You are on page 1of 56

PROCESS CONTROL SYSTEM MODEL WITH FEEDBACK

VOICE OF THE PROCESS

STATISTICAL STATISTICAL METHODS METHODS

PEOPLE EQUIPMENT MATERIAL METHODS MEASUREMENT ENVIRONMENT

THE WAY THE WAY WE WORK WE WORK BLENDING OF BLENDING OF RESOURCES RESOURCES

PRODUCTS OR SERVICES

CUSTOMERS

INPUTS

PROCESS / SYSTEM

OUTPUTS

I IDENTIFYING CHANGING NEEDS NEEDS AND EXPECTATIONS

VOICE OF THE CUSTOMER


In previous lessons we discussed the methodology of understanding the needs and expectations of the customer or voice of the customer and developing process to meet those needs. In this model, the use of statistical methods becomes the voice of the process and the methodology for controlling the process inputs to reduce variation and improve the process.

PROCESS CONTROL AND CAPABILITY


A process control system can be described as a feedback system. Statistical Process Control is one type of feedback system. There are other systems available that are not statistical in nature. Those systems will be discussed in lesson #9. Here we will focus on the statistical methods.

STATISTICAL METHODS FOR ANALYSIS, CONTROL AND PROCESS IMPROVEMENT


INTRODUCTION There are five points that should be made from the onset 1. Gathering data and using statistical methods to interpret them are not ends in themselves. The overall aim should be increased understanding of the process. It is very easy to become experts in the technique without ever realizing any improvements. 2. Measurement systems are critical to proper data analysis and they should be understood before process data are collected. When the measurement system has excessive error, the variation from the system may cause incorrect data to be collected and used and result in inappropriate decisions to be made. 3. The basic concept of studying variation and using statistical signals to improve performance can be applied to any area. Such areas can be on a job site, on the shop floor, or in the office. 4. SPC stands for Statistical Process Control. The focus of SPC is on the process, not the product. Only by focusing on the process can you improve quality, increase productivity, and reduce costs. 5. This should be considered the first step toward the use of statistical methods. It provides generally accepted approaches which may work in many instances. However, there are exceptions where these approaches may not be appropriate.

MEASURES OF CENTRAL TENDENCY AND SPREAD The measure of central tendency is the statistic used to determine the LOCATION of a distribution in other words, at what point are most of the data points centered around. There are three measures of central tendency. They are: Mean Median Mode

The mean is the average of all of the data points and is calculated by adding all of the data points to get the sum, and then divide the sum by the number of data points. The median is the center data point which has the same number of data points above and below it. You can liken it to the median on a highway with two lanes on one side and two lanes on the other side. To calculate the median, first put the data points in sequence from lowest to highest, and then select the data point in the middle that has the same number of data points above it and below it, regardless of what the values are. For an even number of data points, the median would be the value halfway between the two middle points.

The mode is the data point that appears the most frequently. To calculate, simply count which data point value appears the most. Work through the examples below. Calculate the mean, median and mode for each of the three sets of data. 1) 5 9 9 4 3 2) 4 7 6 4 4 3) 5 4 8

Mean = Median = Mode =

Mean = Median = Mode =

Mean = Median = Mode =

There are two measures of the spread of a distribution: Range Standard Deviation

The range is calculated by subtracting the lowest data point value from the highest data point value. The range tells how spread out the data points are from the lowest to the highest. Calculate the range for each of the above sets of data. 1)_________ 2)__________ 3)____________

The standard deviation is a certain calculated distance away from the average in a normal distribution. We will describe how to calculate the standard deviation when we get to the section on process capability.

GATHERING DATA You must first consider the reason or purpose for collecting the data. Is it proper data that is applicable to the application? Will you be able to make the proper evaluation with the data collected? Data for understanding Data for analysis Data for process control Regulating data Know what is happening in the process amount of variation Evaluation of cause and effect Determine normality or presence of special causes Used to set process parameters

Acceptance or rejection data Used for accepting/rejecting product 100% or sampling

Data must be correct and representative of the true value or facts. Data can be Variable: Measurement Data or Attribute Countable Data Data for Analysis: Should be organized for simple analysis Must include all pertinent facts about the part or process (time, date, machine/part number, etc.) CONTROL CHARTS TOOLS FOR PROCESS CONTROL, IMPROVEMENT AND CAPABILITY In his books, Dr. Deming identifies two mistakes frequently made in process control: 1. Ascribe a variation or mistake to a special cause, when in fact it belongs to the system (common cause) 2. Ascribe a variation or mistake to the system when in fact it is a special cause. Over adjusting or tampering is a common example of number one. Never doing anything to try to find a special cause is an example of number two. For effective management of variation during production, there must be an effective way to detect special causes of variation. Histograms can not be used for this purpose. Histograms are a visual representation of the process variation. They can be used to verify that the process variation follows a normal distribution Time-based statistical methods in the form of control charts do provide the necessary means to identify special causes of variation in a process. The control chart was developed by Dr. Shewhart in the 1920s to make the distinction between special causes and common causes of variation during production. Since then, control charts have been widely used in a variety of process control and improvement situations. They have proven to be very effective in distinguishing the special causes of variation when they occur and in reflecting the systemic common causes of the process. When no special causes of process variation are active in the process, it is said to be in statistical control, or simply put, in control. A process in control is stable, consistent and predictable. If special causes are present in the process it is out of control, unstable and unpredictable. With the process distribution being normal, special causes can be detected by significant changes in the process location (mean) or the process spread (range or standard deviation) To see if there has been a change in the process location, one could measure every part and calculate the process average. However, this is very costly. So the most efficient way is to take a sample of the process and calculate the mean of the sample. Then compare the sample mean to the process distribution using a set of statistically calculated control limits that have been set up on a control chart. If the sample mean is outside of the control limits, then it is an indication that a special cause of variation may be present. Also, if the samples inside of the control limits display a non-random pattern, then this is also an indicator of special causes.

