Instrument readings in process plants do not obey basic conservation laws and one has to perform data reconciliation to obtain process variable estimates. Typically, data reconciliation results are corrupted by the presence of biased instruments and leaks, as well as departures from steady state. The accuracy of the data and the quality of the information are related to the algorithm used as well as the sensor network to filter these disturbances appropriately. Therefore, since the onset of the field of data reconciliation in process plants, the issue of detection and identification of gross errors has received central attention. Not only it is desired to have the proper location of gross errors, but it is also advantageous to have an estimate of its size. The former helps instrument maintenance in the case of biases and process monitoring in the case of leaks. The latter constitutes valuable information for production accounting. Three kinds of strategies have been developed for identifying multiple gross errors: a) Serial Elimination identifies one gross error at a time using some statistic tests and eliminates the corresponding measurement until no gross error is detected; b) Serial Compensation which identifies the gross error and its size, compensates the measurement and continues until no error is found; c) Simultaneous or Collective Compensation which proposes the estimation of all gross errors simultaneously. A good survey of these techniques is available in the books by Romagnoli and Sánchez (2000) and, Narasimhan and Jordache (2000). The D-statistic is widely used in Statistical Process Control to reliably detect the out of control status, but by itself it offers no assistance as fault identification tool. Different strategies have been proposed to calculate the contribution of each process variable to the inflated statistic. They work in the original or in the latent variable space. Mason et al. (1995, 1997) presented a strategy to decompose the D-statistic into the contributions of each variable, which is intended to identify the fault source. Due to the combinatorial nature of the formulation, a great number of decompositions are obtained, increasing the complexity of the identification procedure. A straightforward method to decompose the D statistic as a unique sum of each variable contribution was recently developed by Alvarez et al. (2006, 2007), which is called OSS (Original Space Strategy). This decomposition provides a clear understanding about the physical meaning of the negative contributions and estimates a bound for them. In this work, the detection and identification capabilities of the monitoring technique presented by Alvarez et. al (2007), are compared with those corresponding to the most commonly used gross error detection and identification techniques for some benchmarks.

References Alvarez, R; A. Brandolin; Sánchez, M. Process Monitoring: An Alternative Approach for D2 Contribution Plots, Proceedings of XII Interamerican Congreso of Chemical Engineering, 2006. Alvarez, R; A. Brandolin; Sánchez, M. On the variable contributions to the D statistic, Accepted for publication in Chemometrics and Intelligent Laboratory Systems, April 2007. Mason, R. L., N. D. Tracy and J. C. Young, “Decomposition of T2 for multivariate control charts interpretation,” J. Qual. Technol., 27, 99-108, 1995. Mason, R. L., N. D. Tracy and J. C. Young, “A practical approach for interpreting multivariate T2 control chart signals ,” J. Qual. Technol., 29, 396-406, 1997. Narasimhan S.; Jordache, C. Data Reconciliation and Gross Error Detection; Gulf Publishing Company, Houston, 2000. Romagnoli, J.; Sánchez M. Data Processing and Reconciliation for Chemical Process Operations; Academic Press: San Diego, 2000.

Extended Abstract Status: File Uploaded