Bhavik R. Bakshi

Particle filtering and moving horizon estimation

This paper provides an overview of currently available methods for state estimation of linear, constrained and nonlinear systems. The following methods are discussed: Kalman filtering, extended Kalman filtering, unscented Kalman filtering, particle filtering, and moving horizon estimation. The current research literature on particle filtering and moving horizon estimation is reviewed, and the advantages and disadvantages of these methods are presented. Topics for new research are suggested that address combining the best features of moving horizon estimation and particle filters.

Comparison of Multivariate Statistical Process Monitoring Methods with Applications to the Eastman Challenge Problem

Abstract To improve the performance of multivariate statistical process control (MSPC), two advanced methods, moving principal component analysis (MPCA) and DISSIM, have been proposed. In MPCA and DISSIM, an abnormal operation can be detected by monitoring the directions of principal components and the dissimilarity index, respectively. Another important extension of MSPC was made with Multiscale PCA (MS-PCA). The present work investigates the characteristics of several statistical monitoring methods. The monitoring performance is compared with applications to simulated data obtained from a 2×2 process and the Tennessee Eastman process. The superiority of MPCA and DISSIM over the conventional methods comes from the fact that those methods focus on changes in the distribution of process data. Furthermore, the advantages of MPCA or DISSIM over the conventional MSPC and that of MS-PCA are combined, and new methods, termed MS-MPCA and MS-DISSIM, are proposed.

A thermodynamic framework for ecologically conscious process systems engineering

Long term growth and well being of the chemical industry requires economically and ecologically conscious process engineering. Traditional process engineering methods fall short of meeting this need due to their treating the environment as being secondary to economic objectives. Life cycle assessment and design methods broaden the scope of traditional methods, but focus primarily on the environmental impact of emissions while ignoring the contribution of ecological products and services. Taking nature for granted could provide misleading results since natural products and processes are a significant contributor to all industrial products and processes. Methods from systems ecology do account for ecological inputs, but ignore the impact of emissions. This paper presents an original approach for the joint analysis of industrial and ecological systems. This approach considers inputs from both ecological and economic resources, as well as the impact of emissions. It uses thermodynamics to exploit the synergy between methods from process systems engineering, systems ecology, and life cycle assessment to overcome the shortcomings of methods from each field. The approach is based on the fact that growth and sustenance of both industrial and ecological processes are limited by the available energy and its conversion to useful work. Thus, the embodied energy (emergy), that is, the energy used directly or indirectly to make a product or service is a thermodynamic measure of ecological investment or cost, while exergy loss provides a holistic measure of the impact of emissions. Together, emergy and exergy analysis can provide insight into the environmental performance and sustainability of the industrial process or product. The proposed framework is broadly applicable to assist decision making in chemical and other engineering tasks. Challenges and research opportunities for making this framework practical are identified.

Accounting for ecosystem services in life cycle assessment, Part II: Toward an ecologically based LCA.

Despite the essential role of ecosystem goods and services in sustaining all human activities, they are often ignored in engineering decision making, even in methods that are meant to encourage sustainability. For example, conventional Life Cycle Assessment focuses on the impact of emissions and consumption of some resources. While aggregation and interpretation methods are quite advanced for emissions, similar methods for resources have been lagging, and most ignore the role of nature. Such oversight may even result in perverse decisions that encourage reliance on deteriorating ecosystem services. This article presents a step toward including the direct and indirect role of ecosystems in LCA, and a hierarchical scheme to interpret their contribution. The resulting Ecologically Based LCA (Eco-LCA) includes a large number of provisioning, regulating, and supporting ecosystem services as inputs to a life cycle model at the process or economy scale. These resources are represented in diverse physical units and may be compared via their mass, fuel value, industrial cumulative exergy consumption, or ecological cumulative exergy consumption or by normalization with total consumption of each resource or their availability. Such results at a fine scale provide insight about relative resource use and the risk and vulnerability to the loss of specific resources. Aggregate indicators are also defined to obtain indices such as renewability, efficiency, and return on investment. An Eco-LCA model of the 1997 economy is developed and made available via the web (www.resilience.osu.edu/ecolca). An illustrative example comparing paper and plastic cups provides insight into the features of the proposed approach. The need for further work in bridging the gap between knowledge about ecosystem services and their direct and indirect role in supporting human activities is discussed as an important area for future work.