In general, a control chart is set up by calculating the: Centerline = the process average of the characteristic being measured UCL = the upper control limit and LCL = the lower control limit as follows

UCL_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _

_ Centerline_____________________________________________________________ X

LCL _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _

Control charts are very useful tools during all phases of the Improvement Cycle and to aid in bringing the process into statistical control. Using the control chart, data can be compared with the control limits to determine if the variation in the process is normal, from common causes, or not normal, from special causes. If special causes of variation are present they must be identified, resolved and permanently corrected. Once they are removed, the control chart continues to be a tool to monitor the process and identify excessive variation due to common causes that are systemic in the process. These types of causes can only be fixed by management actions. Deming points out that special causes of variation account for about 15% of the problems, while common causes account for 85%. While special causes are being identified and resolved, the control limits need not be recalculated. However, as common causes are resolved and the process improves by reducing the amount of variation, then the control limits on the chart need to be recalculated to more accurately reflect the process performance. Properly used, control charts can: Be used by operators for ongoing control of the process Help the process perform consistently and predictably Allow the process to achieve higher quality, reduced costs and higher capability Provide a common language for discussing process performance Distinguish special from common causes as a guide of the proper action to take

For SPC to be effective management must: Focus the organization on variation reduction Support cross-functional team work to resolve special and common causes Support SPC training for all levels including engineering, management and operators Ensure the optimum use of SPC data and information Assure proper placement of SPC charts for optimum use by the employees

CONSTRUCTING CONTROL CHARTS FOR VARIABLES Variable data are always preferred to attribute data because they give you more useful information than attribute data for the same amount of effort. Variable control charts generally come in pairs because you want to look at the two statistical measurements of a distribution the location and the spread. Therefore the pair of control charts would be an x-bar chart for location and range chart for the spread. You can also do a median and range chart or an x-bar and standard deviation chart. The chart we will use is the X-bar and R chart. A copy of a completed chart is shown on the end of this explanation.

PREPARATORY STEPS 1. Decide what characteristic is to be measured. The characteristic should be representative of the process you want to evaluate, control, and improve. 2. Define the measurement system to be used, including the gage to be used and method for performing the measurement. Ensure that the gage is calibrated and the measurement system is capable.

3. Create a sampling plan. Define the sample size, frequency and sequence of samples taken. The sequence should generally be consecutive parts being checked. For example, if the sample size is five pieces and the frequency is every two hours, then the five pieces should be taken all at one time, five pieces in a row not randomly over the two hour period. This will reduce the piece-topiece variation within the sample and give a more representative picture of what the process is doing. 4. Once these steps are completed, you can begin to gather the data and record it on a blank chart in the space provided at the bottom of the chart.

STEPS FOR COMPLETING THE CHART 1. Complete the information for the header at the top of the chart. This is important information to know the who, what, where, when and how. This is a must. 2. Gather the measurements for each sample of five and record it in the space provided at the bottom of the chart. Each of these samples of five is also known as a sub group. 3. Calculate the average and range for each subgroup (sample of five) that you recorded. 4. Calculate the process average by calculating the average of all of the subgroup averages. This is = the X or (X double bar). __ 5. Calculate the average range from all of the subgroup ranges. This is the R. or (R bar) 6. Based on the statistics from the subgroups, establish the vertical scale for the X-bar chart and the R chart. The scale for the X-bar chart should be set up so the process average is in the middle of the chart. The range chart will always start at zero. 7. Calculate the upper and lower control limit for the X-bar chart and the upper control limit for the range chart. The lower control limit on the range chart will be zero unless your sample size is greater than six.

8. Draw in the control limits for each chart. 9. Draw in the average range on the range chart and the process average on the X-bar chart. 10. Plot the average and range data values for each subgroup. 11. Analyze the charts for any out of control conditions.

FORMULAS NEEDED FOR CALCULATING CONTROL LIMITS X-BAR CHART = __ UCL = X (process average) + A2 x R = __ LCL = X (process average) A2 x R RANGE CHART __ UCL = D4 x R __ LCL = D3 x R

A2, D3 and D4 are statistical factors that are based on the sub group sample size. These were developed by the statistical experts for the ease in calculating the control limits. They save a lot of tedious number crunching.

FACTORS FOR CONTROL LIMITS Sample Size Factor for X-Bar Chart A2 1.880 1.023 0.729 0.577 0.483 0.419 Factors for Range Chart D3 0 0 0 0 0 0.076 Factors For Range chart D4 3.267 2.574 2.282 2.114 2.004 1.924 Factors For Sigma Hat D2 1.128 1.693 2.059 2.326 2.534 2.704

2 3 4 5 6 7

ANALYZING CHARTS FOR OUT OF CONTROL CONDITIONS A process is in control when all plotted points are within the control limits and none of the following out of control conditions are present. An out of control condition is an indicator that a special cause of variation is present. Although there are times when you will get a false indicator, it is important that all out of control conditions be evaluated to determine if special causes exist. The following out of control conditions apply to both the X-bar chart and R chart.

You should always analyze the chart pattern each time you plot a data point. If the pattern shows an out of control condition as shown below, then action to resolve it must be taken immediately to prevent any possible defects from being produced. 1. 2. 3. 4. 5. 6. A single plotted point above the upper control limit or below the lower control limit A run of more than 7 plotted points in a row, either above or below the average line A trend of more than 7 plotted points in a row heading up or down. Cycles Saw tooth pattern Several plotted points close to the control limits.

CONSTRUCTING CONTROL CHARTS FOR ATTRIBUTES Attribute charts serve the same purpose for attribute data as X-bar and R charts do for variable type data. They are designed to track unacceptable parts and identify the presence of special causes of variation in a process that is measured by taking only attribute type data. However, unlike the variables chart, these charts are not based on the normal distribution. They are based on a binomial or poisson distribution that is statistically normalized. Therefore the calculation of the control limits is somewhat different. There are two types of charts: 1. Charts that track the number of defective parts found in the sample 2. Charts that track the number of defects in a single part. The first types of charts are called P-charts or NP-charts. The P chart is used for sample sizes that may vary slightly. (not more that 20%) The NP chart is used when the sample size is the same every time. The second types of charts are called C-charts or U charts. The C chart is used for the numbers of defects found in a single unit, such as a radio. The U chart is used for the number of defects found in a sample from a continuous flow process, such as flaws in woven fabric or rolls of steel. We will go through the steps to construct a P chart, since this is the most frequently used chart. Except for the formulas to calculate the control limits, the steps are the same for the other charts.

