Statistical Process
Nastaran Hajarian; Farzad Movahedi Sobhani; Seyed Jafar Sadjadi
Abstract
One of the most complex and costly systems in the industry is the Gas turbine (GT). Because of the complexity of these assets, various indicators have been used to monitor the health condition of different parts of the gas turbine. Turbine exit temperature (TET) spread is one of the significant indicators ...
Read More
One of the most complex and costly systems in the industry is the Gas turbine (GT). Because of the complexity of these assets, various indicators have been used to monitor the health condition of different parts of the gas turbine. Turbine exit temperature (TET) spread is one of the significant indicators that help monitor and detect faults such as overall engine deterioration and burner fault. The goal of this article is to use data-driven approaches to monitor TET data to detect faults early, as fault detection can have a significant impact on gas turbine reliability and availability. In this study, the TET data of v94.2 GT is measured by six temperature transmitters to show a detailed profile. According to the statistical tests, TET data are high dimensional and time-dependent in the real world industry. Hence, three distinctive methods in the field of the gas turbine are proposed in this study for early fault detection. Conventional principal component analysis (PCA), moving window PCA (MWPCA), and incremental PCA (IPCA) were implemented on TET data. According to the results, the conventional PCA model is a non-adaptive method, and the false alarm rate is high due to the incompatibility of this approach and the process. The MWPCA based on V-step-ahead and IPCA approaches overcame the non-stationary problem and reduced the false alarm rate. In fact, these approaches can distinguish between the normal time-varying and slow ramp fault processes. The results showed that IPCA could detect fault situations faster than MWPCA based on V-step-ahead in this study.
Statistical Process
Teuku Mirwan Saputra; Hernadewita Hernadewita; Akmal Yudha Prawira Saputra; Lien Herliani Kusumah; Hermiyetti ST
Abstract
Enhancing the process capability is a must in quality improvement of process manufacturing in the industry. Usually, Statistical Process Control (SPC) is applied in measure the activity of quality improvement. In this case, the statistical process control used to measure the process capability of molding ...
Read More
Enhancing the process capability is a must in quality improvement of process manufacturing in the industry. Usually, Statistical Process Control (SPC) is applied in measure the activity of quality improvement. In this case, the statistical process control used to measure the process capability of molding machine regarding of force result of inner contact rubber button in plastic industry. The forces of inner contact rubber button which produce by molding machine already become the critical point as a judgment of malfunctions or not. Meanwhile, the forces problem still found in molding machine that affected to the inner contact rubber button as defect product. Furthermore, as shown of SPC, the process capability of molding machine was 0.63. This mean the process capability is still need the improvement. As the result of quality improvement was made by applied of Poka-Yoke, the process capability of molding machine was improved to 1.65.
Statistical Process
Elham Basiri
Abstract
Recently, Rayleigh distribution has received considerable attention in the statistical literature. This paper describes the Bayesian prediction of the one parameter Rayleigh distribution when the data are Type II censored data. Suppose we are planning to collect a Type II censored sample ...
Read More
Recently, Rayleigh distribution has received considerable attention in the statistical literature. This paper describes the Bayesian prediction of the one parameter Rayleigh distribution when the data are Type II censored data. Suppose we are planning to collect a Type II censored sample from the one parameter Rayleigh distribution in order to find a point predictor for a future order statistic with smallest mean squared prediction error (MSPE) among other point predictors. Although, considering large values for failure numbers yields a point predictor with smaller MSPE, the average cost may increase considerably. One question arises here that “How many failure is enough?”. The aim of this paper is finding the optimal value for number of failures in Type II censoring by considering two criteria, total cost of experiment and mean squared prediction error. Towards this end, we find the Bayesian point predictor for the parameter of distribution. Then, the optimal value for number of failures is obtained when the mean squared prediction error and the total cost of experiment are bounded. To show the usefulness of the obtained results, a simulation study is presented.