Lean Sigma: A Practitioners Guide

2017-07-07 02:10:07

43. Statistical Process Control (SPC)

Overview

Statistical Process Control^[82] (SPC) is the use of Control Charts to help control a process. Most novice Belts confuse SPC with Control Charts and vice versa. Control Charts are used in a number of places in the Lean Sigma roadmap, for example:

^[82] Use of Control Charts for Process Control dates back to the 1920's and Dr. Walter Shewhart of Western Electric.

To test for stability before performing a Capability Study in the Measure Phase

To test stability before performing a test of means in the Analyze Phase

To validate stability after project completion

SPC on the other hand is solely a Control tool and appears in the Measure Phase (to check what is currently being controlled) and the Control Phase (to control what should be controlled henceforth).

The key to understanding SPC is an appreciation of the different types of variation in a process,^[83] specifically:

^[83] A wonderful reference here, written in plain language rather than Statspeak, is Donald Wheeler's book, Understanding VariationThe Key To Managing Chaos.

Common Cause. The inherent variation present in every process, produced by the process itself. This is effectively the background noise in the process and can only be removed or lessened by a fundamental change in the process, usually the process physics, chemistry, or technology. A process is Stable, Predictable, and In-Control when only Common Cause variation exists in the process.

Special Cause. The unpredictable variation in a process caused by a unique disturbance or a series of them. Special Cause variation is typically large in comparison to Common Cause variation, but is not part of the underlying physics of the process and can be removed or lessened by basic process control and monitoring. A process exhibiting Special Cause variation is said to be Out-of-Control and Unstable.

Control Charts are used to find the signals (Special Cause variation attributable to assignable causes) in amongst all of the background noise (Common Cause variation).

SPC is placed on critical Xs in the process and uses Control Charts to detect when there are out of the ordinary events in amongst the regular background noise of the process. The Control element of SPC is that once an event is detected, action is taken to identify and remedy the cause. Without these controlling actions, someone accountable to make them and correct placement on the critical Xs, SPC does not exist. What exists instead, which is common in many misinformed groups, is a piece of paper with a graph on it.

Logistics

As mentioned in Chapter 5, "ControlTools Used at the End of All Projects," SPC is part of the Control Plan; the group of all tools, physical changes, procedures, and documentation that is used to ensure that process performance consistently remains at the desired level. SPC is not placed on every single X, it is placed on critical Xs that cannot be designed out of the process or controlled by physical means or with mistake-proofing devices. It clearly also relies on the ability to measure the X and respond accordingly.

The Process Owner should own SPC on an on-going basis, with little to no Belt involvement. If a Belt cannot walk away from the process at the end of a project, then their project isn't complete and more time needs to be spent on a more robust Control Plan and handoff.

Control should be made as close to the process as possible with Control Charts being generated by the operators, special causes detected, and action taken at that level. There needs to be clear management commitment to do this.

Accountability for SPC is typically on a key operator or line supervisor and it should be written into both their role and appraisal criteria.

Roadmap

The roadmap to setting up SPC on a process is as follows:

Step 1.	Identify the critical Xs or KPIVs for the process that are controlled using SPC. The biggest mistake here is to pull them out of thin air. The Xs should be chosen based on the following: They have been determined by analysis to be critical Xs in the process and drive a large percentage of the variability in the major performance characteristics, the Big Ys or KPOVs of the process. They cannot be designed out of the process. They cannot be controlled by physical methods. They cannot be controlled using mistake-proofing devices. They are measurable on a Continuous or Attribute scale.
Step 2.	From the data type of the X (Continuous or Attribute) determine the specific type of Control Chart required (see "Control Charts" in this chapter).

Step 3.	Use a sample of recent historical data to form a base chart. If none is available, then create a blank chart to be populated as data does become available. A common mistake here is to want to invent Control Limits. The process determines these.
Step 4.	Determine the actions required if the process shows an out-of-control condition. This could have been already determined earlier in a Failure Mode & Effects Analysis.
Step 5.	Train the Operators how to read the Chart and the actions required if the process goes out of control.
Step 6.	Transfer ownership of the chart to the Operators or Line Supervisor, whoever is the most appropriate.
Step 7.	Update all procedures and job descriptions to include accountability for the charts.
Step 8.	Hold Operators or Supervisors (as per Step 6) accountable by whatever existing means the business is currently managed.

Interpreting the Output

As mentioned previously, SPC relies on Control Charts to detect when an out of the ordinary event has occurred. See also "Control Charts" in this chapter. Control Charts typically take the form shown in Figure 7.43.1. Data is plotted over time across the x-axis, with the height on the y-axis representing the level of the X in question. From the data in the chart, "Control Limits" are calculated that represent the boundaries of reasonable behavior within the process. A point landing outside of these boundaries is considered special cause (out of the ordinary). The Control Limits are calculated from the process data itself using specific equations based on the data type. A statistical software package does this automatically.

Figure 7.43.1. Structur of a Control Chart.

The odds of a point lying outside of a Control limit are of the order of 300 to 500:1.^[84]

^[84] For an Individuals Chart used for charting normal data, the Control Limits are placed at ±3 Standard Deviations. Analysis of a Normal Distribution shows that approximately 99.73% of all data points should fall between these lines and hence falling outside is an event with probability 0.27%. See "Control Charts" in this chapter.

This is considered an unusual event. Obviously in a process that generates hundreds or thousands of entities, the occasional point falls outside the lines. Two in a row almost guarantees that something highly unusual has occurred or that the process has changed in some way.

Statistical software generally also highlights other unusual points for instances such as:

Points hugging the center line

Points oscillating back and forth

Points continuously rising or falling

For each, statistics are used to determine patterns that occur with odds at least 300:1 against.

The biggest mistakes made with the use of Control Charts for SPC are

Putting Product Specification limits on the Control Chart causes it to become just an inspection toolit is no longer a Control Chart at this point. The Control Limits tend to be ignored and focus is only on the specifications, or even worse the chart is incomprehensible to the Operator and is ignored completely. Understanding how a process performs against specification (known as Process Capability) is important, however, not in this application.

Treating the Control Limits as specification limits. The Control Limits are not directly tied to customer defects. If a point goes out of control it does not necessarily mean that the entity concerned is defective; it just means that the process has changed.

Flooding the system with Control Charts, and then not taking action on the data.

Not following up on unusual negative events to remedy them by determining and eliminating the root causes. In this case there is no "Control" in Statistical Process Control.

Not following up on unusual positive events to learn from them and capture the improvement. Sometimes the process suddenly gets better and the change is statistically unlikely; for example, the data points might suddenly start hugging the centerline. Often the mindset is to feel good about the process improvement, but to do little to understand why the process has improved. It is important in this case that Operators and Supervisors seek to capture the improvement and set this as the new standard going forward.