The preparation steps are the same as for the Variables chart, except that the measurements will be attribute type checks. 1. Complete the information that goes in the header. Very important! 2. Do the checks and collect and record the data in the appropriate spot. Record the sample size and the number of defective parts found 3. Calculate the percent (proportion) defective for each sub group and record it. P = (Percent defective) = number of defective parts X 100 Sample Size 4. Record the date and time. 5. Calculate the average percent defective and record it on the chart. __ P = total number of defectives found Total number of pieces inspected 6. Calculate the control limits using the formulas on the following page. 7. Based on the recorded percent defectives, determine the vertical scale and record it on the chart. 8. Draw in the average percent defective line and control limits. 9. Plot the percent (proportion) defective for each sample taken on the chart. 10. Analyze the plotted points for out control conditions.

FORMULAS FOR CONTROL LIMITS FOR ATTRIBUTE CHARTS For the P-Chart _ UCL = P +3 x _ _ P(1-P) n _ LCL = P 3 x _ _ P(1-P) n

The formulas for all of the other charts can be found on the following pages. Chart analysis is the same process used for the variables X-bar and R chart, except trends toward a lower percent defective are indicators of process improvement and those special causes need to be identified so they can be made a permanent part of the process.

DETERMINING PROCESS CAPABILITY Process capability is the ability of the process to consistently produce parts the meet the blue print specifications. The rules of the normal distribution apply to this method of determining process capability. Process capability is calculated using the process average and the standard deviation. Another word for standard deviation is sigma Before you can begin to calculate process capability you must verify that the process is in statistical control. This is evident from the lack of out of control conditions (special causes of variation) on the control chart. Two formulas are used for calculating the standard deviation or sigma.

First, if you are not using a control chart and want to calculate the sigma for a certain population from which you took the sample: _ _2 S= (X X) N1 The best thing to do here is to calculate it using a statistical calculator. When you are using an X-bar and R chart the formula for sigma hat (the hat means it is estimated) is: _ sigma hat = R_ _ d2 = The R is taken from the range chart and the process average X is taken from the X-bar chart. d2 is another statistical factor for simplifying our calculations. (see the table of factors above)

By applying the process average and sigma to the normal distribution, you can predict the output of the process. Understanding how the sigma and process average work with the normal distribution: We apply it to the product specification to determine the process capability using the following formula. The symbol used to signify process capability is Cp Cp = USL LSL 6 sigma hat This simply tells us if the spread of the process distribution at plus 3 sigma and minus 3 sigma from the process average will fit within the tolerance spread of the specification, or if it is wider than the tolerance spread. If the spread of the process fits within the spread of the tolerance, then the process is capable of producing parts that consistently meet the specification 99.7% of the time. The question now is where is the process located? This is answered by taking the process average into account in calculating process capability. This is the k factor in Cpk The symbol for this is Cpk What happens here is you take the process average in relation to the upper specification limit and lowqer specification limit and use only 3 sigma instead of the six sigma. This tells you if the process is located within the specification limits where you want it to be. = = Cpk upper = USL X OR Cpk lower = X LS 3 sigma hat 3 sigma hat (Which ever number is smallest of the two is the true Cpk value) = NOTE: The text book uses the Greek letter mu (u) in place of the X. These are both the same so you can use either one. So if the Cp = 1.00 or greater, the process is said to be capable because the 6 sigma spread is equal to the tolerance spread. If the Cp =1.33, then the spread of the process at 8 sigma (+/- 4sigma ) is within the tolerance spread, making the process more capable. Therefore, the larger the Cp number the narrower the process spread is in relation to the tolerance spread. This means there is less variation in the process than the tolerance allows for.

This principle applies to the Cpk number as well. When the Cpk is 1.00 or higher, we know for sure that the process spread is within the spread of the specification tolerance because we know the center of the process is located within the specification tolerance. When evaluating the process capability, it is best to use both the Cp and Cpk number. If you have a high Cp number but a low Cpk number, then it tells you that the process is capable but is not located within the specification tolerance. Therefore adjustments need to be made to the process to move the process average to the center of the specification. NOTE: The Cpk number will ALWAYS be equal to or less than the Cp number. If this is not so, then there is a mistake in the calculation somewhere.

The following chart shows the relationship of the Cpk to the yield of good parts from the process and the number of defective parts per million (DPPM) the process may be producing. This gives meaning to the Cpk value. Once you calculate the Cpk index you will be able to tell what your predicted yield of good parts is. Remember that these yield numbers and DPPM numbers are based on the assumption that the distribution from your process is a normal distribution and that the process is in statistical control with no spectial causes of variation present.

Relationship of Cpk to Expected Process Yield and Defects Per Million (DPM) Cpk +/sigma spread 1 2 3 4 5 6 Yield % of good parts 68.2689 95.4499 97.7300 99.9936 99.999942 99.999999 Number of Defects 317,310 45,500 2,699 63 574 2 Per

0.33 0.67 1.00 1.33 1.67 2.0

Million Million Million Million Billion Billion

For attribute data simply multiply the P bar times one million to determine the number of defective parts per million. This chart is based on the assumption that the control charts show that the process is in statistical control and that the output / distribution of the process is a Normal Distribution.

10

PROCESS VARIATION COMMON AND SPECIAL CAUSES


In order to use process control methods and measurement data, it is important to understand the concept of variation. The process control system model shows six inputs to the process and any one or all of them are sources of variation in the process. No two products or characteristics coming from the same process are exactly alike, because any process contains many sources of variability. The difference between products may be relatively large or immeasurably small, but they are always present. The diameter of a machined shaft, for instance, would be susceptible to variation from the machine (clearances, bearing wear), tool (rate of wear, strength), material (diameter, hardness), operator (part feed, accuracy of centering), maintenance (lubrication, replacement of worn parts), environment (temperature, consistency of power supply), and the measurement system accuracy. Another example is that the time required to process an invoice could vary according to the people performing the various steps, the reliability of any equipment they are using, the accuracy and legibility of the invoice itself, the procedures followed, and the volume of other work in the office. Some sources of variation in the process can cause short-term, piece-to-piece differences, such as backlash and clearance within a machine and its fixturing, or the accuracy of a bookkeepers work. Other sources of variation tend to cause changes in the output only over a longer period of time. These changes may occur either gradually as with tool or machine wear, stepwise as with procedural changes, or irregularly as with environmental changes such as power surges. Therefore the time period and conditions over which measurements are made are critical, since they will affect the amount of total variation that will be observed. While the individual measured values may all be different, as a group they form a pattern that can be described as a distribution. This distribution can be characterized by: Location (typical or central value) Spread (span or width of values from smallest to largest) Shape (the pattern of variation whether it is symmetrical, skewed, etc.)

