یک محیط شبیه سازی مبتنی بر مولفه برای تجزیه و تحلیل سیستم های کنترل فرآیند آماری

This paper describes a simulation environment, called Prosim, which permits a user to define components, subsystems, and their interconnections to analyse a statistical process control (SPC) system. The components and systems are defined and analysed interactively. A library of standard SPC objects containing models for the Xbar, range, exponential weighted moving average, p-chart and other SPC techniques have been created which help define the control application. The PC-based tool is tested on theoretical, and real data, and is useful for the design and trouble shooting of a manufacturing system. It is also an effective teaching and research tool.

This paper explores the use of discrete event simulation for the assessment and design of a quality system. The basic premise is that simulation can be used for exploring and investigating the optimal design of a Statistical Process Control (SPC) system. Prosim, an object oriented continuous time simulation system based on the Graph Theoretic Method (GTM), is utilised as the basis of the simulation modelling technique. This approach has the advantage that each component or subsystem can be separately programmed, turned into an object, and handled independently in time. Other discrete event simulation software can also be used. However, Prosim provides the flexibility required for a unified and systemic approach to the design and modelling of SPC systems. Three random number generators based on the rectangular, normal and exponential distributions are developed to represent time-based processes found in typical manufacturing systems and in the service industry. Other distributions including non-parametric and empirical distributions can be read directly from data files. Simple out of control mechanisms are explored by implementing an interactive capability which allows the process to go out of control by a combination of assignable causes such shift or drift of the location (mean), or spread (standard deviation). Processes can also exhibit short-term correlation by the implementation of an autoregressive process. Statistical Process Control techniques such as variable control charts (Xbar, Range and exponential weighted moving average (EWMA) charts), attribute charts (p-chart), process capability, quality costing, and inspection are implemented. These techniques are represented as components which form a toolbox of SPC procedures that can then be configured in the Prosim environment (using drag and drop facilities) to represent any SPC system. The system can be easily extended to include other statistical process control techniques. The purpose of this paper is to illustrate and explore the use of a discrete event simulation system to determine how control charts or other SPC techniques perform in realistic situations. Also, the paper illustrates how a simulation tool can be used as decision support so that the analyst can determine which SPC technique is the most appropriate under a set of given conditions. This is illustrated through a series of examples, presented in Section 4, which use operating characteristic curves, average run length charts, response surfaces and scenario analysis of theoretical and actual manufacturing systems, to outline the potential of implementing the approach to assist the quality engineering of a particular system. 1.1. Statistical process control using Prosim Simulation has been suggested for the assessment of quality characteristics by several authors. (see for example [1], [2], [3], [4] and [5]). The underlying advantage of simulation techniques or tools is their capability to model variation. SPC seeks to control and minimise variation through feedback onto the process thus ensuring a continuous improvement system, based on prevention rather than detection. Several simulation studies have used procedural programming languages such as Fortran and C [4]. Although these languages are extremely capable and flexible programming tools the degree of complexity necessary to develop a comprehensive simulation tool, with time advance mechanisms such as next event scheduling, is a significant limitation. Purpose-built discrete event simulation environments have been developed to investigate the operational characteristics of manufacturing systems. Systems such as Witness™ [6] and Arena™ [7] include excellent graphical interfaces and animation for investigating features such as bottlenecks, length of queues and the rate of utilisation of manufacturing systems. However these systems do not provide an environment flexible enough to easily develop the control charting mechanisms or the out of control characteristics necessary for a detailed simulation of quality control. Discrete Event Simulation packages usually focus on the investigation and analysis of productivity and so it is difficult to isolate SPC issues. Although those packages which include programming capabilities can be programmed to investigate SPC for a specific system they cannot easily be configured for a generic analysis of SPC systems. Prosim offers a useful compromise. It is a time driven system, which means that events can be naturally programmed in time using a modular programming language. Components, handled as objects, can then be connected together to provide a complete system. Thus components such as random number generation with assignable causes, control charts and feedback operations can be coupled together to provide a complete system of quality operations. Once the code has been developed it can be represented as a single component which can be connected with other components to form a system. Fig. 1 shows an example of a system consisting of a Xbar chart monitoring a machine producing random data with a normal distribution. Full-size image (39 K)Fig. 2 shows a more complex SPC system, which extends the system shown in Fig. 1 to include process capability analysis, costing and written output to a file. Once developed, individual components can be stored in a library to form a toolbox of quality components which, using the drag and drop facility of the graphical interface, can be linked together in different system configurations.The key features of Prosim that make it flexible are: (a) independent modelling of each control chart or strategy as a component which is handled by its icon and can be connected with other components to form a system, (b) time-based programming, which keeps track of the system clock as various processes/components perform their actions, and (c) the capability to store the component and systems models in a library or a toolbox for use later. This approach to simulation offers the quality practitioner several important advantages: (a) Simulation models can be used to explore system behaviour. The simulation study can be used to predict the theoretical characteristics for a particular quality control procedure. When attempting to derive theoretical solutions to industrial problems however, the quality practitioner is faced with two significant problems. 1. There is limited access to theoretical results in the industrial environment. Many books concerned with SPC techniques present details of the techniques and their implementation without providing a theoretical justification. Such details are usually confined to classic texts such as Duncan [8] or research publications which are not readily available in the industrial environment. 2. Theoretical solutions are based on ideal theoretical conditions such as normality and single assignable causes and as such limit the applicability of these models in the industrial environment (b) Modularity of simulation approach provides flexibility. The modular approach to programming and system configuration provides the ease of use and flexibility to configure a system for a particular requirement in much less time than the research required to access or develop a theoretical model. Prosim simulation models are not restricted by theoretical considerations as a simulation model can be constructed for any system of known characteristics. Non-Gaussian processes and multiple assignable causes can be programmed as components which more realistically represent an industrial system. (c) A Simulation approach is useful in system design. The Prosim system can be used to evaluate, test or improve an existing system, by investigating the use of alternative SPC techniques, determining when, and where to implement SPC tools and feedback loops, and to study the operational performance of standard techniques on a particular system. These components can be configured and reconfigured very quickly so that alternative scenarios can be investigated. Once the programs have been developed and stored as a library of components the system can be used by a practising engineer to configure and analyse SPC systems. The simulation study can be directed towards improving an existing system or provide detailed analysis for optimising the implementation of a new system. Thus it is possible to perform a detailed simulation of a system before it is implemented to reduce quality problems due to start-up and during the commissioning of a system so that quality characteristics can be better understood. In this way it is possible for the practitioner to build a Prosim model which runs in parallel with an actual system, such that the simulation model accurately reflects the quality characteristics of the actual system as events are occurring. The simulation can then be updated with the most recent information and then “fast forwarded” to investigate future quality problems and to determine the overall control characteristics of the system over a long period of time.

