The performance of a control chart in statistical processcontrol is often quantified in terms of the Average Run Length (ARL). The ARL enables a comparison to be undertaken between various monitoring strategies. Thes...
详细信息
process Analytical Technologies (PAT) are increasingly being explored and adopted by pharma-chem and bio-pharma companies for enhanced process understanding, Quality by Design (QbD) and Real-time-Release (RTR). To ach...
详细信息
Many fed-batch processes can be considered as a class of control-affine nonlinear systems. In this paper, a new methodology of neural networks, called the control Affine Feedforward Neural Network (CAFNN), is proposed...
详细信息
ISBN:
(纸本)0780372980
Many fed-batch processes can be considered as a class of control-affine nonlinear systems. In this paper, a new methodology of neural networks, called the control Affine Feedforward Neural Network (CAFNN), is proposed. It can be trained easily. For constrained nonlinear optimization problems, it offers an effective and simple optimal control strategy by sequential quadratic programming in which the analytic gradient information can be computed directly. The proposed modeling and optimal control schemes are illustrated on an ethanol fermentation process. Compared with a general multilayer neural network, the nonlinear programming problem based on a CAFNN model is solved more accurately and efficiently.
The detection of process changes through a principal component analysis based monitoring scheme can be achieved through the interrogation of two metrics, Hotelling's T/sup 2/ and the Q-statistic. The Q-statistic h...
详细信息
The detection of process changes through a principal component analysis based monitoring scheme can be achieved through the interrogation of two metrics, Hotelling's T/sup 2/ and the Q-statistic. The Q-statistic has been shown to be insensitive to small changes in the process model parameters. In this paper, a modified statistic based on the local approach is proposed to detect changes in model parameters in a principal component analysis monitoring scheme. The performance of the more traditional Q-statistic is compared with the modified statistic through their application to fault detection in a continuous stiffed tank reactor.
This paper presents an analysis of nonlinear extensions to Partial Least Squares (PLS) using error-based minimization techniques. The analysis revealed that such algorithms are maximizing the accuracy with which the r...
详细信息
作者:
Puneet MishraAlison NordonWestCHEM
Department of Pure and Applied Chemistry and Centre for Process Analytics and Control Technology University of Strathclyde Glasgow UK
Calibration-free resolution techniques provide an alternative approach to the development of a calibration model. These combine spectroscopic measurement coupled with mathematical and statistical assumptions and give ...
详细信息
Larimore's state space model derivation and stochastic estimation algorithm, first published in 1983, have been the adopted standard for deriving the state variables and parameters of the five (5) matrices state s...
详细信息
Larimore's state space model derivation and stochastic estimation algorithm, first published in 1983, have been the adopted standard for deriving the state variables and parameters of the five (5) matrices state space model representation which continues to be applied extensively in the literature for applications ranging from controls, system identification and process monitoring. This paper presents an alternate derivation and stochastic estimation algorithm. The paper also discusses how strategic classification of the process inputs may, for some applications, facilitate the use of a simplified stochastic estimation algorithm. The alternative state space modeling approaches demonstrated better fault monitoring statistic performance for specific types of faults simulated. The canonical variate based state space modeling approaches were evaluated on a simulate CSTR process – with recycle through a heat exchanger. The results demonstrates the potential benefits to be derived from using a combined monitoring index based upon monitoring statistics derived from independent state space models for improved overall fault detecting capabilities and reliability of the fault monitoring scheme.
A hybrid recurrent neural network model based on-line re-optimisation control strategy is developed for batch polymerisation reactors. The hybrid model contains a simplified mechanistic model covering material balance...
详细信息
A hybrid recurrent neural network model based on-line re-optimisation control strategy is developed for batch polymerisation reactors. The hybrid model contains a simplified mechanistic model covering material balance and simplified reaction kinetics only and recurrent neural networks. Based on this hybrid neural network model, optimal control policy can be calculated. A difficulty in the optimal control of batch polymerisation reactors is that optimisation effort can be seriously hampered by unknown disturbances such as reactive impurities and reactor fouling. A technique for on-line estimation of reactive impurity and reactor fouling has been developed by Zhang et al. (1999). In this contribution, on-line reactive impurity estimation is combined with batch reactor optimal control to form a novel re-optimisation control strategy. When there exists an unknown amount of reactive impurities, the off-line calculated optimal control profile will be no longer optimal. On-line impurity estimation is applied to estimate the amount of reactive impurities during the early stage of the batch. Based on the estimated amount of reactive impurities, on-line re-optimisation is applied to calculate the optimal reactor temperature profile for the remaining time period of the batch reactor operation. This approach is illustrated on the optimisation control of a simulated batch MMA polymerisation process.
A training method for enhancing neural network model generalisation is proposed. In this method, a neural network is trained and tested alternatively on a training data set and a testing data set. Unlike in convention...
详细信息
A training method for enhancing neural network model generalisation is proposed. In this method, a neural network is trained and tested alternatively on a training data set and a testing data set. Unlike in conventional neural network training where the training and testing data sets are fixed, the training and testing data sets swap roles continuously during network training. Training is terminated when the network prediction errors on both data sets cannot be further reduced. Application examples demonstrate that this neural network training strategy can significantly improve the neural tu network model prediction accuracy, especially the long range prediction accuracy.
暂无评论