at run-time, hence leading to adaptive observers. The implementation of an observer requires a discretization to be performed as regards the continuous-time equations, e.g. a Euler-type algorithm. Although not difficult, discretization must be performed carefully. In the case of low measurement frequency, for example, the use of a continuous/discrete observer shall be preferred to full discretization of a continuous-time observer. Finally, it is worth insisting on the fact that an observer should always be validated prior to using it in a real treatment plant. In particular, systematic and extensive comparisons should be made between the observer predictions and direct on-site measurements (other than those used to calibrate the observer). REFERENCES Andrews, J. (1968). A mathematical model for the continuous culture of microorganisms utilizing inhibitory substrate. Biotechnol. Bioengng., 10:707–723. Bastin, G. and Dochain, D. (1990). On-line Estimation and Adaptive Control of Bioreactors. Elsevier, Amsterdam, The Netherlands. Bernard, O. and Gouz´e, J. L. (2004). Closed loop observers bundle for uncertain biotechnological models. J. Process Contr., 14(7):762–774. Bernard, O., Hadj-Sadok, M. Z., Dochain, D., Genovesi, A. and Steyer, J P. (2001). Dynamical model developmentand parameteridentification foran anaerobicwastewatertreatment process. Biotechnol. Bioengng., 75(4):424–438. Chachuat, B. and Bernard, O. (2006). Probabilistic observers for a class of uncertain biological processes. Int. J. Robust Nonlinear Control, 16(3):157–171. Chachuat, B., Roche, N. and Latifi, M. A. (2003). Reduction of the ASM1 model for optimal control of small-size activated sludge treatment plants. Rev. Sci. Eau., 16(1):5–26. Darouach, M., Zasadzinski, M. and Xu, S. J. (1994). Full-order observers for linear systems with unknown inputs. IEEE Trans. Automat. Contr., 39(3):1068–1072. Gauthier, J. P. and Kupka, I. (2001). Deterministic Observation Theory and Applications. Cambridge University Press, New York. JWBK117-4.1 JWBK117-Quevauviller October 10, 2006 20:31 Char Count= 0 References 263 Gouz´e, J L., Rapaport, A. and Hadj-Sadok, M. Z. (2000). Interval observers for uncertain biolog- ical systems. Ecol. Modell., 133:45–56. Gudi, R. D., Shah, S. L. and Gray, M. R. (1995). Adaptive multirate state and parameter estimation strategies with application to a bioreactor. AIChE J., 41:2451–2464. Hadj-Sadok, M. Z. and Gouz´e, J L. (2001). Estimation of uncertain models of activated sludge process with interval observers. J. Process Contr., 11(3):299–310. Henze, M., Grady, C. P. L., Gujer, W., Marais, G. v. R. and Matsuo, T. (1987). Activated Sludge Model No. 1. Technical report, IAWQ, London. Hou, M. and M¨uller, P. (1991). Design of observers for linear systems with unknown inputs. IEEE Trans. Automat. Contr., 37(6):871–875. Kudva, P., Viswanadham, N. and Ramakrishna, A. (1980). Observers for linear systems with unknown inputs. IEEE Trans. Automat. Contr., 25(1):113–115. Lewis, F. L. (1986). Optimal Estimation (with an Introduction to Stochastic Control Theory). John Wiley & Sons, Ltd, New York. Luenberger, D. G. (1966). Observers for multivariable systems. IEEE Trans. Automat. Contr., 11:190–197. Luenberger, D. G. (1979). Introduction to Dynamic Systems: Theory, Models and Applications. John Wiley & Sons, Ltd, New York. Lukasse, L., Keesman, K. and van Straten, G. (1999). A recursively identified model for short-term predictions of NH 4 /NO 3 concentrations in alternating activated sludge processes. J. Process Contr., 9(1):87–100. Rapaport, A. and Gouz´e, J L. (1999). Practical observers for uncertain affine outputs injec- tion systems. In Proc. ECC’99, Karlsruhe, Germany. European Union Control Association (CD-ROM). Rapaport, A. and Gouz´e, J L. (2003). Parallelotopic and practical observers for non-linear uncer- tain systems. Int. J. Control, 76(3):237–251. Smith, H. L. (1995). Monotone Dynamical Systems: an Introduction to the Theory of Competitive and Cooperative Systems. American Mathematical Society, Providence, Rhode Island. Steyer, J P., Bouvier, J C., Conte, T., Gras, P. and Sousbie, P. (2002). Evaluation of a four year experience with a fully instrumented anaerobic digestion process. Wat. Sci. Technol.,45 (4–5):495–502. Zhao, H. and K¨ummel, M. (1995). State and parameter estimation for phosphorus removal in an alternating activated sludge process. J. Process