Lochner, Robert H. (1987, ASQC) W.A. Golomski & Associates, Chicago, IL
Process and production equipment reliability are critical aspects of quality manufacturing systems. The cause of a production process going out of control can often be traced to failure of manufacturing equipment - either through sudden change or gradual drift. A set of properly done control charts contains the reliability history of the equipment and the process. Unfortunately, that information is seldom utilized. Decisions regarding need for process adjustment are based on control charts which fail to take into account the elapsed time since last adjustment or repair. With some manufacturing processes, especially those where drift or wear is a factor, elapsed time can be a critical factor in equipment performance.In this paper we develop a method for using control chart information to measure the effect elapsed time (since last repair or adjustment) has on production performance. This leads to a blending of control chart data and process reliability estimates to better predict and monitor the manufacturing process.The first step in our procedure involves analysis of historical control chart data. For each point on the control chart the following information is tabulated:- Value of the control chart statistics.- Elapsed time since the process or manufacturing equipment was last adjusted or repaired.- Determination of whether or not the process or equipment failed or went out of control after the sample data for the point was gathered but before the next time for obtaining control chart data.With this information, it is possible to estimate probability of process failure as a function of elapsed time and control chart sample values. Simple graphical and statistical methods can be used to evaluate the usefulness of reliability and control chart information for predicting imminent process failure. This allows us to address the question: "Given that it has been t hours since the process was last adjusted and the value of the control chart statistic is y, what is the probability that the process will go out of control (fail) before the next time data is to be gathered for the control chart?" If this probability is too large (where "too large" can be defined in terms of a cost function) the process should be adjusted.The techniques proposed in this paper allows incorporation of the reliability characteristics of a manufacturing process into its monitoring process using existing data. It provides faster detection of process changes by more effectively extracting information from control chart data.