Statistical process control coincides with the lean values concept that the tools of improvement should enable everyone’s participation. Engineers and managers can improve the inherent capabilities of the process and frontline teams can identify and remove special cause variations as they occur in operation. In the process industries, improvement derived from the efforts of managers and engineers is a basic capability. Therefore, it is likely that the most important new contribution of both lean and SPC is to enable autonomous frontline teams to contribute to the improvement pro- cess. This is the capability that will distinguish you from your competitors and SPC does this in a number of ways.
Assuming that someone, probably a statistician or six-sigma black belt, has provided the operating team with a mathematical model of its process,
the frontline team can now do many things it could not previously do. Normally, the mathematical model is deployed to the frontline in the form of a tool called a “run chart.” In many chemical and process plants, this statistical model is configured within the unit console as control algo- rithms. Whether process technicians perform the analysis or it is done for them as part of the console programming, it is useful for you and your team to understand the nature of the exercise.
Key idea: The run chart continuously refreshes the statistical thinking
and understanding your team has about how its process should oper- ate and how it is actually operating in comparison to expectations.
using a run chart
The run chart is nothing more than a continuous, real-time comparison of the expected performance of the operation as predicted by the statisti- cal model with the actual performance achieved as the unit operates. That comparison is supplemented by a statistical understanding of what the various comparisons mean. The practical element of SPC practice at the frontline is a series of “run rules” that describe the proper action to be taken in response to each operating scenario described by the run chart information. This analysis provides frontline teams with new understand- ing of the real-time performance of their operations and new responses to drive improvement. Let us see how this works in practice.
When the Run Chart Says the Process Is Operating Normally
The process is determined to be operating normally when the perfor- mance achieved is in accord with the expectations of normal performance as described by the statistical model of the process. Whether the produc- tion process is a capable process for the current product or an incapable process, in normal operation, it should produce results within the range of values predicted by the process model and randomly distributed within that range. That is how the process model was constructed. As a result, when the operating team collects current performance data that exhibit the expected range and distribution, its members will know that the pro- cess is operating normally. This knowledge is very valuable in determining the actions of the operating team as prescribed by the run rules.
Run rule: When real-time data demonstrate that the process is oper-
ating normally, there is nothing that the operators can or should do to change process performance. Operators should devote all of their improvement efforts to the care of the equipment (see Chapter 9) or other routine activities.
For operators of capable processes, this run rule makes a lot of sense. Everything is fine. Leave it alone. For capable processes that are running as expected, any adjustment is more likely to do harm than good and the operators can easily understand that.
This run rule is more difficult for operators of incapable processes to understand or accept, but it applies equally to them. If an incapable pro- cess is producing results that are all randomly distributed within the predicted performance range of the process, it is natural and expected for an occasional performance result to fall outside the customer speci- fication. When this happens, most operators want to adjust the process; however, in this case, the statistical understanding remains that such an adjustment is likely to make ongoing performance worse rather than better by introducing the special cause impact of altering the center of the distribution.
For example, if the real-time production sample shows a result that is out of the product specification range to the high side of the range, it is normal for the operators to want to compensate by adjusting the center point of the process toward the low side. Unfortunately, if the process is performing normally, it was originally centered within the specification range of the product. Adjusting the center point toward the low side will indeed reduce the number of results that are normally out of range on the high side. However, it will increase the production of product that is out of range on the low side by an even greater amount because a special cause bias toward low-side production has been created when the process was recentered.
Your frontline teams can make other valuable contributions to improve performance when an incapable process is operating normally. Recognizing that they have that opportunity and further understanding that they should not attempt to produce performance that exceeds the capability of their process allow them to capture the value of those other contributions. Adjusting an incapable process that is operating normally has no value and perhaps makes performance worse.
Key idea: Whether the process is capable or incapable, adjusting the
process when it is operating as it is expected to operate will normally make the situation worse, not better. You and your operators need to understand this or you will be chasing performance that you cannot achieve, perhaps making things worse, and wasting efforts that could be better spent on other things.
When the Run Chart Says the Process Is Producing an Unexpected Result
The second opportunity to use the statistical comparison of real-time results to the process model is when the process produces an unexpected result—that is, when routine monitoring of current process performance produces a result that is not predicted by the process model. Again, whether the process is inherently capable or incapable, the run rule that drives frontline actions is the same. Whenever the process produces a result that is not within the range of predicted results or not properly distributed within that range, the operators can know with certainty that a special cause external to the process is producing that result.