Accounting for Ecosystem Services in Life Cycle Assessment, Part I: A Critical Review

If life cycle oriented methods are to encourage sustainable development, they must account for the role of ecosystem goods and services, since these form the basis of planetary activities and human well-being. This article reviews methods that are relevant to accounting for the role of nature and that could be integrated into life cycle oriented approaches. These include methods developed by ecologists for quantifying ecosystem services, by ecological economists for monetary valuation, and life cycle methods such as conventional life cycle assessment, thermodynamic methods for resource accounting such as exergy and emergy analysis, variations of the ecological footprint approach, and human appropriation of net primary productivity. Each approach has its strengths: economic methods are able to quantify the value of cultural services; LCA considers emissions and assesses their impact; emergy accounts for supporting services in terms of cumulative exergy; and ecological footprint is intuitively appealing and considers biocapacity. However, no method is able to consider all the ecosystem services, often due to the desire to aggregate all resources in terms of a single unit. This review shows that comprehensive accounting for ecosystem services in LCA requires greater integration among existing methods, hierarchical schemes for interpreting results via multiple levels of aggregation, and greater understanding of the role of ecosystems in supporting human activities. These present many research opportunities that must be addressed to meet the challenges of sustainability.

Multiscale analysis and modeling using wavelets

Measured data from most processes are inherently multiscale in nature owing to contributions from events occurring at different locations and with different localization in time and frequency. Consequently, data analysis and modeling methods that represent the measured variables at multiple scales are better suited for extracting information from measured data than methods that represent the variables at a single scale. This paper presents an overview of multiscale data analysis and empirical modeling methods based on wavelet analysis. These methods exploit the ability of wavelets to extract events at different scales, compress deterministic features in a small number of relatively large coefficients, and approximately decorrelate a variety of stochastic processes. Multiscale data analysis methods for off‐line and on‐line removal of Gaussian stationary noise eliminate coefficients smaller than a threshold. These methods are extended to removing non‐Gaussian errors by combining them with multiscale median filtering. Multiscale empirical modeling methods simultaneously select the most relevant features while determining the model parameters, and may provide more accurate and physically interpretable models. Copyright

Representation of process trends. III: Multiscale extraction of trends from process data

Abstract This paper presents a formal methodology for the analysis of process signals and the automatic extraction of temporal features contained in a record of measured data. It is based on the multiscale analysis of the measured signals using wavelets, which allows the extraction of significant temporal features that are localized in the frequency domain, from segments of the record of measured data (i.e. localized in the time domain). The paper provides a concise framework for the multiscale extraction and description of temporal process trends. The resulting algorithms are analytically sound, computationally very efficient and can be easily integrated with a large variety of methods for the interpretation of process trends and the automatic learning of relationships between causes and symptoms in a dynamic environment. A series of examples illustrate the characteristics of the approach and outline its use in various settings for the solution of industrial problems.

Representation of process trends—IV. Induction of real-time patterns from operating data for diagnosis and supervisory control

A methodology for pattern-based supervisory control and fault diagonsis is presented, based on the multi-scale extraction of trends from process data described in Part III of this series (Bakshi and Stephanopoulos, Computers Chem. Engng 17, 1993). An explicit mapping is learned between the features extracted at multiple scales, and the corresponding process conditions, using the technique of induction by decision trees. Simple rules may be derived from the induced decision tree, to relate the relevant qualitative or quantitative features in the measured process data to process conditions. These rules are often physically interpretable and provide physical insight into the process. Industrial case studies from fine chemicals manufacturing, reactive crystallization and fed-batch fermentation are used to illustrate the characteristics of the pattern-based learning methodology and its application to process supervision and diagnosis.

Comparison of statistical process monitoring methods: application to the Eastman challenge problem

Abstract Multivariate statistical process control (MSPC) has been successfully applied to chemical procesess. In order to improve the performance of fault detection, two kinds of advanced methods, known as moving principal component analysis (MPCA) and DISSIM, have been proposed. In MPCA and DISSIM, an abnormal operation can be detected by monitoring the directions of principal components (PCs) and the degree of dissimilarity between data sets, respectively. Another important extension of MSPC was made by using multiscale PCA (MS-PCA). In the present work, the characteristics of several monitoring methods are investigated. The monitoring performances are compared with using simulated data obtained from the Tennessee Eastman process. The results show that the advanced methods can outperform the conventional method. Furthermore, the advantage of MPCA and DISSIM over conventional MSPC (cMSPC) and that of the multiscale method are combined, and the new methods known as MS-MPCA and MS-DISSIM are proposed.

Brief paper: Bayesian estimation via sequential Monte Carlo sampling-Constrained dynamic systems

Nonlinear and non-Gaussian processes with constraints are commonly encountered in dynamic estimation problems. Methods for solving such problems either ignore the constraints or rely on crude approximations of the model or probability distributions. Such approximations may reduce the accuracy of the estimates since they often fail to capture the variety of probability distributions encountered in constrained linear and nonlinear dynamic systems. This article describes a practical approach that overcomes these shortcomings via a novel extension of sequential Monte Carlo (SMC) sampling or particle filtering. Inequality constraints are imposed by accept/reject steps in the algorithm. The proposed approach provides samples representing the posterior distribution at each time point, and is shown to satisfy the same theoretical properties as unconstrained SMC. Illustrative examples show that results of the proposed approach are at least as accurate as moving horizon estimation, but computationally more efficient and in addition, the approach indicates the uncertainty associated with these estimates.