ARTICLES

The effectiveness of SPC applications

Dr Juergen Ude Dr Juergen Ude

Introduction

This article examines the effectiveness and problems of Statistical Process Control (SPC) as popularly applied by Australian Industry.

If we really wish to make competitive improvements, we will need to go beyond slavish adherence to methodologies developed in the 1920s. Many statistical methodologies of the ‘twenties’ were based on assumptions no longer appropriate in an age of readily available computer power.

Yet, to judge by most of the off-the-shelf software, the power of computers has not been fully utilised to achieve the greater power and productivity improvements previously thought possible. The quality profession, more than such professions as chemistry or engineering, is vulnerable to clever catch-phrase marketing. Thus, repeated exposure of the SPC concept accompanied by promises of improved competitiveness through improved quality and productivity gains has resulted in SPC being applied more and more without questioning whether training and the applications are appropriate, or whether the methodologies are appropriate. It is hoped that by examining some of the practices, applications and methodologies, it will be possible to humanise SPC – a need expressed by the AOQ’s Queensland Quality Centre in the introduction to Dr Niven’s article in Quality Australia, July 1991 issue. Readers may wish to question the alternatives suggested here.

Training

An important aspect of SPC is the effectiveness of training. Considerable resources are devoted to training; often everyone in an organisation is taught SPC. One has to question why training personnel other than those directly involved is considered necessary. (What would happen if chemists insisted on everyone in their organisation being trained in chemistry, or engineers insisting on everybody being trained in engineering?)

Companies are better advised to conduct a statistical application study through reputable consultants from industry, tertiary institutions or organisations such as CSIRO. Such application studies should include a training needs analysis. This would ensure that members of an organisation who are not directly affected by SPC would have more time to continue running the business. Such a needs analysis might show, however, the importance of awareness training for personnel not directly involved with SPC. The application study should also identify the tools for implementation. It is one thing to provide training; it is an entirely different matter to identify and provide the necessary tools, such as software.

Off-Line Applications

Although we pride ourselves on accepting the contemporary view that we must ‘get it right first’, most SPC applications seem to be designed for off-line control, where it is too late. While there is a need for off-line applications, many can be ineffective. Off-line applications are based on the belief that through the identification of assignable variation, dramatic quality improvements can be achieved through reductions in variability. Points outside control limits or non-random patterns are said to be due to assignable causes.

Theoretically, if we could simply identify assignable causes, then it would be possible to remove them and reduce variability, with the resulting reductions in variability improving quality and lowering quality costs as explained by Taguchi. In practice there may not be any notable quality improvement, as often there exists a threshold region beyond which further reductions in variability are unimportant. Thus we may unnecessarily divert scarce quality resources away from more beneficial quality improvement projects, and where most companies are running lean, this must become an important consideration. We cannot afford to waste time.

In practice, it is often not possible to identify assignable causes after the event; even Shewhart admitted that assignable causes come and go. Modern processes are frequently so complex that they make it humanly impossible to identify causes of variation after the event, and indeed during the event. The belief that for a process to produce high quality product it must be in a state of statistical control. i.e. no assignable variation is present, has been taken to extremes. There are instances where suppliers with variation less than one-tenth of the tolerance widths have been penalised because a point was outside control limits, whereas suppliers with variation greater than three-quarters of the tolerance limits were not penalised because their process was in control. Instead of commending the suppliers for a good job done, they were demoralised for their efforts. An in-control process is not necessarily better than an out-of-control process.

Another consideration is the economical value of removing assignable variation. It genuinely may be uneconomical to identify assignable causes. Shewhart himself did not consider out-of-control points as assignable causes of variation if it was uneconomical to identify them.

It is important to identify changes, as these could signify the occurrence or the likelihood of an onset of quality problems.

