Journal of Quality Engineering and Management

Abstract We introduce a method for the statistical design of a depth-based control chart, using the percentile-based approach. The proposed control chart is affine invariant and is asymptotically distribution-free. Generally, the performance of a control chart is evaluated with the average run length metric. The average run length metric has a geometric distribution skewed to the right with a large standard deviation and may not be a proper measure for evaluating the control chart. Therefore, we use the statistical design method of control charts with the PL approach, which is an improvement and development on classical statistical design. By employing constraints on average run length, the length of in-control and out-of-control performances are guaranteed with predetermined probabilities and we can ensure that the in-control run length exceeds the desired value and the out-of-control run length is less than the desired value. Simulation studies show that the proposed control chart is more efficient than the average run length approach.

Abstract  Different types of process control methods exist for jointly monitoring of several quality characteristic. Recent studies have shown that multi-stage sampling control charts perform better in detecting shifts in process. Therefore, because of the efficiency of the joint mean and standard deviation charts with multi-stage sampling in detecting shifts in mean and standard deviation of the process, discussion and consideration of these charts is necessary. In this paper, the joint Double Sampling and Triple Sampling mean and standard deviation charts are introduced. The joint Statistical Design of these charts is formulated as an optimization problem and solved using a genetic algorithm. 
 
 

Abstract Identifying significant changes in key indicators of financial processes is one of the points of interest in recent years and after the financial crisis. Control charts are one of the most powerful tools used in this field. A noteworthy point in the practical use of control charts is the estimation of process parameters based on available data. Despite extensive research into the effect of parameter estimation on the performance of control charts, little research has been done on processes with time-dependent observations. Due to the importance and widespread use of the Auto Regressive Conditional Heteroskedasticity (ARCH) time series model in monitoring financial processes, in this paper, the effect of prototype size for estimating model parameters on the performance of phase 2 control charts is investigated. The performance of each control diagram is evaluated using simulation studies based on AARL and SDARL criteria and the results are described.

Abstract Traditional control diagrams are based on the basic assumption that process data are sequentially independent of each other and have a normal distribution. However, in many cases in the real world, including in chemical and continuous processes, this assumption does not exist and there is a kind of autocorrelation between the data collected from the process. The use of traditional control charts in self-correlated processes is unreliable and increases erroneous warnings. One of the methods developed to control autocorrelation processes is to identify the structure of process time series and use residual values ​​to control process. In this paper, a model based on adaptive neural-fuzzy systems to identify time series structure and predict design use. May be. Is in AR (2). Also, the residual control diagrams based on this system for the self-return data of degree 2 and for (EWMA) finally using the simulated data, the efficiency of the proposed method in the moving average diagram are evaluated. Different degrees of correlation are evaluated.

Abstract Traditional control charts are based on the basic assumption that process data are sequentially independent of each other and have a normal distribution. However, in many real-world cases, including chemical and continuous processes, this basic assumption does not exist and there is a kind of autocorrelation between the data collected from the process. The use of traditional control charts in autocorrelated processes is unreliable and increases false alarms. One of the methods developed to control autocorrelated processes is to identify the structure of the process time series and use the residual values ​​to control the process. In this paper, a model based on neuro-fuzzy adaptive systems is designed to identify the structure of the time series and use the prediction. Finally, it is AR(2). Also, residual control charts based on this system for second-order autoregressive data using simulated data, the efficiency of the proposed method in the weighted moving average chart and for different degrees of correlation are evaluated, and it is shown that the proposed method has very good efficiency for data with high correlation.

Abstract As is well known, data obtained from real-world processes are typically autocorrelated, and monitoring such processes requires accounting for this autocorrelation. The control charts developed so far for monitoring the defective proportion (p) are generally based on the assumption that binary observations are independent, which ignores the inherent correlation in the data.
In this paper, a cumulative sum (CUSUM) control chart is proposed that incorporates the autocorrelation between binary observations using a first-order two-state Markov chain model. Furthermore, using the Average Number of Observations to Signal (ANOS) index, it is shown that under 100% inspection conditions, the proposed chart performs better than the Bernoulli CUSUM chart—which assumes independent observations—and, in other words, detects increases in p more quickly.

Abstract Control charts are among the most important tools in statistical process control, used to monitor the extent of variation in a process. When a assignable cause deviation is observed in a control chart, identifying the root causes of the change and determining the time at which the deviation began—referred to as the change point—is crucial and impactful. In some statistical process control problems, the quality of a product or the performance of a process is described by the relationship between a response variable and one or more independent variables, known as a profile. In many applications, such as calibration, this relationship is described by a linear profile, while in other situations more complex models, such as Poisson regression profiles, are required. In this paper, the maximum likelihood estimation method is employed to detect change points in Phase II monitoring of Poisson regression profiles, and its performance is evaluated through simulation.