Run rule: Whenever the real-time process result indicates perfor-
mance that is not predicted by the process model, the operating team knows that a special cause is affecting performance. Therefore, team members must identify and eliminate it as rapidly as possible so that the process can return to normal operation.
When the process is operating normally, the frontline team will make the situation worse if it intervenes; however, when there is a special event, it typically must intervene. Processes affected by a special cause generally require some immediate action by the frontline team in order to return to normal. Even if the special cause is transient and the process will recover by itself as the event passes, prompt action by the team to isolate and iden- tify the special cause while it exists is a necessary proactive step if the team is to prevent that same event from recurring in the future. Some processes experience frequently recurring special events. Resolving a recurring event or preventing an episodic event from becoming a recurring event will have an operating impact similar to inherent improvement of the process.
The teams are assisted in their search for improvement of special causes of variation because they know that whatever caused the abnormal per- formance is not common to their process; therefore, they are looking for something that has changed. Further, if the team continuously and carefully monitors performance, it also knows that whatever is causing the abnor- mal performance is happening at that time. That close correlation between cause and effect is a great benefit to problem identification and resolution. Many teams detect changes in performance so quickly that the life span of an abnormal event before it is identified and resolved is quite short. By reducing the frequency, extent, or duration of abnormal events, teams can again have an impact that is similar to improving the inherent process.
Frontline teams are uniquely able to contribute to this improvement opportunity. Only frontline teams are always present and only they can always correctly answer the questions “What just changed?” or “What new is happening now?” By providing your frontline teams with a model of expected performance that enables them to transform routine process observations into a real-time search for abnormal variation when it exists, you will create a truly powerful improvement capability that you cannot access in any other way.
When the Run Chart Says the Process Is “Nearly Normal,” but Results Are Drifting
The third opportunity for benefiting from the statistical model and statis- tical thinking to help your frontline teams is when the real-time process results are contained within the range of predicted performance, but the results are no longer randomly distributed within that range. Again, this situation and the appropriate responses are common to both capable and incapable processes. Two classic examples are generally used to illustrate this situation:
1. All of the real-time results are within the predicted range, but all of the most recent results are on the same side of the normal middle of the range. This result usually indicates that the process needs to be recen- tered. Although it is inappropriate for the operating team to recenter a unit that is performing normally, when team members detect that performance is abnormal—specifically, when they detect that the center of the range has drifted away from the intended target value— adjustment is not only entirely appropriate but also mandatory.
2. All of the real-time results are within the predicted performance range—they are even fairly distributed on both the high and low sides of the expected center of the distribution—but there is a clear nonrandom pattern to the most recent results. This nonrandom pat- tern is likely to appear as a trend in the results; for example, several recent points may be fairly distributed on both sides of the mean but, considered together, they are demonstrating consistently increas- ing values. Because the statistical model of expected performance predicts that the performance results will be randomly distributed within the range, if the results demonstrate a clear trend within the range, then something is not normal. Unlike the situation where all of the results are suddenly on the same side of the center point, the nature of what has happened is not intuitively apparent, but clearly something has happened and the team needs to intervene to identify the source of the variation and resolve it.
Either of these two situations is only observable through a subtle anal- ysis of performance that certainly would be impossible at the frontline without statistical capabilities to assist the team. With this analysis, team members can intervene to return the process to normal operations, even when no individual measured result has yet demonstrated that the process is not performing normally. By enabling a proactive response to correct an operating problem before any adverse result has occurred, SPC provides your operating teams with a great new capability. In a situation like this, no independent tangible event has yet occurred to indicate the need for a reaction, so no other improvement effort could ever match this result.
Run rule: When the center of the range of production changes or
when the results within the range of current production become non- random or demonstrate a clear trend within the range, some “spe- cial cause” event is happening that requires the team to respond. The response may be a simple recentering or something more complex. The operating team needs to identify and resolve the cause of the abnormal results.
Through careful analysis and monitoring, the team has an opportunity to respond before any production consequences occur. Statistical process
control provides your frontline teams with the ability to recognize and distinguish situations where the process is operating normally and adjust- ment is improper from those where a special cause is affecting produc- tion and adjustment is mandatory. Again, this is a proactive capability that cannot possibly be created within a traditional reactive process manage- ment regime.