Walter Priesnitz Filho

Statistical quality control: A case study research

In this paper, statistical quality control of a production line has been presented using the classical method Shewhart, cumulative sum method (CUSUM) and Exponentially Weighted Moving Average (EWMA). The Shewhart technique can be utilized in controlling the process in which there are big changes in mean. The cumulative sum method is more efficient in detecting small changes in the mean. The EWMA includes the two techniques, as limited cases, since when lambda = 1 the exponential smooth model is equivalent to the Shewhart technique, and when lambda = 0 equivalent to the cumulative sum technique.

An algorithm for quality control charts for autocorrelated data

Currently considerable attention has been given to the effect of data correlation on statistical process control (SPC). Use of traditional SPC methods when observations are correlated often leads to misleading conclusions as to whether or not the process is under control. The objetive of this paper is to develop an algorithm to adjust a model ARMA(p,q), for calculate the run length distribution (RLD), the average run length (ARL), and the standard deviation of the run length (SRL), for residual control charts X(ind) and MR used to monitor autocorrelated processes. The algorithm was used for analysis of real data. We conclude that for negative first-order autocorrelation, the residuals chart is performing better than the Shewhart chart for independent observations.

Kalman Filter an application in process control

In this paper, we present in a systematic way an approach to modelling of Kalman Filter in the Statistical Process Control, formulated in state space. The procedure based on the Kalman filter was superior to the classical procedures like Shewhart and CUSUM control charts.

Study of the layout of a steel manufacturing industry

A layout survey is required to improve the production flow of a steel company. A reliable layout provides improvements in productivity and savings because it states a better distribution for machinery, appliances and human resources. In this paper the layout of a steel company is analyzed. It is found that the production flow does not match with the physical positioning of the machinery that designs the production line of the company. Following this analysis, a reformulation of the layout is proposed to aim the reduction of processing appliance transportation time, leading an improvement in the capacity of production and a reduction in setups times. Pursuing the implementation of the new layout, a search was developed to compare the outputs before and after the redefinition of the layout. The result highlights improvements as shortening processing appliance transportation time, broadening production programming performance and lessening setup times in the lines.

Control charts for monitoring autocorrelated processes based on Neural Networks model

Statistical process control can have different objectives and can be done in different forms (Hawkins, et al, 2003). Currently, considerable attention has been given to the effect of data correlation on the statistical process control (SPC). The use of traditional SPC methods when observations are correlated often leads to misleading conclusions as to whether or not the process is under control. This paper presents the construction of residual based control charts, obtained from Neural Network model, to monitor the mean and dispersion in autocorrelated productive processes. One application with real data and a performance comparison of the residual control charts obtained from the Artificial Neural Network model with that of traditional control charts X(bar) and R presented. It is established that the former procedure is more efficient in detecting changes in the mean and dispersion of the process than the latter.

Statistical process control by model Bayesian

Currently considerable attention has been given to the effect of data correlation on statistical process control (SPC). Use of traditional SPC methods when observations are correlated often leads to misleading conclusions as to whether or not the process is under control. The objective of this paper is to develop an algorithm to adjust a Dynamic Linear Model, to calculate the run length distribution (RLD), the average run length (ARL), standard deviation of the run length (SRL), for residual control charts X macr and R. The algorithm is applied to data collected from a textile company. The results showed that the process had been out of control needing systematic monitoring, with the objective of improving the quality of the products.