(301z) Dynamic Data Reconciliation Using Process Simulator and Wavelet Denoising

Data reconciliation adjusts the process measurements to improve their accuracy and to conform them to the process constraints. Parameters can be estimated simultaneously during the reconciliation or separately. These two approaches are termed as coupled and decoupled reconciliation respectively. These approaches for steady state case were presented by MacDonald and Howat (1988). For dynamic situation, moving-horizon-optimization based approach (Liebman et al (1992)) can be used.

Reconciliation requires process model constraints and their derivates with respect to the measurements. In the case of coupled approach the derivatives of the model constraints with respect to the parameters are also required. Programming the model and the Jacobians requires significant time and effort. In the case of decoupled approach process simulators can be used to provide them.

Using decoupled approach generally results in estimates that are less accurate than those from coupled (Lee (2002)). However, coupled and decoupled approaches will result in similar estimates if the original measurements themselves have low error. To reduce the error in the original measurements, wavelet denoising can be used. The denoised data can then be reconciled. This approach has already been applied in the literature (Kong et al (2002)).

Our hypothesis is that wavelet denoising of the data followed by decoupled dynamic reconciliation with the aid of process simulator will result in accurate estimates of measurements and the parameters. Initial results from testing this idea have been promising. The advantage of this approach is that it significantly reduces the programming effort.

In this work, the above idea is applied for reconciling data from bath distillation process. Batch distillation module in ChemCAD will be used to provide the process model and the Jacobian matrices. Dynamic decoupled data reconciliation will be performed in MathCAD embedded in Microsoft Excel. ChemCAD will interact with Excel (and the MathCAD program in it) through COM interface. These interactions will be programmed using Visual Basic.

1. MacDonald, R.J. and Howat, C.S., 1988. Data Reconciliation and Parameter Estimation in Plant Performance Analysis, AIChE Journal, 34(1): 1-8.

2. Liebman, M. J.; Edgar, T. F. and Lasdon, L. S., 1992. Efficient Data Reconciliation and Estimation for Dynamic Processes Using Nonlinear Programming Techniques. Computers and Chemical Engineering, 16(10/11): 963-986.

3. Lee, T., 2002. Measurement reconciliation and interpretation of a packed distillation column operation. Master's Thesis, University of Kansas.

4. Kong, M; Chen, B. and He, X., 2002. Wavelet-Based Regularization of Dynamic Data Reconciliation. Industrial Engineering Chemistry Research, 41: 3405-3412.