With everything there are exceptions, in general the effectiveness of off-line SPC as a quality improvement tool is limit and debatable. There are other more powerful effective ways of improving quality, such as through the design and analysis of indusrial experiments. A better application for off-line SPC, which may be considered as an indirect quality tool, is to monitor changes in quality levels. Quality practitioners are advised to use a simple off-the-shelf package for quality monitoring. It is important to identify changes, as these could show the occurrence or the likehood of an increase in quality problems. In the food industry a change in a total bacteriological count could indicate a change in sanitary conditions which could result in an increase in micro-organisms; yet only a few food industries apply SPC to the monitoring of bilogical levels. Monitoring is also used in establishing the effectiveness or in-effectiveness of quality improvement processes to detect improvements in the quality of finished product.

Effectiveness of SPC monitoring can be reduced by using Shewhart theory to calculate control limits. Most products are produced in batches or lots. It is practically impossible to centre processes exactly and each lot produced will have been from a different population. This is specially true for batches produced by mother nature (who does not practice SPC). There is unavoidable variation in the parameters of the populations monitored, the assumptions required to calculate Shewhart control limits are violated. Control limits will tend to be too tight and may thus have to be widened. Otherwise, costly action may be taken which will not improve quality but greatly reduce it. Genuine experience can play an important role in setting control limits. It is best to use a software package which can calculate realistic control limits in these situations.

On-Line Applications

Although it seems to be rarely used, on-line SPC is consistent with the “getting it right first” approach. Corrective action should be taken as early as possible, not after the event. On-line SPC is an area where we can make substantial quality improvements by controlling the process to target parameters. Here we are more concerned with taking corrective action, not with identifying assignable causes of variation and then removing them (usually there is insufficient time available for this in real-time mode).

On-line SPC is not easy and is often misapplied; for example, control limits are often placed around the process average. This stems from confusion between off-line and on-line SPC. With off-line SPC there is a need to identify changes – hence using the average as a centre line may be appropriate. For online process control, we should be more interested in maintaining the critical process parameter to a target value, otherwise the process controls itself, not the operator the process. One major problem is that variability often changes during production. In theory this should not happen, and in theory we should rectify the situation. In practice this does happen and the problem is not overcome easily until it is fixed, control limits and targets may have to be adapted accordingly. Such target management is very appropriate for weight and volume control: here the target is set as close as possible to the declared measure without infringing on weights and measures legislation. If variability increases, we have to increase the target, and if it reduces, we lower the target. Without target management, considerable padding will be required, thereby increasing giveaway costs.

For on-line process control, decision rules different from those for off-line control are required. Operators should react only to sustained changes in the process parameters, otherwise they will increase variability. Decision rules tend to be more complex, requiring special computer algorithms. One on-line problem is knowing which variables to control. Little is gained by controlling all sundry variables without knowing which are critical and which are not, and without knowing what the targets should be and their sensitivity. Sensitivity analyses are used to establish the degree of control required. To effectively apply on-line SPC first requires process study, using the techniques of statistical process control. A much overlooked problem is how to adjust a process, which settings do we adjust, what raw materials should we add and how much. There are more complications to consider. Conventional SPC is probably better than no SPC, but if we are to become clever, we will need to consider customised expert systems process control, now being used by several large organisations and yielding savings of over $1 million pa. These systems are easy to use and typically involve work stations networked to a control-room monitor for effective supervision.

Conclusion

SPC has received mythical status through clever marketing strategies. There is no doubt that SPC is a vital tool for all quality professionals: however, it must be creatively and adaptively applied. Using a rigid recipe approach can do more harm than good. Many of the current off-the-shelf software packages are best suited to off-line SPC. And used intelligently, represent a tool all quality professionals should have. For on-line SPC, there is a need to go beyond basics into the area of expert systems.

Download the Inferences APP, comprised of mainstream and machine-powered analytics for statistical analysis

Analytics as a Service (AaaS) for Quality

Drive quality improvement through actionable insights using analytics you can trust! Use up to 200 analytics tools downloadable through a suite of Apps!

FREE usage of the analytics Apps for quality improvement
  • Augmented with machine-powered smarts
  • Always updated with the latest tools and features
  • No licencing or fixed subscriptions - Pay ONLY for the analysis you run from 20 USD cents per analysis, billed monthly! Set a budget so you don't exceed!