Statistical Process Control Instability

Understanding the SPC charts is one of the guides to problem solving quality issues.  However before we start looking at the patterns that can occur on the charts and what they can indicate, lets understand another important feature of control charts.

So for example taking sample 3, the measured dimensions could have been 50.3; 49.7; 50.0; 50.1 and 50.2 = 50.06 mm average and 0.6 mm range.  Those two figures would have been plotted in the sample 3 column of the chart.  Mr Stat used a previous set of plotted points of averages and ranges to work out the position of the control limits.  IN CALCULATING THE CONTROL LIMITS HE IGNORED THE DRAWING TOLERANCES (CAUTION - SEE IMPORTANT NOTE AT END OF THIS SECTION).  Mr Stat looked at the process results and worked out control limits that related to how the process was performing.  This means that when you operate SPC X-bar and R Charts you are checking the performance of the process.  It is important to remember, you are not checking the part against the drawing limits.

The control limits that Mr Stat calculated must be inside the drawing limits otherwise you will continually make some scrap parts.  Remember at the start of this section on statistics Mr Stat conducted a machine capability study to make sure that so long as everything was OK the machine could produce parts within the drawing specification.  In fact he made sure that there was some latitude either side of the body of results.  You could well find that the control limits that Mr Stat has set are very much tighter than the part drawing is requesting, but if you venture outside those control limits you must take action.  The reason for this is if the results drift out then something has changed in the process to cause it.  OK so now lets take a look at some chart patterns.  

Process not stable, due to points outside control limits.

Stop the process and take some action and (IMPORTANT è) record the action taken on the chart  

Possible causes for TREND on X-bar chart

Tool or fixture wear; build up of dirt; gauge wear; gradual deterioration of equipment and operator fatigue; cold start-up.

Possible causes for TREND on R chart

Improvement or deterioration of operator skill; loosening or wearing of fixture; in-coming material or parts changing and operator fatigue.  

Possible causes for RUN on X-bar chart

New set-up; batch change of in-coming material or parts; change of person or inspector; change in method of operation; change in testing or measuring equipment; machine adjusted; improvement made to process but machine not reset to the centre.

Possible causes for RUN on R chart

Batch change of in-coming material or parts; change in operator or inspector; change in method of operation; fixture or tool breakage (springs etc.)  

Process not stable due to CYCLIC PATTERN reoccurring over a period of time.

Possible causes for CYCLIC PATTERN on X-bar chart

Environmental changes; regular rotation of operators; power supply fluctuations; different machine stations being checked in set sequence; different batches of in-coming material or parts being used; machine resetting to different positions.

Possible causes for CYCLIC PATTERN on R chart

Rotation of jigs and fixtures; scheduled preventative maintenance; operator changes due to fatigue.

Process checks resulting in LACK OF VARIATION plot of points.

Possible causes for LACK OF VARIATION on both charts.

Incorrect calculation of control limits; improvement of process without recalculation of control limits; method of measuring is not fine enough to discriminate between different size parts.

Process checks resulting in STRAIGHT LINE plot of points.

Possible causes for LACK OF VARIATION on both charts.

TEST EQUIPMENT OR GAUGES are faulty; change of person checking process who has not been correctly trained.

IMPORTANT NOTE although control limits are calculated from actual process data, the limits can also be set by using the drawing specifications and moving the control limits inboard of them a set amount.  The distance they are moved is based on statistical data gathered from the process.  These limits are called “modified limits”.  Which ever type of limit is being used you must remember that if the results stray outside a control limit you must stop the process, get it fixed and record the action on the SPC chart.