Statistical Process Control Essay, Research Paper

Statistical Process Control

Consistent high quality has become a requirement in today s competitive market. One vital tool for preventing problems is Statistical Process Control (SPC).

SPC is the use of statistics to analyze a process or output so actions can be taken to achieve and maintain statistical control and to improve the consistency of the process. (Jirka, 1995)

Statistical drawing conclusions using a scientific/mathematical approach of analyzing data

Process the whole combination of people, equipment, materials, methods, and environment working together to produce output; any work area that has identifiable, measurable output

Control making something behave in a predictable consistent manner (Jirka, 1995)

Benefits of SPC

Applications of Statistical Process Control in the manufacturing sector have been refined and tested with considerable success. Obvious reductions in production and warranty costs, and significant improvements in employee morale are the proven rewards for implementing SPC.

Using SPC methods, a company can:

+ Identify critical problem areas

+ Reduce variation and monitor for unusual variation

+ Determine the capability of the process

+ Understand and optimize the process

+ Determine the reliability of the product (”Statistical Process Control”, 1998)

New quality requirements such as ISO 9000 have forced many manufacturers to purchase and implement quality systems. Many companies now demand that suppliers practice statistical process control in meeting contractual specifications.

“Most of the major manufacturers are prodding people into using statistical-process-control techniques,” said Phillip Allen, President of Tricor Systems Inc., a small company that supplies hardware, software, and instrumentation products for large industries such as McDonnell Douglas Corp.

According to Allen, the company s assembly errors have been reduced by 43% since implementing SPC. (Chase, 1997)

Cost of SPC

Benefits obtained can quickly repay any investment to implement SPC. The cost largely depends on the method used to collect and process the information. SPC systems generally fall into three main types: manual, semi-automated, and fully-automated.

In a manual SPC system, a person records a small set of readings at regular intervals on a standard SPC form, along with the time and their name. A quick look at the chart for any trends is usually enough to check for potential problems. Advantages are that a manual system is flexible and requires no special equipment. Disadvantages are that it is less reliable than automated systems, more time consuming, and requires operator training.

In a semi-automated SPC system, information is entered directly into a computer, either by typing or by a gauge linked to the computer. This permits automatic alerting of potential problems, improved reliability, and easy access to the information by other people. Advantages include better analysis, reduced training, and easy access. A disadvantage could be the operator is required to use a computer.

In a fully-automated SPC system, computerized monitoring equipment is directly connected to the machine which records both process parameters and machine status. An advantage is minimal operator intervention. Disadvantages are that the system is expensive and difficult to implement.

SPC tools

The cornerstone of statistical process control is measurement. When a process goes unmonitored, unmeasured and uncorrected, very large and unpredictable swings in both quality and quantity often occur. The first step in measuring process control or stability is the use of a control chart, which is the tool most closely associated with SPC.

Control Charts were invented in 1924 by Walter A. Shewhart, who is known as the father of statistical quality control. The control chart represented an initial step toward what Shewhart called “the formulation of a scientific basis for securing economic control.” (”Walter A. Shewhart”, No date)

Control Charts plot observations of the same characteristics as they change over time. By examining the chart for non-random behavior, it can be seen that something is causing the process to act differently than before. The next step is to discover that cause of variation and find a way to eliminate it.

To employ control charts successfully, two things are necessary: good data collection and documentation of the process.

Points on a control chart represent subgroups. A subgroup consists of one or more samples of a characteristic. The mean or average of the samples is plotted on one chart. The spread (difference between the biggest and the smallest measure on a corresponding chart) is also plotted.

The top chart is called an X-bar chart. In mathematical terms, X-bar is the mean (average). The X-bar chart is used to examine the changes from one subgroup to another. The bottom chart is called an R chart. R is an abbreviation for range (the difference between the largest and smallest measurement in each subgroup). The range chart examines the variation in samples of a subgroup.

Control charts include control limits calculated from collected data. When plotted points fall outside of these control limits, it indicates that the process has changed.

There are three types of patterns that may appear on an SPC chart. A run is a repeating pattern of plot points on one side of the control chart s centerline. A cycle is a repeating pattern of plot points that appear to be cyclical or time-dependent. A trend consists of six or more plot points moving consecutively in an upward or downward direction. Regardless of whether a chart shows a run, cycle, trend, or a combination of two or more of these, there has been a meaningful change in the process and the control limits may need to be recalculated.

X-bar and R charts are just one combination of a variety of control charts. Over the years, statisticians developed additional control charts that are more indicative of variation for different circumstances. X-bar and S (standard deviation) charts, individuals charts, moving range charts, moving average charts, moving range charts, exponentially weighted moving range charts, and others each have a proper application. (”Software”, 1997)

Histograms compare the distribution of measurements from an in-control process with the specification limits. Specification limits are the upper and lower limit values specified for a production process. If a data measurement falls outside the specification limits, the product is not being produced according to the product specifications. A benefit of using histograms is that statistically untrained workers can be taught to use a histogram and make good use of the results, without having to call on statisticians in the quality department.

Pareto charts display the relative frequency of quality-related attributes of a process. The frequencies are represented in bars that are ordered in decreasing magnitude. Pareto charts can be used as a first step to help prioritize where the efforts to solve a problem should be concentrated.

If the solution is not obvious, even after a Pareto analysis has narrowed the search, two additional techniques that can be helpful are brainstorming and fishbone diagrams. (Fine, 1997)

In a brainstorming session, members of a group propose possible causes of the problem, calling out ideas either randomly or each in turn. No censorship is allowed; and every idea, no matter how strange, is written down. One member of the group records the possible causes, which are then used as a springboard for additional ideas. After the group has run out of suggestions group members discuss and prioritize the list.

Fishbone diagrams (also called Cause and Effect or Ishikawa Diagrams) are another method for identifying problem causes. The fishbone takes an effect, symptom, or problem and traces its causes using four major categories: materials, manpower, methods, and machinery. (Fine, 1997)

Goal of SPC

In order to be effective, SPC should be understood as a planning process; not just a set of control charts.

Customer satisfaction is key to any company s success. One sure way to satisfy customers is to give them what they asked for. Although it sounds simple, it doesn t always happen. Every time a business upsets a customer, it risks

losing future earnings. Even if the problem is resolved, the probability of repeat sales to the customer has been reduced. True SPC demands that the process be characterized and controlled so that defective parts are prevented, not merely detected after they have been produced. Performing statistical process control on a finished part is too late. Gathering enough information about the process to learn what happens, when it happens, and before the problem occurs, is the goal of statistical process control. (Bohn, 1996)