▼ An Energy Management System (EMS) monitors, evaluates, and controls the performance of different energy-consuming equipment such as motors and compressors and extending to plant-floor machinery. This research explores and develops a systematic framework and statistically-significant analytic models for using electric consumption power variables as an indicator for machine-level health or performance. This is in an effort to explore new techniques for improving the current capabilities of traditional energy management systems. Power data is collected real-time for electrical power consumption usage of machines, under consistent operational conditions. Three levels of performance assessment and associated models are developed based on acquired power signals that effectively consider the power consumed by a machine as an indicator for overall machine performance. The research hypothesis is that a relationship exists between a machine’s electric energy consumption levels and the machine’s level of performance and potential health degradation. An intuitive predictive model is developed to give a power-based performance prediction for one machining cycle or cycle step ahead. The models are successfully implemented and validated on a real-world industrial case study for an injection molding process where electrical power consumption data is collected. A standard moving average method is used to benchmark the results of this analysis. Advisors/Committee Members: Lee, Jay.

▼ Classification is one of the important tasks of data mining. Class imbalance – or differences in class distribution – has been reported to hinder the performance of standard classification models. This dissertation first presents a systematic study to evaluate the impact of class imbalance on several critical steps of learning, namely feature selection, model fitting and performance evaluation. However, study also shows that class imbalance may not be the only cause to blame for the loss of performance, and the underlying complexity of the problem may play a more fundamental role. In this dissertation, K-S tree, a decision tree method based on Kolmogorov-Smirnov statistic, is proposed to segment the data so that the complex problem can be dissected into easier sub-problems and for each sub-problem class imbalance becomes less challenging. K-S tree is also used to perform feature selection, which not only selects relevant variables but also removes redundant ones. After segmentation, a two-way re-sampling will be performed at segment level and the rebalanced data will be used to fit logistic regression models also at segment level. The effectiveness of the proposed method is demonstrated through three case studies – automatic detection of microcalcification in Mammogram, San Diego housing refinance prediction and credit risk assessment. Advisors/Committee Members: Huang, Hongdao.

▼ Fault detection, which involves the estimation of the condition, health or degradation of an equipment or a process and a decision logic to determine whether an event that can be considered as a fault has occurred, is an integral component in prognostics and health management because it is an essential indicator when to perform fault diagnosis and isolation, and it also precedes any performance prediction methodology. The implementation of data-driven fault detection has generally been reliant on unit-specific models which can be less effective with insufficient training data or when used in applications with non-stationary working conditions. The aforementioned scenarios can be alleviated by leveraging on data from similar units experiencing comparable operating regimes. This dissertation investigates the formulation, development and implementation of a cluster-based fault detection to a fleet of similar machines. A two-step approach is introduced: fleet clustering and local cluster fault detection. Fleet clustering verifies, discovers and identifies the group structure of the network of machines. Afterwhich, the health of each unit in the cluster is assessed using peer-to-peer comparison. The approach developed in this dissertation is validated with two case studies: a fleet of industrial welding robots from an automotive manufacturing facility and a group of wind turbines from several wind farms. Advisors/Committee Members: Lee, Jay.

▼ The research presented in this dissertation focuses on a two-stage methodology of virtual machining of parts produced on a three-axis turning center and virtual inspection of the produced parts using a bridge-type CMM. The virtual machining system focuses on a priori predicting the surface profile of the turned part. The surface profile is generated by modeling the effects of the static errors inherent in the turning center, the error in the spindle motion, machine vibrations, tool geometry, process parameters, and tool wear. The model so developed is used to calculate various geometric dimensioning and tolerancing (GD and T) parameters of interest (form error, size, runout, and orientation tolerances). The effect of the various error factors on the GD and T parameters is examined. It is observed that the static errors, spindle motion errors, and tool wear play a significant role on the final profile of the part. The results from the inspection are used to develop quantitative relations between the values of the GD and T parameters and the machining process parameters. These regression equations are used to develop a machining advisor that optimizes the process parameters so as to maximize the adherence of the part to the design specifications. In order to be used in a mass production scenario, a weighted optimization function is used where the machining time is optimized simultaneously with the GD and T parameter. The analysis enables the identification of “sweet spots” on the machine, which, through a particular choice of process parameters and other variables, could yield more accurate products. The optimized process parameters are tested using the virtual machining system, and the results indicate a close match with the estimated values from the regression equations. The virtual inspection system focuses on using the virtual profile generated to analyze the effectiveness of various inspection strategies. The inspection strategy includes the number of sample points, sampling method, and the location of the part on the CMM table for inspection. In addition to these factors, the uncertainty of the CMM due to effects such as probe pre-travel and hysteresis is also considered. The results obtained from this stage are used to study the effect of the various inspection parameters on the accuracy of the inspection results. It is observed that the sample size, location on the CMM table, and the level of CMM uncertainty significantly influence the accuracy of the inspection process. Based on these results, quantitative relations are established between the deviation of the inspection results from the true value of the GD and T parameter and the inspection parameters. These relationships, along with relationships for inspection time, are used to develop an inspection advisor to optimize the inspection parameters with the objective of increasing the inspection accuracy and precision, reducing the inspection time, or a combination of both objectives. The optimized inspection parameters obtained from the inspection advisor are used to inspect a test part for validation. The results obtained indicate a close match with the expected accuracy and precision levels. Finally, a discussion of the procedure to implement these methodologies on the shop floor is presented. The discussion focuses on various scenarios ranging from a single turning center and single CMM to multiple turning centers and CMMs. Overall, the results from the research presented in this dissertation are expected to enable a substantial decrease in the need for physical prototyping for deciding optimal turning and inspection parameters, thereby reducing developmental costs and increasing the profitability of the manufacturing industry. Advisors/Committee Members: Anand, Sundararaman.

