Kihoon Choi

Data-Driven Modeling, Fault Diagnosis and Optimal Sensor Selection for HVAC Chillers

Chillers constitute a significant portion of energy consumption equipment in heating, ventilating and air-conditioning (HVAC) systems. The growing complexity of building systems has become a major challenge for field technicians to troubleshoot the problems manually; this calls for automated ldquosmart-service systemsrdquo for performing fault detection and diagnosis (FDD). The focus of this paper is to develop a generic FDD scheme for centrifugal chillers and also to develop a nominal data-driven (ldquoblack-boxrdquo) model of the chiller that can predict the system response under new loading conditions. In this vein, support vector machines, principal component analysis, and partial least squares are the candidate fault classification techniques in our approach. We present a genetic algorithm-based approach to select a sensor suite for maximum diagnosabilty and also evaluated the performance of selected classification procedures with the optimized sensor suite. The responses of these selected sensors are predicted under new loading conditions using the nominal model developed via the black-box modeling approach. We used the benchmark data on a 90-t real centrifugal chiller test equipment, provided by the American Society of Heating, Refrigerating and Air-Conditioning Engineers, to demonstrate and validate our proposed diagnostic procedure. The database consists of data from sixty four monitored variables of the chiller under 27 different modes of operation during nominal and eight faulty conditions with different severities.

Dynamic Multiple Fault Diagnosis: Mathematical Formulations and Solution Techniques

Imperfect test outcomes, due to factors such as unreliable sensors, electromagnetic interference, and environmental conditions, manifest themselves as missed detections and false alarms. This paper develops near-optimal algorithms for dynamic multiple fault diagnosis (DMFD) problems in the presence of imperfect test outcomes. The DMFD problem is to determine the most likely evolution of component states, the one that best explains the observed test outcomes. Here, we discuss four formulations of the DMFD problem. These include the deterministic situation corresponding to perfectly observed coupled Markov decision processes to several partially observed factorial hidden Markov models ranging from the case where the imperfect test outcomes are functions of tests only to the case where the test outcomes are functions of faults and tests, as well as the case where the false alarms are associated with the nominal (fault free) case only. All these formulations are intractable NP-hard combinatorial optimization problems. Our solution scheme can be viewed as a two-level coordinated solution framework for the DMFD problem. At the top (coordination) level, we update the Lagrange multipliers (coordination variables, dual variables) using the subgradient method. At the bottom level, we use a dynamic programming technique (specifically, the Viterbi decoding or Max-sum algorithm) to solve each of the subproblems, one for each component state sequence. The key advantage of our approach is that it provides an approximate duality gap, which is a measure of the suboptimality of the DMFD solution. Computational results on real-world problems are presented. A detailed performance analysis of the proposed algorithm is also discussed.

Novel Classifier Fusion Approaches for Fault Diagnosis in Automotive Systems

Faulty automotive systems significantly degrade the performance and efficiency of vehicles, and oftentimes are the major contributors of vehicle breakdown; they result in large expenditures for repair and maintenance. Therefore, intelligent vehicle health-monitoring schemes are needed for effective fault diagnosis in automotive systems. Previously, we developed a data-driven approach using a data reduction technique, coupled with a variety of classifiers, for fault diagnosis in automotive systems. In this paper, we consider the problem of fusing classifier decisions to reduce diagnostic errors. Specifically, we develop three novel classifier fusion approaches: class-specific Bayesian fusion, joint optimization of fusion center and of individual classifiers, and dynamic fusion. We evaluate the efficacies of these fusion approaches on an automotive engine data. The results demonstrate that dynamic fusion and joint optimization, and class-specific Bayesian fusion outperform traditional fusion approaches. We also show that learning the parameters of individual classifiers as part of the fusion architecture can provide better classification performance.

