Yi Lu Murphey

Multiclass pattern classification using neural networks

Multiclass neural learning involves finding appropriate neural network architecture, encoding schemes, learning algorithms, etc. We discuss major approaches used in neural networks for classifying multiple classes. The discussion is focused on these architectures using either a system of multiple neural networks or a single neural network. We discuss various learning algorithms, one-again-all, one-against-one, and p-against-q. We also discuss training procedures associated with each approach, implementation and time complexity. These methods are evaluated through their performances on the NlST handwritten digit database.

Multi-class pattern classification using neural networks

Multi-class pattern classification has many applications including text document classification, speech recognition, object recognition, etc. Multi-class pattern classification using neural networks is not a trivial extension from two-class neural networks. This paper presents a comprehensive and competitive study in multi-class neural learning with focuses on issues including neural network architecture, encoding schemes, training methodology and training time complexity. Our study includes multi-class pattern classification using either a system of multiple neural networks or a single neural network, and modeling pattern classes using one-against-all, one-against-one, one-against-higher-order, and P-against-Q. We also discuss implementations of these approaches and analyze training time complexity associated with each approach. We evaluate six different neural network system architectures for multi-class pattern classification along the dimensions of imbalanced data, large number of pattern classes, large vs. small training data through experiments conducted on well-known benchmark data.

Model-based fault diagnosis in electric drives using machine learning

Electric motor and power electronics-based inverter are the major components in industrial and automotive electric drives. In this paper, we present a model-based fault diagnostics system developed using a machine learning technology for detecting and locating multiple classes of faults in an electric drive. Power electronics inverter can be considered to be the weakest link in such a system from hardware failure point of view; hence, this work is focused on detecting faults and finding which switches in the inverter cause the faults. A simulation model has been developed based on the theoretical foundations of electric drives to simulate the normal condition, all single-switch and post-short-circuit faults. A machine learning algorithm has been developed to automatically select a set of representative operating points in the (torque, speed) domain, which in turn is sent to the simulated electric drive model to generate signals for the training of a diagnostic neural network, fault diagnostic neural network (FDNN). We validated the capability of the FDNN on data generated by an experimental bench setup. Our research demonstrates that with a robust machine learning approach, a diagnostic system can be trained based on a simulated electric drive model, which can lead to a correct classification of faults over a wide operating domain.

An intelligent real-time vision system for surface defect detection

In recent years, there is an increased need for quality control in the manufacturing sectors. In the steel making, the rolling operation is often the last process that significantly affects the bulk microstructure of the steel. The cost of having defects on rolled steel is high because it takes more than 5000 KW-Hr to produce a ton of steel. Early detection of defects can reduce product damage and manufacturing cost. This paper describes a real-time visual inspection system that uses support vector machine to automatically learn complicated defect patterns. Based on the experimental results generated from over one thousand images, the proposed system is found to be effective in detecting steel surface detects. The speed of the system for feature extraction and defect detection is less than 6 msec per one-megabyte image.

Intelligent Hybrid Vehicle Power Control—Part II: Online Intelligent Energy Management

This is the second paper in a series of two that describe our research in intelligent energy management in a hybrid electric vehicle (HEV). In the first paper, we presented the machine-learning framework ML_EMO_HEV, which was developed for learning the knowledge about energy optimization in an HEV. The framework consists of machine-learning algorithms for predicting driving environments and generating the optimal power split of the HEV system for a given driving environment. In this paper, we present the following three online intelligent energy controllers: 1) IEC_HEV_SISE; 2) IEC_HEV_MISE ; and 3) IEC_HEV_MIME. All three online intelligent energy controllers were trained within the machine-learning framework ML_EMO_HEV to generate the best combination of engine power and battery power in real time such that the total fuel consumption over the whole driving cycle is minimized while still meeting the drivers demand and the system constraints, including engine, motor, battery, and generator operation limits. The three online controllers were integrated into the Ford Escape hybrid vehicle model for online performance evaluation. Based on their performances on ten test drive cycles provided by the Powertrain Systems Analysis Toolkit library, we can conclude that the roadway type and traffic congestion level specific machine learning of optimal energy management is effective for in-vehicle energy control. The best controller, IEC_HEV_MISE, trained with the optimal power split generated by the DP optimization algorithm with multiple initial SOC points and single ending point, can provide fuel savings ranging from 5% to 19%. Together, these two papers cover the innovative technologies for modeling power flow, mathematical background of optimization in energy management, and machine-learning algorithms for generating intelligent energy controllers for quasioptimal energy flow in a power-split HEV.

