Inhibitors of Statistical Process control

A number of factors can inhibit the implementation of SPC. With SPC, there is not usually the kind of philosophical re­sistance that is common with some aspects of total quality management. However, it is true even with SPC that there must be a management commitment because there will be start-up costs associated with implementation. The most common inhibitor of SPC is lack of resources.

1. Capability in Statistics

Many organizations do not have the in-house expertise in statistics that is necessary for SPC. As SPC is being intro­duced and decisions are made on where to sample, how much to sample, what kinds of control charts to use, and so on, a good statistician is necessary to ensure the validity of the program. If the organization does not employ such an expert, it should either hire one or retain the services of a consultant for the early phases of the SPC implementation.

The danger inherent in not having statistical expertise is developing an SPC program that is statistically invalid— a fact that can easily escape nonstatisticians. The organiza­tion will count on the invalid SPC implementation to control processes when in fact it cannot. A flawed SPC implementa­tion may send messages that make the process control situ­ation worse than it was before. It is important that the initial design of the SPC program be valid. This requires someone
with more than a passing knowledge of statistics. If there are any doubts, get help.

2. Misdirected Responsibility for SPC

Too many companies make the decision to use SPC but then turn it over to the statisticians or the quality assurance de­partment. The value of those departments should not be minimized, but the owner of the process in question should be the person responsible for SPC. This person is the one who can make best use of SPC, and there will be no question about the validity of the data because he or she is the one col­lecting it. The process operators will require help from the statistician and others from time to time, but they are the appropriate owners of SPC for their processes.

When someone else is responsible for SPC (mean­ing collecting and logging data, making corrections, stop­ping out-of-control processes, and getting them fixed), process operators see the entire SPC program as just an­other check on them, and they are very uncomfortable with it. Management tends to see it the same way, but from their particular perspective—a means of checking up on the op­erators. Nothing good will come from such a relationship.

Neither is ownership by the statistician the appropriate answer. If the statistician owns SPC, he or she is more apt to find fascination in the numbers themselves than in what they mean in terms of quality. Even if statisticians are tuned in to the objective, the operator will see them and their SPC charts as just another intrusion.

If operators have the responsibility for SPC, they will be­come familiar with the tasks involved and will see it as a means to help them get the most out of their processes. This is the payoff. All of the others need to observe and review, assisting when needed but never usurping the operator’s ownership.

3. Failure to Understand the Target Process

Unless a process has been flowcharted recently and charac­terized, the odds are good that the people designing the SPC system for it do not know how the process actually works. Most processes have evolved over many years, changing now and then to meet the requirements of the market or the de­sires of management or operators. Few are adequately docu­mented. People are usually astonished when flowcharts reveal the complexity of processes they thought to be straightfor­ward. A good SPC system cannot be designed for a process that isn’t fully understood.

4. Failure to Have Processes Under Control

Before SPC can be effective, any special causes of variation must be removed. This was discussed earlier, but it is ap­propriate to mention it here again as an inhibitor of SPC. Remember, by definition, a process is not in control if any special cause is working on it. The use of control charts as­sumes an in-control process. Their use will set off visual alarms whenever a new special cause is introduced. But the real work of process improvement can come about only when nothing but the common causes are active. This is why SPC is so powerful. It will show when common causes are the only causes of variation so that improvements to the fun­damental process can be made. Will special causes still come up from time to time? Certainly, but this is different from trying to control a process with special causes constantly present, masking both the common causes and each other.

5. Inadequate Training and Discipline

Everyone who will be involved in the SPC program must be trained not only in data acquisition, plotting, and interpreta­tion for control charts, but also in the use of the seven tools. Not everyone needs to be a statistics expert, but all need to know enough so that with a statistician’s help, the program can be designed and operated.

Training should teach that SPC and tweaking do not make a good pair. If tweaking is permitted, SPC data will be meaningless. The process may appear to be more stable than it is if the person doing the tweaking is an expert, or it may show more variation than if left alone. It must be understood that operators and engineers alike are to let the process run essentially hands-off until an out-of-limits condition is de­tected. Variation between the limits is not to be tweaked out. The only acceptable means of reducing the variation is a real process improvement that will narrow the limits permanently.

6. Measurement Repeatability and Reproducibility

SPC data are the result of measurement or count. In the case of the variables data (the measurements), the data become meaningless when the measurements are not repeatable. For example, a worn instrument or a gauge with insufficient precision and resolution might yield measurements over an unacceptably wide range when measuring the same object repeatedly. This is not satisfactory. The data taken from all measurements must be accurate to the degree specified, and repeatable, or there is no point in recording them.

Nothing should be taken for granted. Before any gauge is used for SPC, it should be calibrated and its repeatability certi­fied. It is also important that different operators obtain the same readings. This is known as reproducibility. Before getting them involved with SPC, certify all gauges and train all operators.

7. Low Production Rates

Although it is more convenient to implement SPC with pro­cesses that have continuous flow, or high rates of product output, it is by no means impossible to apply SPC to low- rate production of the type that is often found in a job shop setting. In a factory that produces several hundred printed circuit boards per day, sampling schemes are relatively easy.

A job shop might produce only a few boards in a day, often with gaps between production days. Sampling there must be done differently. Low-rate production provides an opportunity for taking a 100% sample. It is possible to take a sample from every board. In such an application, a computer-generated random x-y table can designate a spe­cific small area of a board for inspection of solder joints or other attributes. From that, a number representing the frac­tion defective may be developed. Control charts are easily constructed for fraction-defective data. Low production rates are not a good excuse for avoiding SPC.

Source: Goetsch David L., Davis Stanley B. (2016), Quality Management for organizational excellence introduction to total Quality, Pearson; 8th edition.

One thought on “Inhibitors of Statistical Process control

Leave a Reply

Your email address will not be published. Required fields are marked *