Data Reduction Techniques for Intelligent Fault Diagnosis in Automotive Systems

Faults in automotive systems significantly degrade the performance and efficiency of vehicles, and often times are the major causes of vehicle break-down leading to large expenditure for repair and maintenance. An intelligent fault diagnosis system can ensure uninterrupted and reliable operation of vehicular systems, and aid in vehicle health monitoring. Due to cost constraints, the current electronic control units (ECUs) for control and diagnosis have 1-2 MB of memory and 24 -50 MHz of processor speed. Therefore, intelligent data reduction techniques and partitioning methodology are needed for effective fault diagnosis in automotive systems. In this paper, we propose a data- driven approach using a data reduction technique, coupled with a variety of classifiers, for an automotive engine system. Adaptive boosting (AdaBoost) is employed to improve the classifier performance. Our proposed techniques can be used for any vehicle systems without the need to tune the classification algorithms for a specific vehicle model. Our proposed fault diagnosis scheme results in significant reduction in data size (25.6 MBrarr12.8 KB) without loss of accuracy in classification.

Data Visualization, Data Reduction and Classifier Fusion for Intelligent Fault Diagnosis in Gas Turbine Engines

In this paper, we investigate four key issues associated with data-driven approaches for fault classification using the Pratt and Whitney commercial dual-spool turbofan engine data as a test case. The four issues considered here include the following. (1) Can we characterize, a priori, the difficulty of fault classification via self-organizing maps? (2) Do data reduction techniques improve fault classification performance and enable the implementation of data-driven classification techniques in memory-constrained digital electronic control units (DECUs)? (3) When does adaptive boosting, an incremental fusion method that successively combines moderately inaccurate classifiers into accurate ones, help improve classification performance? (4) How to synthesize classifier fusion architectures to improve the overall diagnostic accuracy? The classifiers studied in this paper are the support vector machine, probabilistic neural network, k-nearest neighbor, principal component analysis, Gaussian mixture models, and a physics-based single fault isolator. As these algorithms operate on large volumes of data and are generally computationally expensive, we reduce the data set using the multiway partial least squares method. This has the added benefits of improved diagnostic accuracy and smaller memory requirements. The performance of the moderately inaccurate classifiers is improved using adaptive boosting. These results are compared to the results of the classifiers alone, as well as different fusion architectures. We show that fusion reduces the variability in diagnostic accuracy, and is most useful when combining moderately inaccurate classifiers.

An Optimization-Based Method for Dynamic Multiple Fault Diagnosis Problem

Imperfect test outcomes, due to factors such as unreliable sensors, electromagnetic interference, and environmental conditions, manifest themselves as missed detections and false alarms. The main objective of our research on on-board diagnostic inference is to develop near-optimal algorithms for dynamic multiple fault diagnosis (DMFD) problems in the presence of imperfect test outcomes. Our problem is to determine the most likely evolution of fault states, the one that best explains the observed test outcomes. Here, we develop a primal-dual algorithm for solving the DMFD problem by combining Lagrangian relaxation and the Viterbi decoding algorithm in an iterative way. A novel feature of our approach is that the approximate duality gap provides a measure of suboptimality of the DMFD solution.

Data Visualization, Data Reduction and Classifier Fusion for Intelligent Fault Detection and Diagnosis in Gas Turbine Engines

In this paper, we investigate four key issues associated with data-driven approaches for fault classification using the Pratt and Whitney commercial dual-spool turbofan engine data as a test case. The four issues considered here include: (1) Can we characterize, a priori, the difficulty of fault classification via self-organizing maps? (2) Do data reduction techniques improve fault classification performance and enable the implementation of data-driven classification techniques in memory-constrained digital electronic control units (DECUs)?, (3) When does adaptive boosting, an incremental fusion method that successively combines moderately inaccurate classifiers into accurate ones, help improve classification performance?, and (4) How to synthesize classifier fusion architectures to improve the overall diagnostic accuracy? The classifiers studied in this paper are the support vector machine (SVM), probabilistic neural network (PNN), k-nearest neighbor (KNN), principal component analysis (PCA), Gaussian mixture models (GMM), and a physics-based single fault isolator (SFI). As these algorithms operate on large volumes of data and are generally computationally expensive, we reduce the dataset using the multi-way partial least squares (MPLS) method. This has the added benefits of improved diagnostic accuracy and smaller memory requirements. The performance of the moderately inaccurate classifiers is improved using adaptive boosting (AdaBoost). These results are compared to the results of the classifiers alone, as well as different fusion architectures. We show that fusion reduces the variability in diagnostic accuracy, and is most useful when combining moderately inaccurate classifiers.

