A non-linear principal components analysis algorithm is proposed for process performance monitoring. Before assessing the capabilities of the monitoring scheme on an industrial dryer, the data is first pre-processed t...
详细信息
A non-linear principal components analysis algorithm is proposed for process performance monitoring. Before assessing the capabilities of the monitoring scheme on an industrial dryer, the data is first pre-processed through application of a wavelet filter to remove noise and spikes. The wavelet coefficients are then used as the inputs into the non-linear algorithm. Performance monitoring charts with non-parametric control limits were then constructed to identify the occurrence of non-conforming operation prior to interrogating non-linear contribution plots to help identify the potential source of the fault. Encouraging results were achieved.
Neural component analysis (NCA) is one of the latest nonlinear multivariate statistical process control (MSPC) methods, which consists in combining artificial neural networks (ANNs) with principal component analysis (...
详细信息
Neural component analysis (NCA) is one of the latest nonlinear multivariate statistical process control (MSPC) methods, which consists in combining artificial neural networks (ANNs) with principal component analysis (PCA). However, NCA cannot handle the non-Gaussian feature and the extracted principal components (PCs) in NCA may not be the key information in the process data. Herein, we propose an improved NCA (INCA) which introduces a new cost function based on kurtosis to restrict the Gaussianity of PCs. We also propose a novel PC selection mechanism based on the information of PCs in the original data space rather than in the PC data space. INCA achieves almost 100% detection rates in three different types of faults in a simulation model test, and it can detect the fault in the thermal power plant process more than 1 min ahead of orthogonal nonlinear PCA (ONLPCA) and NCA.
In this study, a new data-driven multivariate multiscale statisticalprocess monitoring method based on singular spectrum analysis (SSA) and empirical mode decomposition (EMD) is proposed for fault detection in chemic...
详细信息
In this study, a new data-driven multivariate multiscale statisticalprocess monitoring method based on singular spectrum analysis (SSA) and empirical mode decomposition (EMD) is proposed for fault detection in chemical process systems. SSA extracts the trends of process signals using the eigenvalues of trajectory matrices while EMD uses the intrinsic mode functions (IMFs) to capture the signal trends through sifting process. The results obtained from the industrial and simulated case studies showed that SSA and conventional multivariatestatisticalprocess monitoring technique such as principal component analysis (PCA) failed to extract the nonstationary and nonlinear trends in the signal effectively. As an alternative, in this study, SSA is combined with EMD decomposition prior to the process monitoring procedure using PCA. The efficiency of EMD in analyzing the nonstationary and nonlinear signals enhanced the performance of linear SSA techniques by combining the two techniques in this study. Experimental and simulation results also revealed that fault detection using EMD is comparable to the combined technique.
Most applications of MSPC have tended to focus upon the manufacture of a single product with separate models being developed to monitor individual recipes. With process manufacturing trends being influenced by custome...
详细信息
Most applications of MSPC have tended to focus upon the manufacture of a single product with separate models being developed to monitor individual recipes. With process manufacturing trends being influenced by customer demands there has been an increase in the manufacture of a wide variety of products, there is a real need for process models which allow a range of products, grades or recipes to be monitored using a single process model. With increasing attention now being paid to the FDA process Analytical Technologies (PAT) initiative, the use of spectro-chemical information for enhanced monitoring of reactions and is now gaining impetus. An application of the performance monitoring of a multi-recipe multi-reactor industrial batch polymer manufacturing is discussed in which NIR spectroscopic data is also integrated with process data to provide enhanced batch monitoring.
Three - way data collected from batch processes and from transitions of continuous processes are dynamic in nature; the process variables in such processes are both auto correlated and cross correlated. Empirical mode...
