15. Control Charts

What is a Control Chart?

A control chart is a graphical representation of data used to analyze variation in different processes. Control charts can be used to: • Measure current performance • Analyze causes of variation • Identify opportunities to reduce variation • Control a stable process • And identify occasions when maintenance, tool changes, and adjustments are required

Types of variation

Before we move further into the discussion of control charts, it is imperative that we familiarize ourselves with two types of variation present in any process: Special Cause and Common Cause. Special cause variation arises in a periodic fashion and is somewhat unpredictable. Examples of special causes are; operator errors, broken or worn out machines or major IT system breakdown. This type of variation is not critical and represents only a small fraction of the total variation that we may find in a process and can be dealt with at the operator or machine level. Common cause variation is inherent in the system. They are always present and affect the output of the process. Examples of common causes of variation are poor training, inappropriate production methods, poor work station design etc. People erroneously assume that defects, mistakes or accidents are caused by special causes when in fact they are most likely caused by common causes which lie in the system.

When to use control charts

Control charts can play a pivotal role in problem solving methodologies. Now, let us identify the usage or control charts during the different steps of PDCA, DMAIC, 8D and DMADV. Starting with PDCA, control charts can be used to analyze the past data to devise a course of action in the plan phase. Also in check phase control charts can be used to analyze outputs and results thus help us evaluate performance. Moving to DMAIC, control charts can be used to collect data in Measure phase. Process behavior charts can be used for the analysis of the implemented actions and finally in control phase control charts can be used for continuous process monitoring. Looking at 8Ds, control charts can be used in D4,D7 and D8 to investigate, implement and verify the process. Control charts can also be used in the verify stage of DMADV to evaluate the performance of a new design.

8 Tests to check if a process is in control or not

If any of the following eight tests is verified then the process is marked out of control and requires immediate action. Test # 1 states that if 1 sample point lies outside the UCL or LCL, the process is not operating within the allowed limits. Test # 2 states that If 9 consecutive points lie on the same side of the center line, the process is rendered out of control. Such kind of pattern is also referred to as a SHIFT. Test # 3 This kind of pattern is also called a trend and the test states that a process will be out of control if 6 points in a row all increasing or decreasing are spotted on the control chart. Test # 4 states that a process will be out of control if 14 consecutive points alternating up and down appear on the chart. It is also an example of a CYCLE. Test # 5 states that If 15 points in a row show up within the green bands on either side of the CL, then the process is termed out of control. Test # 6 tells us If 4 out of 5 consecutive points lie on one side of the CL in the green band or beyond green band, the process is out of control. Test # 7 says, If 2 out of 3 consecutive points lie on one side of the CL, beyond the yellow band then the process needs attention. This test is another variation of a CYCLIC process. Test # 8 states that If 8 points in a row lie beyond green band on either side of the CL then the process is out of control.

Components of a Control Chart

Let us take a look at the three core components of a typical control chart: The central Line also called CL is the horizontal line that marks the historical mean of the process. The upper control limit also known as UCL is the horizontal line above the center line, that marks the upper limit of where the process tends to operate when it is in control The lower control limit or LCL is the horizontal line below the center line, that marks the lower limit of where the process tends to operate when it is in control. Data pointsare the recorded sample points from the process output, plotted on a control chart. They normally represent the sample mean values. Moving along, control chart is normally divided into 6 equal bands. • The outer bold lines indicate the upper and lower control limits and are the most critical places on the control chart. • The Orange bands next to red lines are somewhat less critical. Every now and then a point or two are recorded in this region even in a controlled process. • The adjacent Yellow bands show even less critical zone. Some sample points may lie in this region in a controlled process. • And finally the Green bands, closest bands to the central line are the least critical regions. Typically in a statistically controlled process around 65% of the points lie in the green bands.

Reading a Control chart

Now let us have a look at a manufacturing process example to further clarify the idea and usage of a control chart. Consider that an operator's goal is to ensure that the oven temperature is maintained within the prescribed limits. The operator collects sample temperature readings over time and constructs a control chart. On the chart the central line shows the desired oven temperature. The upper and lower control limits show the allowable tolerances for temperature variation. The operator observes that almost 65% of the plotted points lie within close proximity of the central line in the green bands. This is common for a statistically controlled process. The operator also notices some points in the yellow and a couple of points in the orange bands, even though yellow and orange bands are more critical. This tells you that variation is a part of every system and isn't necessarily a bad thing. Its normal to have a point or two in close proximity of the control limits every now and then. Finally, looking at the holistic view a process is said to be in statistical control if 99.73% of the plotted points lie within the prescribed tolerances, thus operator can be satisfied that his process is in control. Next let us have a look at the 8 tests we can use to check the same process if it is in control or not.

What is variation?

Variation is defined as change in the expected behavior, reliability, quality or quantity of a product, service or process. Variation is inherent in every process. Only when it exceeds the defined tolerances, we see its adverse affects on our process outputs. For instance an employee's arrival time to work. His time of arrival varies from day to day but it causes a problem only if he is late to work. Now you might be wondering why should I care about variation Well to put it simply, analyzing and controlling variation will; • Enable you to verify whether the process is in control or not. • Help you detect abnormalities in the process. • Assist you in investigating the causes of these abnormalities.