This paper has demonstrated how discrete event simulation can be employed for the design and implementation of statistical process control systems. Prosim provides a powerful platform on which to base the simulation. The component-based structure of Prosim provides the facility to build rigorous system models with the minimum of effort and the flexibility of a time-based programming language allows complex structures to be developed. The tool is based on the concepts of the graph theoretic systems modelling approach, which has wide acceptance in dynamic systems modelling including manufacturing systems. Prosim provides a flexible and user friendly environment for creating, testing, and planning different quality control strategies in an interactive manner. The simulation approach presented permits a systematic method for investigating the different configurations and SPC strategies that are beyond a theoretical treatment. It permits a quality practitioner to investigate complex (existing or planned) systems. Although until now simulation of quality systems has been firmly in the hands of the researcher it is hoped that the further development of simulation tools such as Prosim will provide a powerful decision making tool for the practising quality engineer. It is envisaged that this can be achieved by developing a flexible and comprehensive “toolbox” of techniques which once programmed by an “expert” can provide the practising engineer with the facility to configure a system to his or her particular requirements. Future development of the SPC toolbox to include a wider range of control charts such as CUSUM charts, adaptive control charts, and charts for autocorrelated processes. The simulation can be used in situations where theoretical results are not available either because of lack of material or due to the complexity of the real industrial situation. This is demonstrated in the examples presented in Section 4. Here Prosim has been shown to accurately replicate theoretical results as well as extend these in more complex situations where theoretical solutions are not possible. The use of Prosim and discrete event simulation inherently promotes a systems approach to statistical process control and it is hoped that the studies presented here provide the basis of a holistic approach to quality. It is envisaged that further development of the simulation facility will focus on an event-driven approach where machine breakdown, material flow, and manufacturing strategies can be more accurately modelled. In keeping with a systems approach it is hoped that the models can be extended to the enterprise level to reflect the strategic management decisions, the interactions, and inter-dependencies among the elements of the system.