Knowing that there is a certain degree of part-to-part variation, the design engineers establish tolerances or specifications that the parts must meet. Thus, from the standpoint of minimum requirements, the issue of variation is often simplified: parts within the specification tolerances are acceptable, parts beyond specification tolerances are not acceptable; reports on time are acceptable, late reports are not acceptable. However, the goal should be to maintain the location to a target value with minimal variability. To manage any process and reduce variation, the variation must be traced back to its sources. The first step is to make a distinction between common and special causes of variation. Common causes refer to the many sources of variation that consistently act on the process. Common causes within a process produce a stable and repeatable distribution over time. This is called in a state of statistical control, in statistical control, or sometimes just in control. Common causes yield a stable system of chance causes. If only common causes of variation are present and do not change, the output of a process is predictable. In most processes common causes of variation are inherent in the system and beyond the control of the operator. Special causes (often called assignable causes) refer to any factors causing variation that affect only some of the process output. They are often intermittent and unpredictable. Special causes are signaled by one or more points beyond the control limits or by non-random patterns of points within the control limits of a process control chart. (These charts will be covered in detail in lesson seven.) Unless all of the special causes of variation are identified and acted upon, they may continue to affect the process output

in unpredictable ways. If special causes of variation are present, the process out put will not be stable over time and may result in defective or unacceptable product. The changes in the process distribution due to special causes can be either detrimental or beneficial. When detrimental, they must be understood and permanently removed. When beneficial, they should be understood and made a permanent part of the process.

PROCESS CONTROL AND CAPABILITY


A process control system can be described as a feedback system. Statistical Process Control is one type of feedback system. There are other systems available that are not statistical in nature. Those systems will be discussed in lesson #9. Here we will focus on the statistical methods.

STATISTICAL METHODS FOR ANALYSIS, CONTROL AND PROCESS IMPROVEMENT


INTRODUCTION There are five points that should be made from the onset 1. Gathering data and using statistical methods to interpret them are not ends in themselves. The overall aim should be increased understanding of the process. It is very easy to become experts in the technique without ever realizing any improvements. 2. Measurement systems are critical to proper data analysis and they should be understood before process data are collected. When the measurement system has excessive error, the variation from the system may cause incorrect data to be collected and used and result in inappropriate decisions to be made. 3. The basic concept of studying variation and using statistical signals to improve performance can be applied to any area. Such areas can be on a job site, on the shop floor, or in the office. 4. SPC stands for Statistical Process Control. The focus of SPC is on the process, not the product. Only by focusing on the process can you improve quality, increase productivity, and reduce costs. 5. This should be considered the first step toward the use of statistical methods. It provides generally accepted approaches which may work in many instances. However, there are exceptions where these approaches may not be appropriate.

MEASURES OF CENTRAL TENDENCY AND SPREAD The measure of central tendency is the statistic used to determine the LOCATION of a distribution in other words, at what point are most of the data points centered around. There are three measures of central tendency. They are: Mean Median Mode

The mean is the average of all of the data points and is calculated by adding all of the data points to get the sum, and then divide the sum by the number of data points. The median is the center data point which has the same number of data points above and below it. You can liken it to the median on a highway with two lanes on one side and two lanes on the other side. To calculate the median, first put the data points in sequence from lowest to highest, and then select the data point in the middle that has the same number of data points above it and below it, regardless of what the values are. For an even number of data points, the median would be the value halfway between the two middle points.

The mode is the data point that appears the most frequently. To calculate, simply count which data point value appears the most. Work through the examples below. Calculate the mean, median and mode for each of the three sets of data. 1) 5 9 9 4 3 2) 4 7 6 4 4 3) 5 4 8

Mean = Median = Mode =

Mean = Median = Mode =

Mean = Median = Mode =

There are two measures of the spread of a distribution: Range Standard Deviation

The range is calculated by subtracting the lowest data point value from the highest data point value. The range tells how spread out the data points are from the lowest to the highest. Calculate the range for each of the above sets of data. 1)_________ 2)__________ 3)____________

The standard deviation is a certain calculated distance away from the average in a normal distribution. We will describe how to calculate the standard deviation when we get to the section on process capability.

GATHERING DATA You must first consider the reason or purpose for collecting the data. Is it proper data that is applicable to the application? Will you be able to make the proper evaluation with the data collected? Data for understanding Data for analysis Data for process control Regulating data Know what is happening in the process amount of variation Evaluation of cause and effect Determine normality or presence of special causes Used to set process parameters

Acceptance or rejection data Used for accepting/rejecting product 100% or sampling

Data must be correct and representative of the true value or facts. Data can be Variable: Measurement Data or Attribute Countable Data Data for Analysis: Should be organized for simple analysis Must include all pertinent facts about the part or process (time, date, machine/part number, etc.) CONTROL CHARTS TOOLS FOR PROCESS CONTROL, IMPROVEMENT AND CAPABILITY In his books, Dr. Deming identifies two mistakes frequently made in process control: 1. Ascribe a variation or mistake to a special cause, when in fact it belongs to the system (common cause) 2. Ascribe a variation or mistake to the system when in fact it is a special cause. Over adjusting or tampering is a common example of number one. Never doing anything to try to find a special cause is an example of number two. For effective management of variation during production, there must be an effective way to detect special causes of variation. Histograms can not be used for this purpose. Histograms are a visual representation of the process variation. They can be used to verify that the process variation follows a normal distribution Time-based statistical methods in the form of control charts do provide the necessary means to identify special causes of variation in a process. The control chart was developed by Dr. Shewhart in the 1920s to make the distinction between special causes and common causes of variation during production. Since then, control charts have been widely used in a variety of process control and improvement situations. They have proven to be very effective in distinguishing the special causes of variation when they occur and in reflecting the systemic common causes of the process. When no special causes of process variation are active in the process, it is said to be in statistical control, or simply put, in control. A process in control is stable, consistent and predictable. If special causes are present in the process it is out of control, unstable and unpredictable. With the process distribution being normal, special causes can be detected by significant changes in the process location (mean) or the process spread (range or standard deviation) To see if there has been a change in the process location, one could measure every part and calculate the process average. However, this is very costly. So the most efficient way is to take a sample of the process and calculate the mean of the sample. Then compare the sample mean to the process distribution using a set of statistically calculated control limits that have been set up on a control chart. If the sample mean is outside of the control limits, then it is an indication that a special cause of variation may be present. Also, if the samples inside of the control limits display a non-random pattern, then this is also an indicator of special causes.