▼ This research was initiated as a result of a project funded by the United States Environmental Protection Agency with the goals to: (a) examine the pollution prevention strategies to increase lead recycling and recovery and (b) identify the U.S. facilities receptive to implement such strategies. Since the majority of recycled lead is extracted from lead acid batteries (LAB), this work is centered on identifying the major sources of lead pollution in secondary lead smelters from battery recycling and identifying and evaluating readily deployable pollution prevention technologies for the secondary lead industry. Currently, there is limited information on recycling rates for LAB in the published literature. The Battery Council International determined that the recycling rates for recovering lead from spent LAB is approaching 99.2% for the 1999 to 2003 time period. Although very encouraging, such rates appear to be significantly higher than those reported for other countries. Therefore, its recycling efforts in the US has been unclear so as to determine the maximum opportunities for metal recovery and recycling in the face of significant demands for LAB particularly in the auto industry. In the first part of the research an evidence-based approach is utilized to: (1) determine recycling rates for lead recovery in the LAB product lifecycle for the U.S. market; and (2) quantify and identify opportunities where lead recovery and recycling can be improved. It was observed that 1) lead recovery and recycling has been stable between 1999-2006; (2) lead consumption has increased at an annual rate of 2.25%, thus, the values derived in this study for opportunities dealing with lead recovery and recycling underestimate the amount of lead in scrap and waste generated; and (3) the opportunities for maximizing lead recovery and recycling are centered on spent batteries left with consumers, mishandled LAB sent to auto wreckers, slag resulting from recycling technology process inefficiencies, and lead lost in municipal waste. The second part of this research examines pollution prevention and waste minimization practices and technologies which meet the following criteria: (a) reduce/recover/recycle the largest quantities of lead currently being disposed of as waste, (b) technically and economically viable, that is, ready to be diffused and easily transferable, and have (c) strong industry interest. The documented methodology reveals that it is technically and economically feasible to implement integrated environmental solutions to increase lead recovery and recycling among US smelters. The third part of this research quantifies the recycling rates for lead recovery in the LAB product lifecycle for the U.S. market throughout the history of recycling efforts (i.e., based on all historical data available); and, identifies opportunities for significant lead recovery and recycling. The finding of this research suggests that future efforts be directed into ways to establish sustainable development for the secondary smelting industry. This is important to achieve for the following reasons: (a) to maintain sustained growth in turbulent times such as those experienced in today’s global financial industries and felt throughout other industries through, among other things, the tightening of supply of credit; and (b) to significantly reduce dependence upon primary lead smelting and imports to compensate for the declining efforts in lead recovery and recycling. Advisors/Committee Members: Lim, Teik.

▼ Offshore manufacturing creates national defense concerns for the United States of America. National defense depends on four manufacturing elements; people, innovation, production, and logistics. Before discussing those elements, with respect to defense, it is essential to understand them and their relationships with each other. Government is another topic to consider when examining the effects of offshore manufacturing on national defense. Trends in the elements and government, ashore and abroad, threaten the economic and military dominance enjoyed by the United States for decades. The objective of this research is to show that offshore manufacturing trends are detrimental to the nation defense of the United States of America. Advisors/Committee Members: Hall, Ernest.

▼ The degradation process of a complex system may be affected by many unknown factors, such as unidentified fault modes, unmeasured operational conditions, engineering variance, environmental conditions, etc. These unknown factors not only complicate the degradation behaviors of the system, but also lower the quality of the collected data for modeling. Due to lack of knowledge and incomplete measurements, certain important context information (e.g. fault modes, operational conditions) of the collected data will be missing. Therefore historical data of the system with a large variety of degradation patterns will be mixed together. With such data, learning a global model for Remaining Useful Life (RUL) prediction becomes extremely hard. This has led us to look for advanced RUL prediction techniques beyond the traditional global models. In this thesis, a novel RUL prediction method inspired by the Instance Based Learning methodology, called Trajectory Similarity Based Prediction (TSBP), is proposed. In TSBP, the historical instances of a system with life-time condition data and known failure time are used to create a library of degradation models. For a test instance of the same system whose RUL is to be estimated, similarity between it and each of the degradation models is evaluated by computing the minimal weighted Euclidean distance defined on two degradation trajectories. Based on the known failure time, each of the degradation models will produce one RUL estimate for the test instance. The final RUL estimate can then be obtained by aggregating the multiple RUL estimates using a density estimation method. A case study using the turbofan engine degradation simulation data supplied by NASA Ames is provided to study the performance of TSBP. In this study, the TSBP method has demonstrated significant improvement in performance over a Neural Network based prediction method. Advisors/Committee Members: Lee, Jay.