Susanta Kumar Gauri

Recognition of control chart patterns using improved selection of features

Recognition of various control chart patterns (CCPs) can significantly reduce the diagnostic search process. Feature-based approaches can facilitate efficient pattern recognition. The full potentiality of feature-based approaches can be achieved by using the optimal set of features. In this paper, a set of seven most useful features is selected using a classification and regression tree (CART)-based systematic approach for feature selection. Based on these features, eight most commonly observed CCPs are recognized using heuristic and artificial neural network (ANN) techniques. Extensive performance evaluation of the two types of recognizers reveals that both these recognizers result in higher recognition accuracy than the earlier reported feature-based recognizers. In this work, various features are extracted from the control chart plot of actual process data in such a way that their values become independent of the process mean and standard deviation. Thus, the developed feature-based CCP recognizers can be applicable to any general process.

Optimization of Correlated Responses of EDM Process

Electrical discharge machining (EDM) process has several important performance measures (responses), some of which are correlated. For example, material removal rate (MRR) and electrode wear rate (EWR) are highly correlated. No reported research work on EDM process has taken into consideration the possible correlation between the response variables while determining the optimal process conditions. Thus, the results achieved by the past researchers are often suboptimal. In the recent past, a few multiresponse optimization methods have been proposed that make use of the principal component analysis (PCA) to take into account the possible correlation between the responses. So, ideally, these methods should be more effective for optimizing the EDM process. However, the relative optimization performances of these methods are unknown and therefore, the process engineers may face the difficulty in selecting the most appropriate method for optimizing an EDM process. In this article, two sets of past experimental data on EDM processes are analyzed using four PCA-based optimization methods. The optimization performances of these methods are compared with the results achieved by the past researchers, considering expected total signal-to-noise (S/N) ratio as the utility measure. It is found that the PCA-based approaches, in general, lead to better optimization performance and among the four PCA-based approaches, PCA-based proportion of quality loss reduction (PQLR) method results in the best optimization performance. So the PCA-based PQLR method can be applied for optimizing multiple responses of EDM process.

Multi-Response Optimization Using Multiple Regression–Based Weighted Signal-to-Noise Ratio (MRWSN)

ABSTRACT Several methods for optimization of multiple responses have been developed. Most of them have mathematical complexities and are not easily implemented by engineers. In this article, we propose a new method that integrates multiple regression technique and Taguchis signal-to-noise (SN) ratio concept and is also easily implemented. Two sets of experimental data are analyzed using the proposed method. The proposed method is found to be superior to other methods with respect to total SN ratio as well as closeness of individual responses to their respective target values, which are reflected in the expected mean square error (MSE) values for the individual responses.

Assessing effectiveness of the various performance metrics for multi-response optimization using multiple regression

Several methods for optimization of multiple response problems using planned experimental data have been proposed in the literature. Among them, an integrated approach of multiple regression-based optimization using an overall performance criteria has become quite popular. In this article, we examine the effectiveness of five performance metrics that are used for optimization of multiple response problems. The usefulness of these performance metrics are compared with respect to a utility measure, namely, the expected total non-conformance (NC), for three experimental datasets taken from the literature. It is observed that multiple regression-based weighted signal-to-noise ratio as a performance metric is the most effective in finding an optimal solution for multiple response problems.

Optimisation of the correlated responses of EDM process using modified principal component analysis-based utility theory

One of the most extensively used non-traditional machining processes is electrical discharge machining (EDM). Achieving the best quality performance from the EDM process requires simultaneous optimisation of multiple performance characteristics (responses). Researchers have applied different techniques of multi-response optimisation for optimising the EDM processes. But those approaches do not take care of the possible correlation between the responses, whereas some of the responses of EDM process are usually correlated. In recent times, different principal component analysis (PCA)-based approaches for multi-response optimisation have been proposed, which take into account the correlations among the responses. One of such approaches is the PCA-based utility theory (UT) approach. This paper highlights one weakness of the PCA-based UT approach and proposes the necessary modifications. Two sets of past experimental data on EDM processes are analysed using the modified procedure. The results show that the modified PCA-based UT approach leads to better optimisation performance than that obtained by the earlier researchers. This implies that the modified PCA-based UT approach can be very useful technique for optimising the EDM processes.