In general, a control chart is set up by calculating the: Centerline = the process average of the characteristic being measured UCL = the upper control limit and LCL = the lower control limit as follows

UCL_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _

_ Centerline_____________________________________________________________ X

LCL _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _

Control charts are very useful tools during all phases of the Improvement Cycle and to aid in bringing the process into statistical control. Using the control chart, data can be compared with the control limits to determine if the variation in the process is normal, from common causes, or not normal, from special causes. If special causes of variation are present they must be identified, resolved and permanently corrected. Once they are removed, the control chart continues to be a tool to monitor the process and identify excessive variation due to common causes that are systemic in the process. These types of causes can only be fixed by management actions. Deming points out that special causes of variation account for about 15% of the problems, while common causes account for 85%. While special causes are being identified and resolved, the control limits need not be recalculated. However, as common causes are resolved and the process improves by reducing the amount of variation, then the control limits on the chart need to be recalculated to more accurately reflect the process performance. Properly used, control charts can: Be used by operators for ongoing control of the process Help the process perform consistently and predictably Allow the process to achieve higher quality, reduced costs and higher capability Provide a common language for discussing process performance Distinguish special from common causes as a guide of the proper action to take

For SPC to be effective management must: Focus the organization on variation reduction Support cross-functional team work to resolve special and common causes Support SPC training for all levels including engineering, management and operators Ensure the optimum use of SPC data and information Assure proper placement of SPC charts for optimum use by the employees

CONSTRUCTING CONTROL CHARTS FOR VARIABLES Variable data are always preferred to attribute data because they give you more useful information than attribute data for the same amount of effort. Variable control charts generally come in pairs because you want to look at the two statistical measurements of a distribution the location and the spread. Therefore the pair of control charts would be an x-bar chart for location and range chart for the spread. You can also do a median and range chart or an x-bar and standard deviation chart. The chart we will use is the X-bar and R chart. A copy of a completed chart is shown on the end of this explanation.

PREPARATORY STEPS 1. Decide what characteristic is to be measured. The characteristic should be representative of the process you want to evaluate, control, and improve. 2. Define the measurement system to be used, including the gage to be used and method for performing the measurement. Ensure that the gage is calibrated and the measurement system is capable.

3. Create a sampling plan. Define the sample size, frequency and sequence of samples taken. The sequence should generally be consecutive parts being checked. For example, if the sample size is five pieces and the frequency is every two hours, then the five pieces should be taken all at one time, five pieces in a row not randomly over the two hour period. This will reduce the piece-topiece variation within the sample and give a more representative picture of what the process is doing. 4. Once these steps are completed, you can begin to gather the data and record it on a blank chart in the space provided at the bottom of the chart.

STEPS FOR COMPLETING THE CHART 1. Complete the information for the header at the top of the chart. This is important information to know the who, what, where, when and how. This is a must. 2. Gather the measurements for each sample of five and record it in the space provided at the bottom of the chart. Each of these samples of five is also known as a sub group. 3. Calculate the average and range for each subgroup (sample of five) that you recorded. 4. Calculate the process average by calculating the average of all of the subgroup averages. This is = the X or (X double bar). __ 5. Calculate the average range from all of the subgroup ranges. This is the R. or (R bar) 6. Based on the statistics from the subgroups, establish the vertical scale for the X-bar chart and the R chart. The scale for the X-bar chart should be set up so the process average is in the middle of the chart. The range chart will always start at zero. 7. Calculate the upper and lower control limit for the X-bar chart and the upper control limit for the range chart. The lower control limit on the range chart will be zero unless your sample size is greater than six.

8. Draw in the control limits for each chart. 9. Draw in the average range on the range chart and the process average on the X-bar chart. 10. Plot the average and range data values for each subgroup. 11. Analyze the charts for any out of control conditions.

FORMULAS NEEDED FOR CALCULATING CONTROL LIMITS X-BAR CHART = __ UCL = X (process average) + A2 x R = __ LCL = X (process average) A2 x R RANGE CHART __ UCL = D4 x R __ LCL = D3 x R

A2, D3 and D4 are statistical factors that are based on the sub group sample size. These were developed by the statistical experts for the ease in calculating the control limits. They save a lot of tedious number crunching.

FACTORS FOR CONTROL LIMITS Sample Size Factor for X-Bar Chart A2 1.880 1.023 0.729 0.577 0.483 0.419 Factors for Range Chart D3 0 0 0 0 0 0.076 Factors For Range chart D4 3.267 2.574 2.282 2.114 2.004 1.924 Factors For Sigma Hat D2 1.128 1.693 2.059 2.326 2.534 2.704

2 3 4 5 6 7

ANALYZING CHARTS FOR OUT OF CONTROL CONDITIONS A process is in control when all plotted points are within the control limits and none of the following out of control conditions are present. An out of control condition is an indicator that a special cause of variation is present. Although there are times when you will get a false indicator, it is important that all out of control conditions be evaluated to determine if special causes exist. The following out of control conditions apply to both the X-bar chart and R chart.

You should always analyze the chart pattern each time you plot a data point. If the pattern shows an out of control condition as shown below, then action to resolve it must be taken immediately to prevent any possible defects from being produced. 1. 2. 3. 4. 5. 6. A single plotted point above the upper control limit or below the lower control limit A run of more than 7 plotted points in a row, either above or below the average line A trend of more than 7 plotted points in a row heading up or down. Cycles Saw tooth pattern Several plotted points close to the control limits.

