A Fuzzy Expert System for Controlling Individual Measurements

Shewhart charts are the main tools for statistical process control. They are used for detecting assignable causes which affect quality of process output. From them, X and MR charts are two univariate control charts for monitoring mean and variation of measurable quality characteristics. The main drawbacks of these charts are: weakness of X chart against non-normal distribution of process data, autocorrelation of consecutive MR values, and inability of MR chart in detecting decrease in process variation. These drawbacks have declined their effectiveness. This paper presents a simple method for monitoring two parameters of normal distribution, based on fuzzy logic, resolving above drawbacks. The proposed method can detect non-random patterns of the process data. Here a fuzzy expert system is developed to assess individual measurements. This results in considerable improvement in decreasing out-of-control average run length versus the traditional control charts. Moreover, performance of fuzzy control chart for recognizing unnatural process behaviors has investigated using Mont-Carlo simulation and advantages of this novel method have been investigated.