Contr., 5(5):341–351. JWBK117-4.2 JWBK117-Quevauviller October 10, 2006 20:31 Char Count= 0 4.2 Industrial Wastewater Quality Monitoring Olivier Thomas and Marie-Florence Pouet 4.2.1 Regulatory Context 4.2.2 Characteristics of Industrial Wastewater 4.2.3 Monitoring of Industrial Wastewater 4.2.4 Variability 4.2.5 Accident Detection and Source Identification References 4.2.1 REGULATORY CONTEXT Among the regulation texts related to the management of industrial wastewater, the European directive of 21 May 1991 (see Chapter 1.1) concerning urban wastewater treatment (91/271/EEC) (Annex I, C) states that industrial wastewater entering col- lecting systems and urban wastewater treatment plants shall be subject to such pre- treatment as is required in order to: r protect the health of staff working in collecting systems and treatment plants; r ensure that collecting systems, wastewater treatment plants and associated equip- ment are not damaged; r ensure that the operation of the wastewater treatment plant and the treatment of sludge are not impeded; Wastewater Quality Monitoring and Treatment Edited by P. Quevauviller, O. Thomas and A. van der Beken C  2006 John Wiley & Sons, Ltd. ISBN: 0-471-49929-3 JWBK117-4.2 JWBK117-Quevauviller October 10, 2006 20:31 Char Count= 0 266 Industrial Wastewater Quality Monitoring r ensure that discharges from the treatment plants do not adversely affect the en- vironment, or prevent receiving water from complying with other Community Directives; r ensure that sludge can be disposed of safely in an environmentally acceptable manner. When an industrial wastewater discharges directly, after treatment, in the receiving medium, it shall respect the regulation compliance depending on the industry nature, the effluent characteristics, the minimum efficiency of treatment and the support ca- pacity of the receiving medium. For example, these considerations are precised in the USA, in the National Pollution Discharge Elimination System (NPDES) pro- gram and in the total maximum daily load (TMDL) of the Clean Water Act (see Chapter 1.1). 4.2.2 CHARACTERISTICS OF INDUSTRIAL WASTEWATER Contrary to the treatment of industrial wastewater, there exist very few studies and books about industrial wastewater characteristics. The composition of wastewater varies with the industrial activity (Table 4.2.1), but also with the size and location of enterprises. Small and medium enterprises (SMEs) of urban area are generally connected to the sewer network after or without wastewater treatment, but larger industries are often outside the urban area and equipped with their own treatment plant with a specific discharge in the receiving medium. Moreover, there is a trend to recycle water in processes and thus to reduce and minimise the effluent load (Trebuchon et al., 2000; Gomes et al., 2006). For all industries, the regulation gives threshold limits in concentration and daily load, depending on the existence of a treatment plant, parameters and industrial activity. A way of characterised industrial wastewater is also to consider the ratio of pollution [for example, biological oxygen demand (BOD)] on the production unit (for example ton of paper). This ratio is the specific load. One key characteristic of industrial wastewater is its daily, weekly or seasonal variation in compositionand loadwhich can be expressed by a statistical distribution, plotting the value of a given parameter (concentration, load or specific load) against the percent of time for which the value of the parameter is equal or less than a given one (Eckenfelder, 2001). The composition variation can also be estimated by the variability explained in Section 4.2.4. If the characteristics of raw wastewater of a given enterprise exceed the limits, the industry must install a pretreatment step before the discharge into the sewer network. Nevertheless, considering the composition of some industrial wastewater, its impact on sewer and treatment plant can be effective, particularly if the industry is close to the treatment plant. The main impact is related to shock loads and toxicity effects on [...]