S. Lakshminarayanan

Identification of Hammerstein models using multivariate statistical tools

The iterative Narendra-Gallman algorithm (NGA) for the identification of a nonlinear system representable by the Hammerstein structure is extended to perform simultaneous structure determination and parameter estimation of multivariable chemical process systems. The parameters of the linear system obtained in state space form using canonical correlations analysis and the coefficients of the polynomial type nonlinear elements are alternately adjusted, until convergence, to obtain the model. The theory is illustrated using data from an experimental heat exchanger and the simulation example of a realistically complex acid-base neutralization tank.

New product design via analysis of historical databases

Abstract A methodology is presented to define a set of operating conditions to produce a desired product, given a database of historical operating conditions and the product quality that they produced. This approach relies on the generation of a reliable model that can be used to predict the quality variables (the Y block) from the decision variables (the X block). Genetic programming (GP) is used to automatically generate accurate nonlinear models relating latent vectors for the X and Y blocks. The GP has the capability to carry out simultaneous optimization of model relationship structures and parameters, as well as to identify the most important basis functions. Once an adequate model is generated, it is used to predict the required process conditions to meet the new quality target by reverse mapping.

Critical evaluation of image processing approaches for real-time crystal size measurements

Abstract Monitoring and control of particulate processes is quite challenging and has evoked recent interest in the use of image-based approaches to estimate product quality (e.g. size, shape) in real-time and in situ. Crystal size estimation from video images, especially for high aspect-ratio systems, has received much attention. In spite of the increased research activity in this area, there is little or no work that demonstrates and quantifies the success of the image analysis (IA) techniques to any reasonable degree. This is important because, although image analysis techniques are well developed, the quality of images from inline sensors is variable and often poor, leading to incorrect estimation of the process state. The present paper, to our knowledge, the first large-scale size estimation study with Lasentecs in-process video imaging system, PVM, seeks to fill this void by focusing on one key step in IA viz. segmentation. Using manual segmentation of particles as an independent measure of the particle size, we have devised metrics to compare the accuracy of automated segmentation during IA. These metrics provide a quantitative measure of the quality of results. Based on these metrics, a sensitivity study of IA parameters has also been performed and “optimal” parameter settings identified. A Monosodium Glutamate seeded cooling crystallization process is used to illustrate that, with proper settings, IA can be used to accurately track the size within ∼8% error.

Data-Driven Optimization of Metabolomics Methods Using Rat Liver Samples

The aim of metabolomics is to identify, measure, and interpret complex time-related concentration, activity, and flux of metabolites in cells, tissues, and biofluids. We have used a metabolomics approach to study the biochemical phenotype of mammalian cells which will help in the development of a panel of early stage biomarkers of heat stress tolerance and adaptation. As a first step, a simple and sensitive mass spectrometry experimental workflow has been optimized for the profiling of metabolites in rat tissues. Sample (liver tissue) preparation consisted of a homogenization step in three different buffers, acidification with different strengths of acids, and solid-phase extraction using nine types of cartridges of varying specificities. These led to 18 combinations of acids, cartridges, and buffers for testing for positive and negative ions using mass spectrometry. Results were analyzed and visualized using algorithms written in MATLAB v7.4.0.287. By testing linearity, repeatability, and implementation of univariate and multivariate data analysis, a robust metabolomics platform has been developed. These results will form a basis for future applications in discovering metabolite markers for early diagnosis of heat stress and tissue damage.

Monitoring Batch Processes Using Multivariate Statistical Tools: Extensions and Practical Issues

Abstract Extensions and practical issues in the application of a statistical technique, namely Partial Least Squares (PLS), to the monitoring, product quality prediction and fault detection of batch/semibatch processes is considered. The approach of Nomikos and MacGregor (1994a, 1994b) is explored further to include : (1) multirate sampling and (2) normal batches with varying run lengths. The theoretical development presented here is illustrated using simulated data from a fed-batch bioreactor.