CONSTRUCTING CONTROL CHARTS FOR ATTRIBUTES Attribute charts serve the same purpose for attribute data as X-bar and R charts do for variable type data. They are designed to track unacceptable parts and identify the presence of special causes of variation in a process that is measured by taking only attribute type data. However, unlike the variables chart, these charts are not based on the normal distribution. They are based on a binomial or poisson distribution that is statistically normalized. Therefore the calculation of the control limits is somewhat different. There are two types of charts: 1. Charts that track the number of defective parts found in the sample 2. Charts that track the number of defects in a single part. The first types of charts are called P-charts or NP-charts. The P chart is used for sample sizes that may vary slightly. (not more that 20%) The NP chart is used when the sample size is the same every time. The second types of charts are called C-charts or U charts. The C chart is used for the numbers of defects found in a single unit, such as a radio. The U chart is used for the number of defects found in a sample from a continuous flow process, such as flaws in woven fabric or rolls of steel. We will go through the steps to construct a P chart, since this is the most frequently used chart. Except for the formulas to calculate the control limits, the steps are the same for the other charts.

The preparation steps are the same as for the Variables chart, except that the measurements will be attribute type checks. 1. Complete the information that goes in the header. Very important! 2. Do the checks and collect and record the data in the appropriate spot. Record the sample size and the number of defective parts found 3. Calculate the percent (proportion) defective for each sub group and record it. P = (Percent defective) = number of defective parts X 100 Sample Size 4. Record the date and time. 5. Calculate the average percent defective and record it on the chart. __ P = total number of defectives found Total number of pieces inspected 6. Calculate the control limits using the formulas on the following page. 7. Based on the recorded percent defectives, determine the vertical scale and record it on the chart. 8. Draw in the average percent defective line and control limits. 9. Plot the percent (proportion) defective for each sample taken on the chart. 10. Analyze the plotted points for out control conditions.

FORMULAS FOR CONTROL LIMITS FOR ATTRIBUTE CHARTS For the P-Chart _ UCL = P +3 x _ _ P(1-P) n _ LCL = P 3 x _ _ P(1-P) n

The formulas for all of the other charts can be found on the following pages. Chart analysis is the same process used for the variables X-bar and R chart, except trends toward a lower percent defective are indicators of process improvement and those special causes need to be identified so they can be made a permanent part of the process.

DETERMINING PROCESS CAPABILITY Process capability is the ability of the process to consistently produce parts the meet the blue print specifications. The rules of the normal distribution apply to this method of determining process capability. Process capability is calculated using the process average and the standard deviation. Another word for standard deviation is sigma Before you can begin to calculate process capability you must verify that the process is in statistical control. This is evident from the lack of out of control conditions (special causes of variation) on the control chart. Two formulas are used for calculating the standard deviation or sigma.

First, if you are not using a control chart and want to calculate the sigma for a certain population from which you took the sample: _ _2 S= (X X) N1 The best thing to do here is to calculate it using a statistical calculator. When you are using an X-bar and R chart the formula for sigma hat (the hat means it is estimated) is: _ sigma hat = R_ _ d2 = The R is taken from the range chart and the process average X is taken from the X-bar chart. d2 is another statistical factor for simplifying our calculations. (see the table of factors above)

By applying the process average and sigma to the normal distribution, you can predict the output of the process. Understanding how the sigma and process average work with the normal distribution: We apply it to the product specification to determine the process capability using the following formula. The symbol used to signify process capability is Cp Cp = USL LSL 6 sigma hat This simply tells us if the spread of the process distribution at plus 3 sigma and minus 3 sigma from the process average will fit within the tolerance spread of the specification, or if it is wider than the tolerance spread. If the spread of the process fits within the spread of the tolerance, then the process is capable of producing parts that consistently meet the specification 99.7% of the time. The question now is where is the process located? This is answered by taking the process average into account in calculating process capability. This is the k factor in Cpk The symbol for this is Cpk What happens here is you take the process average in relation to the upper specification limit and lowqer specification limit and use only 3 sigma instead of the six sigma. This tells you if the process is located within the specification limits where you want it to be. = = Cpk upper = USL X OR Cpk lower = X LS 3 sigma hat 3 sigma hat (Which ever number is smallest of the two is the true Cpk value) = NOTE: The text book uses the Greek letter mu (u) in place of the X. These are both the same so you can use either one. So if the Cp = 1.00 or greater, the process is said to be capable because the 6 sigma spread is equal to the tolerance spread. If the Cp =1.33, then the spread of the process at 8 sigma (+/- 4sigma ) is within the tolerance spread, making the process more capable. Therefore, the larger the Cp number the narrower the process spread is in relation to the tolerance spread. This means there is less variation in the process than the tolerance allows for.

This principle applies to the Cpk number as well. When the Cpk is 1.00 or higher, we know for sure that the process spread is within the spread of the specification tolerance because we know the center of the process is located within the specification tolerance. When evaluating the process capability, it is best to use both the Cp and Cpk number. If you have a high Cp number but a low Cpk number, then it tells you that the process is capable but is not located within the specification tolerance. Therefore adjustments need to be made to the process to move the process average to the center of the specification. NOTE: The Cpk number will ALWAYS be equal to or less than the Cp number. If this is not so, then there is a mistake in the calculation somewhere.

The following chart shows the relationship of the Cpk to the yield of good parts from the process and the number of defective parts per million (DPPM) the process may be producing. This gives meaning to the Cpk value. Once you calculate the Cpk index you will be able to tell what your predicted yield of good parts is. Remember that these yield numbers and DPPM numbers are based on the assumption that the distribution from your process is a normal distribution and that the process is in statistical control with no spectial causes of variation present.

Relationship of Cpk to Expected Process Yield and Defects Per Million (DPM) Cpk +/sigma spread 1 2 3 4 5 6 Yield % of good parts 68.2689 95.4499 97.7300 99.9936 99.999942 99.999999 Number of Defects 317,310 45,500 2,699 63 574 2 Per

0.33 0.67 1.00 1.33 1.67 2.0

Million Million Million Million Billion Billion

For attribute data simply multiply the P bar times one million to determine the number of defective parts per million. This chart is based on the assumption that the control charts show that the process is in statistical control and that the output / distribution of the process is a Normal Distribution.