Dynamic fusion of classifiers for fault diagnosis

This paper considers the problem of temporally fusing classifier outputs to improve the overall diagnostic classification accuracy in safety-critical systems. Here, we discuss dynamic fusion of classifiers which is a special case of the dynamic multiple fault diagnosis (DMFD) problem [1]-[3]. The DMFD problem is formulated as a maximum a posteriori (MAP) configuration problem in tri-partite graphical models, which is NP-hard. A primal-dual optimization framework is applied to solve the MAP problem. Our process for dynamic fusion consists of four key steps: (1) data preprocessing such as noise suppression, data reduction and feature selection using data-driven techniques, (2) error correcting codes to transform the multiclass data into binary classification, (3) fault detection using pattern recognition techniques (support vector machines in this paper), and (4) dynamic fusion of classifiers output labels over time using the DMFD algorithm. An automobile engine data set, simulated under various fault conditions [4], was used to illustrate the fusion process. The results demonstrate that an ensemble of classifiers, when fused over time, reduces the classification error as compared to a single classifier and static fusion of classifiers trained over the entire batch of data. The results for sliding window dynamic fusion are also provided.

Fault detection, diagnosis, and data-driven modeling in HVAC chillers

Heating, Ventilation and Air Conditioning (HVAC) systems constitute the largest portion of energy consumption equipment in residential and commercial facilities. Real-time health monitoring and fault diagnosis is essential for reliable and uninterrupted operation of these systems. Existing fault detection and diagnosis (FDD) schemes for HVAC systems are only suitable for a single operating mode with small numbers of faults, and most of the schemes are systemspecific. A generic real-time FDD scheme, applicable to all possible operating conditions, can significantly reduce HVAC equipment downtime, thus improving the efficiency of building energy management systems. This paper presents a FDD methodology for faults in centrifugal chillers. The FDD scheme compares the diagnostic performance of three data-driven techniques, namely support vector machines (SVM), principal component analysis (PCA), and partial least squares (PLS). In addition, a nominal model of a chiller that can predict system response under new operating conditions is developed using PLS. We used the benchmark data on a 90-ton real centrifugal chiller test equipment, provided by the American Society of Heating, Refrigerating and Air-Conditioning Engineers (ASHRAE), to demonstrate and validate our proposed diagnostic procedure. The database consists of data from sixty four monitored variables under nominal and eight fault conditions of different severities at twenty seven operating modes.

Novel classifier fusion approahces for fault diagnosis in automotive systems

Faulty automotive systems significantly degrade the performance and efficiency of vehicles and are often major contributors of vehicle breakdown; they result in large expenditures for repair and maintenance. Therefore, intelligent vehicle health-monitoring schemes are needed for effective fault diagnosis in automotive systems. Previously, we developed a data-driven approach using a data-reduction technique, coupled with a variety of classifiers, for fault diagnosis in automotive systems. In this paper, we consider the problem of fusing classifier decisions to reduce diagnostic errors. Specifically, we develop three novel classifier fusion approaches: 1. class-specific Bayesian fusion; 2. joint optimization of the fusion center and individual classifiers; and 3. dynamic fusion. We evaluate the efficacies of these fusion approaches on an automotive engine data. The results demonstrate that the proposed fusion techniques outperform traditional fusion approaches. We also show that learning the parameters of individual classifiers as part of the fusion architecture can provide better classification performance.