Robust Monitoring Strategy

The elements of the robust monitoring strategy builds on the methods discussed previously (e.g., PC A in 3.1 and signal filtering in 6.2.3). Here, only the key variations will be introduced. 

Nonlinear PCA To address the nonlinearity in the identity mapping of multivariate data, a nonlinear counterpart of the PCA can be used (see Section 3.6.1). As the versions of NLPCA make use of the neural network (NN) concept to address the nonlinearity, they suffer from the known overparameterization problem in the case of noise corrupted data. Data with small SNR will also give rise to extensive computations during the training of the network. Shao et al. [266] used wavelet filtering to pre-process the data followed by IT-net to detect the non-conforming trends in an industrial spray drier. 

The approach presented here is based on Kramer s work [150] where his method uncovers both linear and nonlinear correlations independent of the 

The set of vectors T, T2, , Tp constitutes an orthogonal set. The procedure requires designation of a vector from which all other vectors are constructed so as to be orthogonal to the initially chosen vector. In this case, there is no restriction on which vector has to be chosen first. 

The technique is referred to as Backward Substitution for Sensor Identification and Reconstruction (BSSIR) and it is based on the principle that process upsets and sensor failures can be identified in the presence of redundancy among sensor arrays. Due to process characteristics, these measurements may have strong correlations among each other, particularly the ones in close proximity and measuring the same variable. Therefore, when a disturbance affects the process, it would be sensed by a group of sensors rather than just by one. However, if a sensor malfunctions (e.g., due to complete failure, bias, precision degradation, or a drift), then this will only affect the individual sensor performance, at least initially. If the malfunc- 

---

Figure 7.33. The schematic of robust monitoring strategy. Reprinted from [60]. Copyright 2001 with permission from Elsevier.

In the design of the proce.ss, it is important to develop cut point strategies which lead to robust control of the product pool. Use of strategies which combine monitoring the elution volume and detector response can lead to robust control strategies which avoid the collection and analysis of fractions. The development of strategies which are sufficiently robust to accept the variability as.sociated with the feed composition, the degradation of the stationary phase, temperature variations, and mobile-phase composition, is not trivial and an important step in the development process. 

The cut point location strategy in chromatography is important because it determines the product purity, yield and product concentration to be processed forward. Use of detectors which allow on-line monitoring of the column effluent facilitates development of robust pooling strategies. Common strategies for identification of the start of the pool... 

F Doymaz, J Chen, JA Romagnoli, and A Palazoglu. A robust strategy for real-time process monitoring. J. Process Control, 11 343-359, 2001. 

The monitoring and control strategy is based primarily on existing methods and systems in use or planned for use at other CSDP facilities or similar commercial installations (e.g., industrial wastewater treatment plants). All of the instruments and control elements are standard industrial hardware with field-proven high reliability and robustness. The overall system consists of the BPCS, ESS, and individual equipment PECs. The BPCS contains microprocessor-based controllers. The ESS is a separate, dedicated safety system consisting of PECs or microprocessor-based controllers that will provide protective logic and enable safe shutdown of the facility. 

Doymaz, F., Chen, J., Romagnoli, J.A., Palayoglu, A., A Robust Strategy for Real-Time Process Monitoring, Journal of Process Control, 11,2(X)1,343-359. 

---