PROCESS VARIATION COMMON AND SPECIAL CAUSES


In order to use process control methods and measurement data, it is important to understand the concept of variation. The process control system model shows six inputs to the process and any one or all of them are sources of variation in the process. No two products or characteristics coming from the same process are exactly alike, because any process contains many sources of variability. The difference between products may be relatively large or immeasurably small, but they are always present. The diameter of a machined shaft, for instance, would be susceptible to variation from the machine (clearances, bearing wear), tool (rate of wear, strength), material (diameter, hardness), operator (part feed, accuracy of centering), maintenance (lubrication, replacement of worn parts), environment (temperature, consistency of power supply), and the measurement system accuracy. Another example is that the time required to process an invoice could vary according to the people performing the various steps, the reliability of any equipment they are using, the accuracy and legibility of the invoice itself, the procedures followed, and the volume of other work in the office. Some sources of variation in the process can cause short-term, piece-to-piece differences, such as backlash and clearance within a machine and its fixturing, or the accuracy of a bookkeepers work. Other sources of variation tend to cause changes in the output only over a longer period of time. These changes may occur either gradually as with tool or machine wear, stepwise as with procedural changes, or irregularly as with environmental changes such as power surges. Therefore the time period and conditions over which measurements are made are critical, since they will affect the amount of total variation that will be observed. While the individual measured values may all be different, as a group they form a pattern that can be described as a distribution. This distribution can be characterized by: Location (typical or central value) Spread (span or width of values from smallest to largest) Shape (the pattern of variation whether it is symmetrical, skewed, etc.)

Knowing that there is a certain degree of part-to-part variation, the design engineers establish tolerances or specifications that the parts must meet. Thus, from the standpoint of minimum requirements, the issue of variation is often simplified: parts within the specification tolerances are acceptable, parts beyond specification tolerances are not acceptable; reports on time are acceptable, late reports are not acceptable. However, the goal should be to maintain the location to a target value with minimal variability. To manage any process and reduce variation, the variation must be traced back to its sources. The first step is to make a distinction between common and special causes of variation. Common causes refer to the many sources of variation that consistently act on the process. Common causes within a process produce a stable and repeatable distribution over time. This is called in a state of statistical control, in statistical control, or sometimes just in control. Common causes yield a stable system of chance causes. If only common causes of variation are present and do not change, the output of a process is predictable. In most processes common causes of variation are inherent in the system and beyond the control of the operator. Special causes (often called assignable causes) refer to any factors causing variation that affect only some of the process output. They are often intermittent and unpredictable. Special causes are signaled by one or more points beyond the control limits or by non-random patterns of points within the control limits of a process control chart. (These charts will be covered in detail in lesson seven.) Unless all of the special causes of variation are identified and acted upon, they may continue to affect the process output

in unpredictable ways. If special causes of variation are present, the process out put will not be stable over time and may result in defective or unacceptable product. The changes in the process distribution due to special causes can be either detrimental or beneficial. When detrimental, they must be understood and permanently removed. When beneficial, they should be understood and made a permanent part of the process.

T O P S PROBLEM SOLVING PROCESS


BECOME AWARE OF THE PROBLEM

1. USE THE TEAM APPROACH

2. DESCRIBE THE PROBLEM ANALYZE EXISTING DATA SEPARATE THE PROBLEM

3. IMPLEMENT AND VERIFY INTERIM CONTAINMENT ACTIONS Verify effectiveness of the actions

4. DEFINE AND VERIFY THE ROOT CAUSE IDENTIFY POTENTIAL CAUSES Use C/E diagram, 5 why and other tools Preview and improve problem description Evaluate each potential cause by comparison to problem description

SELECT THE MOST LIKELY CAUSES TEST AND EVALUATE THE LIKELY CAUSES TO DETERMINE THE ROOT CAUSE ONCE ROOT CAUSE IS DETERMINED, IDENTIFY ALTERNATIVE SOLUTIONS

5. VERIFY CORRECTIVE ACTIONS Evaluate solutions for problem resolution Evaluate the degree of problem reduction or elimination

6. IMPLEMENT PERMANENT CORRECTIVE ACTIONS Establish an action plan Establish a contingency plan

7. PREVENT RECURRENCE

8. CONGRATULATE YOUR TEAM

TOPS TEAM ORIENTED PROBLEM SOLVING


This is a very simple disciplined problem solving model which can be used with or without putting a team together. Speaking from years of experience in business and industry, as a manager you will be spending most of your time solving problems. To be successful, you must have a good problem solving process that you can use for any situation that is effective and part of your problem solving mindset. This approach has been in use in industry throughout the world for the past 25 years and is accepted as the basis for any problem solving situation. While at Ford Motor Company, I taught this process to about 100 suppliers and have seen it in action. I can say that it is a very effective process for permanently solving any type of problem whether you use a team or not. The objective of this process is to put in place a permanent corrective action that will ensure the problem is fixed at the root cause (accomplished in steps 1-6) and to ensure that the problem does not recur (accomplished with step 7).

STEP 1 - USE THE TEAM APPROACH


OBJECTIVE: Establish a small group of people with the process/product knowledge, allocated time, authority and skill in required technical disciplines to solve the problem and implement permanent corrective actions. The group must have a designated champion. When should you use the team approach to problem solving? When you dont know what the root cause of the problem is; when you get only one chance to correct the problem; when the solution must be the right one and permanent.

Question: Answer:

If a team is not necessary, simply follow the other steps of problem identification through prevention to ensure the permanent and proper solution is implemented. The team should consist of no more than ten members and should be organized with a team champion, team leader, scribe etc. The organization can be flexible with team members leaving and joining in. The team champion is the person with the power and authority to allocate the necessary resources (people, time, and money) and implement the permanent corrective action to resolve the problem. When using a team, it is important that the team leader sets a schedule, makes assignments to the right team members and follows up with them. Decision making should be by consensus, in that although all team members may not agree, once the course of action is decided all team members will fully support it. Upon completion of the problem solving process, it is important that the team be given proper recognition throughout the organization and if possible by some type of reward.

STEP 2 DESCRIBE THE PROBLEM


OBJECTIVE: Specify the internal or external customer problem by identifying in quantifiable terms who, what, where, when, why, how and how many (5W and 2H) of the problem.