Variable predictive models-A new multivariate classification approach for pattern recognition applications

Many pattern recognition algorithms applied in literature exhibit data specific performances and are also computationally intense and complex. The data classification problem poses further challenges when different classes cannot be distinguished just based on decision boundaries or conditional discriminating rules. As an alternate to existing methods, inter-relations among the feature vectors can be exploited for distinguishing samples into specific classes. Based on this idea, variable predictive model based class discrimination (VPMCD) method is proposed as a new and alternative classification approach. Analysis is carried out using seven well studied data sets and the performance of VPMCD is benchmarked against well established linear and non-linear classifiers like LDA, kNN, Bayesian networks, CART, ANN and SVM. It is demonstrated that VPMCD is an efficient supervised learning algorithm showing consistent and good performance over these data sets. The new VPMCD method has the potential to be effectively and successfully extended to many pattern recognition applications of recent interest.

Algorithm Note: Variable predictive model based classification algorithm for effective separation of protein structural classes

Variable predictive model based class discrimination (VPMCD) algorithm is proposed as an effective protein secondary structure classification tool. The algorithm mathematically represents the characteristics amino acid interactions specific to each protein structure and exploits them further to distinguish different structures. The new concept and the VPMCD classifier are established using well-studied datasets containing four protein classes as benchmark. The protein samples selected from SCOP and PDB databases with varying homology (25-100%) and non-uniform distribution of class samples provide challenging classification problem. The performance of the new method is compared with advanced classification algorithms like component coupled, SVM and neural networks. VPMCD provides superior performance for high homology datasets. 100% classification is achieved for self-consistency test and an improvement of 5% prediction accuracy is obtained during Jackknife test. The sensitivity of the new algorithm is investigated by varying model structures/types and sequence homology. Simpler to implement VPMCD algorithm is observed to be a robust classification technique and shows potential for effective extensions to other clinical diagnosis and data mining applications in biological systems.

Empirical modelling and control of processes with recycle: some insights via case studies

Abstract Issues related to the identification and control of systems with recycle of material and/or energy are investigated via simple case studies. We aim for high performance control using recycle compensators that can be easily implemented on industrial DCS systems. The modelling strategy centers around system identification and/or decomposing the overall process dynamics into a direct effect component and a recycle component. When fundamental process models are available, such decomposition is relatively straightforward. In their absence, the models required for controller design may only be gained via plant tests and subsequent system identification. This study will examine the issues related to the empirical identification of integrated systems such as the proper use of available sensor information and appropriate implementation of the recycle compensator. Though only two case studies are presented here, the principles and conclusions are broadly applicable to industrial chemical process systems.

VPMCD: variable interaction modeling approach for class discrimination in biological systems.

Data classification algorithms applied for class prediction in computational biology literature are data specific and have shown varying degrees of performance. Different classes cannot be distinguished solely based on interclass distances or decision boundaries. We propose that inter‐relations among the features be exploited for separating observations into specific classes. A new variable predictive model based class discrimination (VPMCD) method is described here. Three well established and proven data sets of varying statistical and biological significance are utilized as benchmark. The performance of the new method is compared with advanced classification algorithms. The new method performs better during different tests and shows higher stability and robustness. The VPMCD is observed to be a potentially strong classification approach and can be effectively extended to other data mining applications involving biological systems.

Closed loop identification and control loop reconfiguration: an industrial case study

Abstract The results of a joint university–industry collaborative project for control loop reconfiguration using closed loop experimental data from a fuel gas pressure control system are described in this paper. The fuel gas pressure was being regulated using a butane stream. For economic reasons, it was necessary to switch control to the ethane stream. Previous attempts at effecting this changeover had proved unsuccessful. In this study, a powerful system identification technique namely Canonical Variate Analysis (CVA) was employed to obtain the empirical plant models. A PI controller was then designed using the direct synthesis method. Acceptable closed loop behavior was obtained with little online tuning.