3-Hour Virtual Seminar on Statistical Process Control (SPC), Process capability, And Non-normal Distributions

This 3-session seminar will cover:

• Process capability and process performance, which reflect the ability of a process to meet specifications. A Six Sigma process will, when centered on the nominal, have only two nonconformances or defects per billion opportunities. (3.4 defects per million opportunities assumes a major process shift of 1.5 standard deviations.)
• Statistical process control (SPC) is closely associated with process capability analysis because (1) the data collection and analysis procedure is essentially identical and (2) a state of statistical control is a prerequisite for a meaningful capability study.
• Traditional SPC and capability analysis assume the availability of large quantities of data from well-established processes. Short run SPC is however usable for start-up processes as well as processes that make parts with multiple characteristics, and/or characteristics with different specifications. These charts support kanban and mixed model production control systems.

Attendees will receive copies of the slides and accompanying notes, and also Excel spreadsheets with the example data used in the presentations.

Why You Should Attend:

• To improve the writing of SOPs for more effective training and reduction of errors
• To understand the regulatory implications of what is written in an SOP
• To learn the parameters of an effective SOP
• To distinguish a well-written SOP from a poorly written one
• To define processes better by effective interaction with the SOP process owner and/or author
• To integrate the SOP effectively into the position curricula of employees
• To explore the connection between SOPs and job performance, and measure retention of SOP content at the training stage

Seminar Agenda:

Session 1: Process capability and process performance

• Relationship between variation and quality; a process that behaves like a rifle is better able to meet specifications (hit the target) than one that behaves like a musket.
• Processes, especially batch operations as opposed to single-unit flow, may include long-term variation such as batch to batch variation that is not accounted for by traditional samples. We must accordingly select the rational subgroup, a sample that represents all variation sources, correctly.
• Traditional process capability and performance calculations rely on the assumption that the process follows the normal or bell curve distribution, which is far more common in textbooks than in the real world. There are however off the shelf methods to obtain process performance indices for non-normal distributions.
• A capability study relies on the assumption that the process is in a state of statistical control, i.e. special or assignable variation causes are not present. An SPC chart should therefore be part of any capability study.
• A sample capability and performance study will be provided.

Session 2: Traditional statistical process control

• SPC is a visual control that makes the status of the process obvious without the need to interpret tables of data. It is a graphical hypothesis test that tests the starting assumption (the process is in control and does not require adjustment) against the alternate hypothesis that the process’ mean has shifted—i.e. the “rifle” is no longer sighted in on the bulls-eye (nominal)—or the variation has increased, i.e. our “rifle” has been replaced by a smoothbore musket.
• As with process capability studies, the rational subgroup must be selected correctly. A batch process is likely to have between-batch as well as within-batch variation which will result in charts that seem to be out of control for no identifiable reason. The nature of these charts (points outside the lower and upper control limits) shows that the control limits are based on only the within-batch variation.
• Traditional SPC also relies on the assumption that the process follows the normal or bell curve distribution. When the distribution is non-normal but identifiable (e.g. gamma, Weibull, lognormal), however, control charts can be created to reflect their performance accurately.
• Control chart setup includes calculation of the control limits based on the process statistics.
• Control charts are often part of a process’ control plan, and action on out of control signals is mandatory.

Session 3: Short Run SPC

• Traditional or “textbook” SPC relies on the collection of data from 20 or more rational subgroups to obtain good estimates of the process’ mean and standard deviation. This kind of data is often not available in job shops, processes that make a wide variety of parts, and of course new processes with no prior history. The Automotive Industry Action Group’s CQI-26 SPC Short Run Supplement and also in Stephen Wise’s and Douglas Fair’s Innovative Control Charting (1998) offer practical solutions for these circumstances.
• The DNOM (Deviation from Nominal) chart assumes the process is centered on its nominal or target value, and also makes informed assumptions about the standard deviation (e.g. from similar processes).
• Charts that assume equal variances for parts with different specifications are very suitable for kanban production control systems and mixed model production (e.g. ABCABC versus long runs of parts A, B, and C respectively). These charts can be deployed in Excel.
• Charts for multiple product characteristics, and group charts, are available for parts that have multiple features for which specifications must be met. The group chart can be handled by StatGraphics.

Topic Background:

There is a direct interrelationship between variation in product characteristics and quality, and variation is why purportedly interchangeable parts are sometimes not interchangeable. Process capability indices measure quantitatively the amount of variation in a process, and therefore the process’ ability to meet specifications. Statistical process control (SPC) meanwhile provides advance warning that a process’ mean is drifting away from the nominal, or the process’ variation has increased, and an SPC chart is an important supporting element for a good capability study.

Customers often request process performance metrics from their suppliers, and this may be mandatory in the automotive sector (IATF 16949:2016 clause 9.1.1.1). Even if it is not mandatory, it is extremely useful to assess the process’ ability to meet specifications. It is therefore vital to know more than just the mathematical mechanics of their computation. These metrics can be highly misleading, and even off by orders of magnitude in terms of the process’ nonconforming fraction, because of:

• Incorrect selection of the rational subgroup (a sample that reflects all variation sources in the process). As an example, within-batch variation, as might be measured by a sample from a single lot or batch, does not reflect between-batch variation. When the estimated long term variation exceeds the short term variation, this is a likely cause.
• A non-normal process distribution. The traditional calculation relies on the assumption that the process data conform to a bell curve, but performance indices can be calculated for non-normal distributions as well.
• A process capability study also relies on the assumption that the process is in a state of statistical control because assignable or special cause variation will affect the estimated standard deviation (variation) and/or the process mean. An SPC chart should therefore accompany every capability study.

Process capability and process performance assessment are meanwhile synergistic with statistical process control. SPC can provide advance warning, in the form of points outside the control limits, that process capability and performance have been reduced by (1) an increase in variation or (2) a shift in the process mean. If the process is no longer centered on the nominal, then the fraction of nonconforming work will increase.

• Traditional SPC is also, like process capability analysis, subject to the need to select the rational subgroup properly and also the normality assumption. There are also off-the-shelf SPC charts for processes that do not meet this assumption.
• Traditional SPC also assumes that enough data are available to characterize the process mean and standard deviation. Short run SPC methods are however available for start-up processes, processes that make small quantities of parts, and processes that make parts with different specifications.

Who Should Attend:

• Quality Assurance Departments
• Quality Control Departments
• Research and Development Departments
• Manufacturing Departments
• Engineering Departments
• Operations Departments
• Production Departments
• QA/QC Technicians
• Manufacturing Technicians

Course Director: WILLIAM LEVINSON ‎

William (Bill) A. Levinson, P.E., is the principal of Levinson Productivity Systems, P.C. He is an ASQ Fellow, Certified Quality Engineer, Quality Auditor, Quality Manager, Reliability Engineer, and Six Sigma Black Belt. He holds degrees in chemistry and chemical engineering from Pennsylvania State and Cornell Universities, and night school degrees in business administration and applied statistics from Union College, and he has given presentations at the ASQ World Conference, TOC World 2004, and other national conferences on productivity and quality. Mr. Levinson is also the author of several books on quality, productivity, and management. Henry Ford’s Lean Vision is a comprehensive overview of the lean manufacturing and organizational management methods that Ford employed to achieve unprecedented bottom line results, and Beyond the Theory of Constraints describes how Ford’s elimination of variation from material transfer and processing times allowed him to come close to running a balanced factory at full capacity. Statistical Process Control for Real-World Applications shows what to do when the process doesn’t conform to the traditional bell curve assumption.