An issue often raised in multivariate statistical processcontrol, when using statistical projection-based techniques to define nominal process behaviour, is that of the assured identification of the variables causing...
详细信息
An issue often raised in multivariate statistical processcontrol, when using statistical projection-based techniques to define nominal process behaviour, is that of the assured identification of the variables causing an out-of-statistical-control signal. One approach which has been adopted is that once a change in process operating conditions has been detected, the contribution of the individual variables to the principal component scores or squared prediction error, the Q-statistic, are examined. Adopting this approach, it is important that those variables responsible for, or contributing to, the process change are clearly identifiable. In process modelling and estimation studies, confidence bounds are typically placed around the model predictions. Currently confidence bounds are not used to identify the limits of normal behaviour for the individual multivariate statistical contributions, resulting in the interpretation of the contribution plot being left to the user. This paper presents a potential solution to the definition of confidence bounds for contribution plots. The methodology is based on bootstrap estimates of the standard deviations of the loading matrix. The proposed approach is evaluated using data from a benchmark simulation of a continuous stirred tank reactor system. The preliminary results are encouraging. Copyright (C) 2000 John Wiley and Sons, Ltd.
Projection to Latent Structures (PLS) is a linear regression technique for nondynamic problems where the data is noisy, highly correlated and where there are a limited number of observations. Methodologies have been p...
详细信息
Projection to Latent Structures (PLS) is a linear regression technique for nondynamic problems where the data is noisy, highly correlated and where there are a limited number of observations. Methodologies have been proposed to integrate the nonlinear features within a linear PLS framework resulting in a non-linear algorithm. PLS has also been extended to include dynamic processes. This paper presents a non-linear dynamic PLS algorithm which incorporates polynomial or neural network functions that are integrated within the PLS algorithm through weight updating of the inner/outer models. The modelling capabilities are assessed through comparisons on a pH neutralisation process.
A new filtering method is presented which extends the SureShrink algorithm by eliminating the peak noise in the wavelet transformed signal to improve the overall filtering properties. Data from industrial plants alway...
详细信息
A new filtering method is presented which extends the SureShrink algorithm by eliminating the peak noise in the wavelet transformed signal to improve the overall filtering properties. Data from industrial plants always contain some peak noise, but ‘denoise’ algorithms such as ‘SureShrink’ can have difficulty in handling sudden large excursions in the corrupting noise. In the new algorithm the peak noise is reduced prior to filtering using the SureShrink algorithm. The pre screened data can be used to build a number of projections to latent structures regression models. Data from an industrial fluidized bed reactor is used to evaluate the new algorithm, which demonstrates improved performance in terms of improved modeling capability through use of the new data pre filtering algorithm.
Principal Component Analysis and Partial Least Squares have been used extensively for Multivariate Statistical processcontrol (MSPC) with increasing numbers of applications in the chemical manufacturing industries. I...
详细信息
Principal Component Analysis and Partial Least Squares have been used extensively for Multivariate Statistical processcontrol (MSPC) with increasing numbers of applications in the chemical manufacturing industries. In contrast the multivariate statistical technique of Canonical Correlation Analysis (CCA), despite its conceptual similarities with Principal Component Analysis and Partial Least Squares, has not found analogous applications. In this paper a library of metrics based on Hotelling’s T and the Squared Prediction Error ( SPE ) are developed for CCA. The various monitoring tools and techniques are validated and compared through application to a simulated spray-drying process.
An industrial ethylene propylene rubber compounding process is used to illustrate some of the issues that arise in the monitoring of the manufacturing performance of a process comprising both batch and continuous unit...
详细信息
An industrial ethylene propylene rubber compounding process is used to illustrate some of the issues that arise in the monitoring of the manufacturing performance of a process comprising both batch and continuous unit operations. The key issues relate primarily to the different formats of the data that are routinely collected. For example on batch type processes, measurements are collected for a fixed quantity of produced mass whilst for the continuous units, measurements are recorded for each variable at fixed time points. Thus for the development of a process performance monitoring scheme, it was necessary to provide a common sampling base.
This paper examines how the dynamic information present in a multivariate set of data can be used to improve the performance of a monitoring scheme of the multivariate mean. Optimum performance of such a scheme requir...
详细信息
This paper examines how the dynamic information present in a multivariate set of data can be used to improve the performance of a monitoring scheme of the multivariate mean. Optimum performance of such a scheme requires the use of the covariance/correlation matrix of the time-lagged variables. In the general case, this would require variables to be lagged with respect to one another. A Statistical processcontrol scheme is described where the Mahalanobis distance is monitored based on optimally lagged variables. The benefits of the method are demonstrated using Monte Carlo simulations.
This study contributes to the comparison of partial least squares (PLS) and canonical variate analysis (CVA) for the identification of dynamic systems. Two model forms, autoregressive with exogenous inputs and state s...
详细信息
This study contributes to the comparison of partial least squares (PLS) and canonical variate analysis (CVA) for the identification of dynamic systems. Two model forms, autoregressive with exogenous inputs and state space representations, are developed with PLS and CVA being used to calculate the model parameters. The different models are compared using two case studies: a benchmark simulation of a binary distillation column and an industrial fluidised bed reactor.
Previous publications have focused upon the application of principal components analysis (PCA) for sensor fault identification through data reconstruction. These reconstruction based methods do not address the problem...
详细信息
Previous publications have focused upon the application of principal components analysis (PCA) for sensor fault identification through data reconstruction. These reconstruction based methods do not address the problem of fault propagation to other sensor measurements and as a consequence misleading fault identification can result. A novel method for the identification of multiple sensor disturbances during process monitoring is proposed based upon the T 2 -statistic. By using the T 2 -statistic in conjunction with the associated principal component score contribution plot, multiple sensor disturbances can be identified in a sequential manner. The methodology is demonstrated on two industrial plant simulations.
Traditionally Principal Components Analysis (PCA) has been viewed as a single group method and in particular, in multivariate statistical performance monitoring (MSPM), it has been used as a monitoring and diagnostic ...
详细信息
Traditionally Principal Components Analysis (PCA) has been viewed as a single group method and in particular, in multivariate statistical performance monitoring (MSPM), it has been used as a monitoring and diagnostic tool for single product production. An extension to PCA is presented which enables a number of similar products or product grades to be monitored through a single multi-group model. The method is applied to a semi-discrete industrial batch manufacturing process. The industrial application illustrates that the detection and diagnostic capabilities of the generic model are comparable to those of a single group model.
暂无评论