Ratna Babu Chinnam

HMMs for diagnostics and prognostics in machining processes

Despite considerable advances over the last two decades in sensing instrumentation and information technology infrastructure, monitoring and diagnostics technology has not yet found its place in health management of mainstream machinery and equipment. This is in spite of numerous studies reporting that the expected savings from widespread deployment of condition-based maintenance (CBM) technology would be in the tens of billions of dollars in many industrial sectors as well as in governmental agencies. It turns out that a prerequisite to widespread deployment of CBM technology and practice in industry is cost efficient and effective diagnostics and prognostics. This paper presents a novel method for employing hidden Markov models (HMMs) for carrying out both diagnostic as well as prognostic activities for metal cutting tools. The methods employ HMMs for modelling sensor signals emanating from the machine (or features thereof), and in turn, identify the health state of the cutting tool as well as facilitate estimation of remaining useful life. This paper also investigates some of the underlying issues of proper HMM design and training for the express purpose of effective diagnostics and prognostics. The proposed methods were validated on a physical test-bed, a vertical drilling machine. Experimental results are very promising.

mr2PSO: A maximum relevance minimum redundancy feature selection method based on swarm intelligence for support vector machine classification

This paper presents a hybrid filter-wrapper feature subset selection algorithm based on particle swarm optimization (PSO) for support vector machine (SVM) classification. The filter model is based on the mutual information and is a composite measure of feature relevance and redundancy with respect to the feature subset selected. The wrapper model is a modified discrete PSO algorithm. This hybrid algorithm, called maximum relevance minimum redundancy PSO (mr^2PSO), is novel in the sense that it uses the mutual information available from the filter model to weigh the bit selection probabilities in the discrete PSO. Hence, mr^2PSO uniquely brings together the efficiency of filters and the greater accuracy of wrappers. The proposed algorithm is tested over several well-known benchmarking datasets. The performance of the proposed algorithm is also compared with a recent hybrid filter-wrapper algorithm based on a genetic algorithm and a wrapper algorithm based on PSO. The results show that the mr^2PSO algorithm is competitive in terms of both classification accuracy and computational performance.

A fuzzy logic based approach to reliability improvement estimation during product development

Abstract During early stages of product development process, a vast amount of knowledge and information is generated. However, most of it is subjective (imprecise) in nature and remains unutilized. This paper presents a formal structure for capturing this information and knowledge and utilizing it in reliability improvement estimation. The information is extracted as improvement indices from various design tools, experiments, and design review records and treated as fuzzy numbers or linguistic variables. Fuzzy reasoning method is used to combine and quantify the subjective information to map their impact on product reliability. The crisp output of the fuzzy reasoning process is treated as new evidence and incorporated into a Bayesian framework to update the reliability estimates. A case example is presented to demonstrate the proposed approach.

A neuro-fuzzy approach for estimating mean residual life in condition-based maintenance systems

This paper presents a framework for online reliability estimation of physical systems utilising degradation signals. Most prognostics methods promoted in the literature for estimation of mean-residual-life of individual components utilise trending or forecasting models in combination with mechanistic or empirical failure definition models. In the absence of sound knowledge for the mechanics of degradation and/or adequate failure data, it is not possible to establish practical failure definition models. However, if there exist domain experts with strong experiential knowledge, one can establish fuzzy inference models for failure definition. This paper presents a neuro-fuzzy approach for performing prognostics under such circumstances. The proposed approach is evaluated on a cutting tool monitoring problem. In particular, the method is used to monitor high-speed-steel drill-bits used for drilling holes in stainless steel metal plates.

Support vector machines for recognizing shifts in correlated and other manufacturing processes

Traditional statistical process control (SPC) techniques of control charting are not applicable in many process industries where the data from the facilities are often autocorrelated. This is often true in piece-part manufacturing industries that are highly automated and integrated. Several attempts have been made in the literature to extend traditional SPC techniques to deal with autocorrelated parameters. However, these extensions pose several serious limitations. The literature discusses several machine-learning methods based on radial basis function (RBF) networks and multi-layer perceptron (MLP) networks to address the limitations, with some success. This paper demonstrates that support vector machines (SVMs) can be extremely effective in minimizing both Type-I errors (probability that the method would wrongly declare the process to be out of control or generate a false alarm) and Type-II errors (probability that the method will be unable to detect a true shift or trend present in the process) in these autocorrelated processes. Even while employing the simplest type of polynomial kernels, the SVMs were extremely good at detecting shifts in papermaking and viscosity datasets (available in the literature) and performed as well or better than traditional as well as machine learning methods. It was also observed that SVMs are good at minimizing both Type-I and Type-II errors even in monitoring non-correlated processes. When tested on datasets available in the literature, they once again performed as well or better than the classical Shewhart control charts and other machine learning methods.