Intelligent Hybrid Vehicle Power Control—Part I: Machine Learning of Optimal Vehicle Power

In this series of two papers, we present our research on intelligent energy management for hybrid electric vehicles (HEVs). These two papers cover the modeling of power flow in HEVs, the mathematical background of optimization in energy management in HEVs, a machine learning framework that combines dynamic programming (DP) with machine learning to learn about roadway-type- and traffic-congestion-level-specific energy optimization, machine learning algorithms, and real-time quasi-optimal control of energy flow in an HEV. This first paper presents our research on machine learning for optimal energy management in HEVs. We will present a machine learning framework ML_EMO_HEV developed for the optimization of energy management in an HEV, machine learning algorithms for predicting driving environments, and the generation of an optimal power split for a given driving environment. Experiments are conducted based on a simulated Ford Escape Hybrid vehicle model provided by Argonne National Laboratorys Powertrain Systems Analysis Toolkit (PSAT). Based on the experimental results on the test data, we can conclude that the neural networks trained under the ML_EMO_HEV framework are effective in predicting roadway type and traffic congestion levels, predicting driving trends, and learning optimal engine speed and optimal battery power from DP.

Automotive signal fault diagnostics - part I: signal fault analysis, signal segmentation, feature extraction and quasi-optimal feature selection

The paper describes our research in vehicle signal fault diagnosis. A modern vehicle has embedded sensors, controllers and computer modules that collect a large number of different signals. These signals, ranging from simple binary modes to extremely complex spark timing signals, interact with each other either directly or indirectly. Modern vehicle fault diagnostics very much depend upon the input from vehicle signal diagnostics. Modeling vehicle engine diagnostics as a signal fault diagnostic problem requires a good understanding of signal behaviors relating to various vehicle faults. Two important tasks in vehicle signal diagnostics are to find what signal features are related to various vehicle faults, and how can these features be effectively extracted from signals. We present our research results in signal faulty behavior analysis, automatic signal segmentation, feature extraction and selection of important features. These research results have been incorporated in a novel vehicle fault diagnostic system, which is described in another paper (see Yi Lu Murphey et al., ibid., p.1076-98).

Intelligent Vehicle Power Control Based on Machine Learning of Optimal Control Parameters and Prediction of Road Type and Traffic Congestion

Previous research has shown that current driving conditions and driving style have a strong influence over a vehicles fuel consumption and emissions. This paper presents a methodology for inferring road type and traffic congestion (RT&TC) levels from available onboard vehicle data and then using this information for improved vehicle power management. A machine-learning algorithm has been developed to learn the critical knowledge about fuel efficiency on 11 facility-specific drive cycles representing different road types and traffic congestion levels, as well as a neural learning algorithm for the training of a neural network to predict the RT&TC level. An online University of Michigan-Dearborn intelligent power controller (UMD_IPC) applies this knowledge to real-time vehicle power control to achieve improved fuel efficiency. UMD_IPC has been fully implemented in a conventional (nonhybrid) vehicle model in the powertrain systems analysis toolkit (PSAT) environment. Simulations conducted on the standard drive cycles provided by the PSAT show that the performance of the UMD_IPC algorithm is very close to the offline controller that is generated using a dynamic programming optimization approach. Furthermore, UMD_IPC gives improved fuel consumption in a conventional vehicle, alternating neither the vehicle structure nor its components.

Driver's style classification using jerk analysis

This paper presents an innovative approach to classifying the drivers driving style by analyzing the jerk profile of the driver. Driving style is a dynamic behavior of a driver on the road. At times a driver can be calm but aggressive at others. The information about drivers dynamic driving style can be used to better control fuel economy. We propose to classify drivers style based on the measure of how fast a driver is accelerating and decelerating. We developed an algorithm that classifies drivers style utilizing the statistical information from the jerk profile and the road way type and traffic congestion level prediction. Our experiment results show that our approach generates more reasonable results than those generated by using other published methods.

Automotive fault diagnosis - part II: a distributed agent diagnostic system

For pt.I see Crossman, J.A. et al., ibid., p.1063-75. We describe a novel diagnostic architecture, distributed diagnostics agent system (DDAS), developed for automotive fault diagnosis. The DDAS consists of a vehicle diagnostic agent and a number of signal diagnostic agents, each of which is responsible for the fault diagnosis of one particular signal using either a single or multiple signals, depending on the complexity of signal faults. Each signal diagnostic agent is developed using a common framework that involves signal segmentation, automatic signal feature extraction and selection, and machine learning. The signal diagnostic agents can concurrently execute their tasks; some agents possess information concerning the cause of faults for other agents, while other agents merely report symptoms. Together, these signal agents present a full picture of the behavior of the vehicle under diagnosis to the vehicle diagnostic agent. DDAS provides three levels of diagnostics decisions: signal-segment fault; signal fault; vehicle fault. DDAS is scalable and versatile and has been implemented for fault detection of electronic control unit (ECU) signals; experiment results are presented and discussed.