... Count= 0 5.1 Quality Survey of Wastewater Discharges Marie-Florence Pouet, Genevi` ve Marcoux and Olivier Thomas e 5.1.1 Characteristics and Impact of Wastewater Discharges 5.1.2 Chemical Monitoring 5.1.3 Biological Monitoring 5.1.4 In Practice References 5.1.1 CHARACTERISTICS AND IMPACT OF WASTEWATER DISCHARGES A wastewater discharge concerns, most of the time, the treated effluents of a wastewater treatment... raw wastewater and on the type and efficiency of the applied treatment (physico-chemical or biological), particulate matter (coarse colloids and suspended matter) may represent Wastewater Quality Monitoring and Treatment Edited by P Quevauviller, O Thomas and A van der Beken C 2006 John Wiley & Sons, Ltd ISBN: 0-471-49929-3 JWBK117-5.1 JWBK117-Quevauviller 276 October 10, 2006 20:31 Char Count= 0 Quality. .. which is more and more studied (Barcelo, 2005) 4.2.3 MONITORING OF INDUSTRIAL WASTEWATER The monitoring of industrial wastewater quality does not differ significantly from that of urban wastewater The objectives, the main parameters and the methodologies are similar However, there are some differences due to the nature and production mode of industrial wastewater: r The parameter values (concentration... classes; and (iii) the validation that these substances are the key toxicants Unfortunately, the numerous and complex manipulations and analysis of this approach are limitations to its application Finally, further efforts are needed in research and development for the characterisation and monitoring of industrial wastewater REFERENCES Barcelo, D (2005) Emerging Organic Pollutants in Wastewater and Sludge... heavy rainfall, raw wastewater can also be discharged, without treatment, from combined sewer overflows The aim of this section is to present the main characteristics and impacts of wastewater discharges and the monitoring tools available, particularly the alternative ones for an on-site/on-line use Wastewater discharges are mixtures of organic and mineral compounds, mainly dissolved and colloidal However,... reuse and the variation in flow and load to undergo wastewater treatment (Eckenfelder, 2001) Generally, the same parameters as for urban wastewater are acquired [temperature, pH, conductivity, turbidity, TOC, chemical oxygen demand (COD), BOD, total suspended solids (TSS), N and P forms], plus some specific parameters depending on the nature of the expected industrial wastewater discharges, such as oil and. .. rather stable and low concentrated matrix (Vaillant, 2000) Typical residual constituents of treated wastewater effluents and their potential impacts on receiving medium are summarized in Table 5.1.1 Table 5.1.1 Main residual constituents found in treated wastewater effluents and their impacts (Adapted from Metcalf and Eddy, 2003) Residual constituents Suspended and colloidal solids (inorganic and organic)... industrial wastewater (in mg/l) (Adapted from Eckenfelder, 2001; Baur` s, 2002; Metcalf and Eddy, e 2003; Pons et al., 2004; Degr´ mont, 2005) e JWBK117-4.2 Char Count= 0 JWBK117-4.2 268 JWBK117-Quevauviller October 10, 2006 20:31 Char Count= 0 Industrial Wastewater Quality Monitoring biological treatments The potential risk must be evaluated from sampling campaign and measurement, analysis and tests... Yuxia, C., Huihua, S., Zhonghai, D and Hongjun J (2004) Ecotoxicol Environ Saf., 57, 426–430 Jolibois, B and Guerbet, M (2005) Mutat Res., 565, 151–162 Juliastuti, S.R., Baeyens, J and Creemers, C (2003) Environ Engin Sci., 20(2), 79–90 Metcalf and Eddy (2003) Wastewater Engineering, Treatment and Reuse, 4th Edn McGraw Hill, Boston Muret, C., Pouet, M.F., Touraud, E and Thomas, O (2000) Water Sci Technol,... VARIABILITY One key point of industrial wastewater monitoring is that composition and load variability can be important, depending on the size and activity of the industry For industries with a single line production (pulp and paper, metal industry and some agro-food plants) the variability can be rather low, contrary to more complex industries (refinery, fine chemicals and textile) or SMEs with a very variable . and treatment plants; r ensure that collecting systems, wastewater treatment plants and associated equip- ment are not damaged; r ensure that the operation of the wastewater treatment plant and. treatment plant and the treatment of sludge are not impeded; Wastewater Quality Monitoring and Treatment Edited by P. Quevauviller, O. Thomas and A. van der Beken C  2006 John Wiley & Sons, Ltd 0 4.2 Industrial Wastewater Quality Monitoring Olivier Thomas and Marie-Florence Pouet 4.2.1 Regulatory Context 4.2.2 Characteristics of Industrial Wastewater 4.2.3 Monitoring of Industrial Wastewater 4.2.4