Sung-Shick Kim

One-class support vector machines: an application in machine fault detection and classification

Fast incipient machine fault diagnosis is becoming one of the key requirements for economical and optimal process operation management. Artificial neural networks have been used to detect machine faults for a number of years and shown to be highly successful in this application area. This paper presents a novel test technique for machine fault detection and classification in electro-mechanical machinery from vibration measurements using one-class support vector machines (SVMs). In order to evaluate one-class SVMs, this paper examines the performance of the proposed method by comparing it with that of multilayer perception, one of the artificial neural network techniques, based on real benchmarking data.

Neural networks fusion to overlay control system for lithography process

This paper presents a neural network based overlay control system for lithography process. The control system is structured to be compatible with the existing control system. The two main components of the control system are neural network prediction module for metrology prediction and a control module for various control methods. The prediction module utilizes various overlay metrologies and other process related parameters to assess the process conditions and make accurate predictions of the output metrologies. Based on the prediction results, control module calculates the appropriate control parameter settings. The prediction module is constructed using the Levenberg-Marquardt method to compensate for the small to medium size neural network and the demand for speed. The control module incorporates both popular control methods and specific engineering process control (EPC). Evaluation results are presented to illustrate the control system performance.

Update Cycle Detection Method of Control Limits using Control Chart Performance Evaluation Model

Statistical process control (SPC) is an important technique for monitoring and managing the manufacturing process. In spite of its easiness and effectiveness, some problematic sides of application exist such that the SPC techniques are hardly reflect the changes of the process conditions. Especially, update of control limits at the right time plays an important role in acquiring a reasonable performance of control charts. Therefore, we propose the control chart performance evaluation index (CPEI) based on count data model to monitor and manage the performance of control charts. The CPEI could indicate the degree of control chart performance and be helpful to detect the proper update cycle of control limits in real time. Experiments using real manufacturing data show that the proper update intervals are made by proposed method.

Contract-collaboration network method for modeling manufacturing resource control workflows

This paper presents contract-collaboration network (CC-Net) method that is developed to model manufacturing resource control workflows. The CC-Net is an object-oriented class diagram. It depicts the contract-collaboration relationships among the classes in a manufacturing system, with constraints. The CC-Net method uses a primitive modeling block called collaboration module by which the CC-Net is established systematically. This idea is very similar to that of the Lego® block toy. Unlike most workflow modeling methods, the CC-Net method views workflow modeling as a constraint satisfaction process. That is, describing the set of constraint recovery rules corresponding to the constraint violations is regarded as workflow modeling. The obtained set of workflow rules is free from process deadlock and considers all the events of triggering the workflow. We explore the use of the CC-Net method for the workflow modeling of a flexible manufacturing system.

Model Parameter Based Fault Detection for Time-series Data

The statistical process control (SPC) assumes that observations follow the particular statistical distribution and they are independent to each other. However, the time-series data do not always follow the particular distribution, and most of cases are autocorrelated, therefore, it has limit to adopt the general SPC in tim series process. In this study, we propose a MPBC (Model Parameter Based Control-chart) method for fault detection in time-series processes. The MPBC builds up the process as a time-series model, and it can determine the faults by detecting changes parameters in the model. The process we analyze in the study assumes that the data follow the ARMA (p,q) model. The MPBC estimates model parameters using RLS (Recursive Least Square), and -control chart is used for detecting out-of control process. The results of simulations support the idea that our proposed method performs better in time-series process.

A Study of Estimation Method for Auto-Regressive Model with Non-Normal Error and Its Prediction Accuracy

Bo Mi Lim ․Cheong-Sool Park ․Jun Seok Kim․Sung-Shick Kim ․Jun-Geol BaekSchool of Industrial Management Engineering, Korea UniversityWe propose a method for estimating coefficients of AR (autoregressive) model which named MLPAR (Maximum Likelihood of Pearson system for Auto-Regressive model). In the present method for estimating coefficients of AR model, there is an assumption that residual or error term of the model follows the normal distribution. In common cases, we can observe that the error of AR model does not follow the normal distribution. So the normal assumption will cause decreasing prediction accuracy of AR model. In the paper, we propose the MLPAR which does not assume the normal distribution of error term. The MLPAR estimates coefficients of auto-regressive model and distribution moments of residual by using pearson distribution system and maximum likelihood estimation. Comparing proposed method to auto-regressive model, results are shown to verify improved perfor-mance of the MLPAR in terms of prediction accuracy.

Fault Detection of Unbalanced Cycle Signal Data Using SOM-based Feature Signal Extraction Method

In this paper, a feature signal extraction method is proposed in order to enhance the low performance of fault detection caused by unbalanced data which denotes the situations when severe disparity exists between the numbers of class instances. Most of the cyclic signals gathered during the process are recognized as normal, while only a few signals are regarded as fault; the majorities of cyclic signals data are unbalanced data. SOM(Self-Organizing Map)-based feature signal extraction method is considered to fix the adverse effects caused by unbalanced data. The weight neurons, mapped to the every node of SOM grid, are extracted as the feature signals of both class data which are used as a reference data set for fault detection. kNN(k-Nearest Neighbor) and SVM(Support Vector Machine) are considered to make fault detection models with comparisons to Hotelling`s Control Chart, the most widely used method for fault detection. Experiments are conducted by using simulated process signals which resembles the frequent cyclic signals in semiconductor manufacturing.

Dispatching to Minimize Flow Time for Production Efficiency in Non-Identical Parallel Machines Environment with Rework

Reducing waste for the efficiency of production is becoming more important because of the rapidly changing market circumstances and the rising material and oil prices. The dispatching also has to consider the characteristic of production circumstance for the efficiency. The production circumstance has the non-identical parallel machines with rework rate since machines have different capabilities and deterioration levels in the real manufacturing field. This paper proposes a dispatching method, FTLR (Flow Time Loss Index with Rework Rate) for production efficiency. The goal of FTLR is to minimize flow time based on such production environments. FTLR predicts the flow time with rework rate. After assessing dominant position of expected flow time per each machine, FTLR performs dispatching to minimize flow time. Experiments compare various dispatch methods for evaluating FTLR with mean flow time, mean tardiness and max tardiness in queue.

Improved Dispatching Algorithm for Satisfying both Quality and Due Date

The manufacturing industry seeks for improvements in efficiency at the manufacturing process. This paper presents a method for effective real time dispatching for parallel machines with multi product that minimizes mean tardiness and maximizes the quality of the product. What is shown in this paper is that using the Rolling Horizon Tabu search method in the real time dispatching process, mean tardiness can be reduced to the minimum. The effectiveness of the method presented in this paper has been examined in the simulation and compared with other dispatching methods. In fact, using this method manufacturing companies can increase profits and improve customer satisfaction as well.