A Quantitative Approach for Detection of Unstable Processes Using a Run Chart

Abstract Nearly all process data have a time variable, representing the time each data point is measured. A process can be said stable if the parameter(s) of the distribution of a process/product characteristic remain constant over time and there is no autocorrelation. Only a stable process has the ability to perform in a predictable manner over time. Statistical analysis of process data usually assume that these data are obtained from stable processes. In the absence of control charts, the hypothesis of process stability is usually assessed by visually examining the pattern in the run chart. In this paper, a measure for process stability called process stability indicator (PSI) is proposed based on two shape features of run chart pattern, using which an unstable process can be detected objectively from the run chart of considerably shorter length. Important properties of PSI are derived and cutoff values of PSI for run charts of different lengths are determined. The effectiveness of the proposed approach is evaluated and compared with the existing quantitative approaches using simulated data. The results show that the proposed PSI method, in general, results in much better performance than the other approaches. But it is relatively less effective in detection of the unstable process conditions that leads to cyclic pattern.

A test for suitability of the preliminary samples for constructing control limits of X¯ chart

Unless the preliminary m subgroups of small samples are drawn from a stable process, the estimated control limits of X¯ chart in phase I can be erroneous, due to which the performance of the chart in phase II can be significantly affected. In this work, a quantitative approach based on extraction of the shape features of control chart patterns in the X¯ chart is proposed for evaluating the stability of the process mean, while the preliminary samples were drawn and thus, the subjectivity associated with the visual analysis of the patterns is eliminated. The effectiveness of the test procedure is evaluated using simulated data. The results show that the proposed approach can be very effective for m≥48. The power of the test can be improved by identifying a new feature that can more efficiently discriminate the cyclic pattern of smaller periodicity from the natural pattern and by redefining the test statistic.

A study on the multi-response optimisation of EDM processes

Determination of the parametric settings that can simultaneously optimise multiple responses of electric discharge machining (EDM) process is an important issue to the engineers. Researchers have usually preferred to apply grey relational analysis (GRA)-based approaches for optimising the multiple responses of EDM process. The advantage of GRA-based approaches is that they are easily comprehendible and computationally simple. Literature survey reveals that there are few other simple methods for multi-response optimisation which can be easily implemented using EXCEL worksheet. The aim of this paper is to explore whether any of these methods can give better optimisation performance than the commonly used GRA-based approaches. In this paper, two sets of past experimental data on EDM processes are analysed using four different methods and their relative performances are then compared. The results show that weighted signal-to-noise ratio (WSN) and utility theory methods give better overall optimisation performance than GRA-based and other approaches.

Multiresponse optimisation of correlated responses: some analysis and results

Although there are several methods to resolve multiresponse optimisation problems, only a few of them take care of the possible correlation among the responses. The relative performance of these methods is unknown and therefore, selection of the appropriate method becomes an important issue to the engineers. In this paper, the optimisation performance of three methods dealing with the multiple correlated responses, e.g., weighted principal component (WPC) method, principal component analysis (PCA)-based grey relational analysis (GRA) method and PCA-based technique for order preference by similarity to ideal solution (TOPSIS) method are compared. It is found that PCA-based GRA method result in the best optimisation when the correlations among the responses are quite weak. But the WPC and PCA-based TOPSIS methods result in equivalent and better optimisation when at least two responses are strongly correlated. However, WPC method is preferable because of its simpler computational procedure.

Modeling product-mix planning for batches of melt under multiple objectives in a small scale iron foundry

Product-mix of castings of different types/sizes can be produced in a heat (batch of melt) if all these castings require similar raw material composition. However, inappropriate product-mix may lead to under utilization of furnace capacity or failure in timely delivery or overproduction of castings or may call for deployment of excess resources for packing of molds to enable starting of the pouring operation in time. Some objectives, again, can be conflicting in nature. This paper presents a weighted integer goal programming model for the product-mix planning, developed in the context of a small scale iron foundry. Implementation of the model is illustrated using real life data from an Indian foundry.