详细信息
Three - way data collected from batch processes and from transitions of continuous processes are dynamic in nature; the process variables in such processes are both auto correlated and cross correlated. Empirical models developed for the statisticalprocesscontrol of these processes should be capable of capturing the auto and cross correlation of the process variables. statisticalprocesscontrol checks deviations from a nominal behaviour. Therefore for the statisticalprocesscontrol of batch processes and transitions we should look at deviations of process variable trajectories from their nominal trajectories and from their nominal auto/cross correlations. This paper addresses issues related to modelling three way data collected from such processes using projection methods, such as principal component analysis (PCA) and partial least squares (PLS).
Abstract A new fault detection method is proposed to realize adaptive high-performance monitoring and efficient maintenance of the system. The proposed method is data-driven and called Just-In-Time statisticalprocess...
详细信息
Abstract A new fault detection method is proposed to realize adaptive high-performance monitoring and efficient maintenance of the system. The proposed method is data-driven and called Just-In-Time statisticalprocesscontrol (JIT-SPC). JIT-SPC focuses on the distance from the current operation data to the normal operation data stored in the database, because fault detection depends essentially on whether normal data exist near the current data or not. Since JIT-SPC is a model-free technique without impractical assumptions that conventional methods make, it has a potential for realizing practical, adaptive, high-performance monitoring. In addition, fault identification can be done through contribution plot in the framework of JIT-SPC. The usefulness of JIT-SPC and its contribution plot is demonstrated through a case study of the vinyl acetate monomer process. The results show that JIT-SPC can cope with changes in operating condition and can detect faults earlier than the conventional MSPC.
The sub-prime crisis started from November 2006 to February 2008 is a global crisis that affected almost all economy activities in the world. In this study, we used the covariance stability test for exploring its impa...
详细信息
The sub-prime crisis started from November 2006 to February 2008 is a global crisis that affected almost all economy activities in the world. In this study, we used the covariance stability test for exploring its impact towards foreign exchange rate among 15 currencies. Box?s M control chart and its root causes analysis are employed to understand the behaviour and interrelationship of FOREX?s structure among America and Europe continents. From the analysis, it shows that the structures of covariance from Jan, 2006 to Dec, 2008 are not stable. To be detail, if there is any shift on USD during April-June 2007, the nearest currencies that will received the impact are Argentine Peso, Chilean Peso and Rusia Ruble.
The ever-present drive to safer, more cost-effective and cleaner processes motivates the exploration of a variety of process monitoring methods. In the domain of data-driven approaches, random forest models present a ...
详细信息
The ever-present drive to safer, more cost-effective and cleaner processes motivates the exploration of a variety of process monitoring methods. In the domain of data-driven approaches, random forest models present a nonlinear framework. Random forest models consist of ensembles of classification and regression trees in which the model response is determined by voting committees of independent binary decision trees. Data-driven approaches to fault diagnosis often involve summarizing potentially large numbers of process variables in lower dimensional diagnostic sequences. Random forest feature extraction allows for the monitoring of process in feature and residual spaces, while random forest variable importance measures can potentially be used to identify process variables contribution to fault conditions. In this study, a framework for diagnosing steady state faults with random forests is proposed and demonstrated with a simple nonlinear system and the benchmark Tennessee Eastman process.
Principal component analysis (PCA) has been applied widely for monitoring plant performance across a range of industrial processes. PCA is a linear technique and it is therefore not strictly applicable for handling in...
详细信息
Principal component analysis (PCA) has been applied widely for monitoring plant performance across a range of industrial processes. PCA is a linear technique and it is therefore not strictly applicable for handling industrial problems which exhibit significant non-linear behaviour. A novel non-linear PCA method is proposed based upon the Input-Training neural network. multivariate statistical process control charts with non-parametric control limits are then defined to overcome the limitations of the conventional approach of defining the limits based upon the assumption of normality. A contribution plot capable of identifying the potential source of the fault in a non-linear situation is then proposed prior to applying the methodology to a continuous industrial reactor. (C) 1998 Elsevier Science Ltd. All rights reserved.
暂无评论