MASCF: A generic process-centered methodological framework for analysis and design of multi-agent supply chain systems

Multi-agent systems (MAS) are becoming popular for modeling complex systems such as supply chains. However, development of multi-agent systems remain quite involved and extremely time consuming. Currently, there exist no generic methodologies for modeling supply chains using multi-agent systems. In this research, we propose a generic process-centered methodological framework, Multi-Agent Supply Chain Framework (MASCF), to simplify MAS development for supply chain (SC) applications. MASCF introduces the notion of process-centered organization metaphor, and creatively adopts Supply Chain Operations Reference (SCOR) model to a well-structured generic MAS analysis and design methodology, Gaia, for multi-agent supply chain system (MASCS) development. The popular Tamagotchi case was designed and analyzed using MASCF. The validity of the framework was established by implementing MASCF output of Tamagotchi SC using the Java Agent DEvelopment Framework (JADE).

An Industrial Strength Novelty Detection Framework for Autonomous Equipment Monitoring and Diagnostics

This paper presents a practical framework for autonomous monitoring of industrial equipment based on novelty detection. It overcomes limitations of current equipment monitoring technology by developing a “generic” structure that is relatively independent of the type of physical equipment under consideration. The kernel of the proposed approach is an “evolving” model based on unsupervised learning methods (reducing the need for human intervention). The framework employs procedures designed to temporally evolve the critical model parameters with experience for enhanced monitoring accuracy (a critical ability for mass deployment of the technology on a variety of equipment/hardware without needing extensive initial tune-up). Proposed approach makes explicit provision to characterize the distinct operating modes of the equipment, when necessary, and provides the ability to predict both abrupt as well as gradually developing (incipient) changes. The framework is realized as an autonomous software agent that continuously updates its decision model implementing an unsupervised recursive learning algorithm. Results of validation of the proposed methodology by accelerated testing experiments are also discussed.

Dynamic routing under recurrent and non-recurrent congestion using real-time ITS information

In just-in-time (JIT) manufacturing environments, on-time delivery is a key performance measure for dispatching and routing of freight vehicles. Growing travel time delays and variability, attributable to increasing congestion in transportation networks, are greatly impacting the efficiency of JIT logistics operations. Recurrent and non-recurrent congestion are the two primary reasons for delivery delay and variability. Over 50% of all travel time delays are attributable to non-recurrent congestion sources such as incidents. Despite its importance, state-of-the-art dynamic routing algorithms assume away the effect of these incidents on travel time. In this study, we propose a stochastic dynamic programming formulation for dynamic routing of vehicles in non-stationary stochastic networks subject to both recurrent and non-recurrent congestion. We also propose alternative models to estimate incident induced delays that can be integrated with dynamic routing algorithms. Proposed dynamic routing models exploit real-time traffic information regarding speeds and incidents from Intelligent Transportation System (ITS) sources to improve delivery performance. Results are very promising when the algorithms are tested in a simulated network of South-East Michigan freeways using historical data from the MITS Center and Traffic.com.

Autonomous diagnostics and prognostics through competitive learning driven HMM-based clustering

A prerequisite to effective wide-spread deployment of condition-based maintenance (CBM) practices is effective diagnostics and prognostics. This paper presents a novel method for employing HMMs for autonomous diagnostics as well as prognostics. The diagnostics module exploits competitive learning to achieve HMM-based clustering. The prognostics module builds upon the diagnostics module to compute joint distributions for health-state transition times. The proposed methods were validated on a physical test bed; a drilling machine.

Efficient exact optimization of multi-objective redundancy allocation problems in series-parallel systems

This paper proposes a decomposition-based approach to exactly solve the multi-objective Redundancy Allocation Problem for series-parallel systems. Redundancy allocation problem is a form of reliability optimization and has been the subject of many prior studies. The majority of these earlier studies treat redundancy allocation problem as a single objective problem maximizing the system reliability or minimizing the cost given certain constraints. The few studies that treated redundancy allocation problem as a multi-objective optimization problem relied on meta-heuristic solution approaches. However, meta-heuristic approaches have significant limitations: they do not guarantee that Pareto points are optimal and, more importantly, they may not identify all the Pareto-optimal points. In this paper, we treat redundancy allocation problem as a multi-objective problem, as is typical in practice. We decompose the original problem into several multi-objective sub-problems, efficiently and exactly solve sub-problems, and then systematically combine the solutions. The decomposition-based approach can efficiently generate all the Pareto-optimal solutions for redundancy allocation problems. Experimental results demonstrate the effectiveness and efficiency of the proposed method over meta-heuristic methods on a numerical example taken from the literature.