Problem definition clearly describes the problem in a way that has meaning to everyone involved. Deming called this Operational Definitions. Proper definition serves as the basis for problem solving. The problem solving process starts with a clear problem definition and is vital to arriving at the root cause. The process of describing the problem is begun by gathering the necessary data to answer the 5W and 2H. Once the

problem is clearly defined, you can continue gathering data and information for further evaluation and verification. Irrelevant data can be discarded, and the team is more likely to stay on track because they agree on the basic elements of the problem. At this stage of the process, do not get caught in the trap of speculating or making decisions about causes and corrective actions. That will come later in the root cause analysis stage. Some guidelines for problem description activities are: Find the deviation: What Should Be versus What Actually Is = the problem Gather relevant information using such tools as the process flow diagram, and the What is and What is not diagram. Describe in comparisons using the 5Ws and 2Hs Subdivide the problem into smaller sub problems, if possible. Write the definition using an operational definition that is interpreted the same way by all involved

PROBLEM DESCRIPTION PROCESS FLOW

FIND THE DEVIATION What should be - - - - - -What is

GATHER RELEVANT DATA Process flow charts 5Ws and 2Hs Stratification Factors using the What is and What is not diagram

OTHER STATISTICAL TOOLS TO USE IN THIS STEP OF THE PROCESS Check Sheets Pareto Chart Affinity Diagram Histogram Flow chart

WHAT IS AND IS NOT DIAGRAM PROBLEM STATEMENT - (What is wrong with what?)

Description of the problem

IS

IS NOT

Need to get more information

WHAT The object

Deviation

Where Seen on object

Seen geographically

WHEN First seen

When else seen

When seen in process Or life cycle

HOW BIG Number of objects Affected Number of problems

Size of the problem

STEP 3 IMPLEMENT AND VERIFY INTERIM CONTAINMENT ACTIONS


OBJECTIVE: Define and implement containment actions to isolate the problem from any internal or external customer until permanent corrective actions are available. Verify the effectiveness of the action.

The purpose of implementing interim corrective actions is to contain the problem and prevent it from further affecting the customer. This action is a temporary Band Aid fix until the permanent corrective action is implemented. This step gives you the time to properly develop the root cause and implement the permanent corrective action.

Interim actions often involve 100% inspection, sorting, rework, repair or additional hand work. These activities are time consuming and expensive and must be removed as soon as the permanent corrective action is put in place.

Some guidelines for this step are: Establish the decision making criteria to be used in choosing the indicators Verify that the interim actions you have chosen actually do work to the customers satisfaction. Develop an action plan to implement the interim actions. Note: All action plans must include who is responsible to do what and when is it to be done. Monitor results to verify the interim action is working. Continue with the root cause analysis.

IMPLEMENT AND VERIFY INTERIM CONTAINMENT ACTIONS FLOW CHART ESTABLISH CONTAINMENT CRITERIA AND ACTIONS

VERIFY CONTAINMENT IS EFFECTIVE (Customer is isolated)

DEVELOP IMPLEMENTATION PLAN AND IMPLEMENT ACTIONS

MONITOR INTERIM CONTAINMENT ACTIONS

OTHER STATISTICAL TOOLS TO USE IN THIS STEP OF THE PROCESS PDPC chart Interrelationship Diagraph Control Charts Hisogram

STEP 4 DEFINE AND VERIFY THE ROOT CAUSE(S)


OBJECTIVE: Identify all potential causes which could explain why the problem occurred. Isolate and verify the root cause by testing each potential cause against the problem description and test data. Identify alternative corrective actions to eliminate the root cause.

This stage of the process is done in two phases. First, determine what the root cause is and second, identify the best solution or corrective action. The basic steps in this stage are to: Identify the potential causes, Then from the potential causes select the most likely causes

Investigate the selected most likely causes to arrive at the real root cause(s), Next identify the alternative solutions and select the one that best fits the situation.

Some guidelines to use for identifying potential and most likely causes are: Brainstorm and use a cause and effect diagram to identify all potential causes Do a comparative analysis to help uncover new information and differences Examine each difference discovered for changes Evaluate timing of events and changes a time line Ask the question Why as many times as necessary Five Whys group process Check for variation in the process inputs Evaluate several potential causes to arrive at the most likely cause Investigate the most likely causes to determine the root cause. Consult with experts Use the decision making process to help determine root cause and alternative solutions How do you know when you have the root cause defined? When the cause you are investigating addresses all of the elements stated in the problem description. There may be more that one root cause. This must be supported by the data and information gathered NOT just a GUT FEELING or a HUNCH. This is the reason for having a good detailed problem description.

Question: Answer:

Once the root cause is determined, the second phase is to identify possible solutions that will permanently resolve the root cause. Evaluate the possible solutions considering costs, effectiveness and feasibility to determine the best possible solution. NOTE: DO NOT allow the interim containment actions to become a substitute for a permanent solution.

DEFINE AND VERIFY ROOT CAUSE (S) PROCESS FLOW CHART

IDENTIFY ALL POTENTIAL CAUSES Brainstorm, use affinity diagram Use the Cause and Effect (Fishbone) diagram Use an interrelationship diagraph

FROM THE POTENTIAL CAUSES, SELECT THE MOST LIKELY CAUSES

THROUGH ANALYSIS AND TESTING SELECT AND VERIFY THE ROOT CAUSE (S) USE THE 5 WHY PROCESS

IDENTIFY THE ALTERNATIVE SOLUTIONS USE A TREE DIAGRAM

OTHER STATISTICAL TOOL TO USE IN THIS STEP OF THE PROCESS Cause and Effect diagram Scatter Diagram Pareto Chart Affinity diagram Tree diagram Flow chart Interrelationship diagraph PDPC chart

FIVE WHYS GROUP PROCESS PROCESS: STEP 1: THINK about the focus [question or statement]. STEP 2: ASK yourself, WHY? and write down your answer. STEP 3: LOOK at your first answer and ASK yourself, WHY? WRITE down your second answer. STEP 4: LOOK at your second answer and ASK yourself, WHY? and write down your answer. CONTINUE asking yourself, WHY until you have answered the question FIVE TIMES.

QUESTION OR STATEMENT: WHY?

WHY? WHY? WHY? WHY?

STEP 5: VERIFY PERMANENT CORRECTIVE ACTIONS


OBJECTIVE: Through trial and testing programs, quantitatively confirm that the selected permanent corrective actions will resolve the problem for the customer and will not cause undesirable side effects. Define contingency actions as needed based on the risk assessment.

In this step conduct tests, experiments, etc. to ensure that the chosen corrective actions will really solve the problem. Be sure tests are not influenced by the interim containment actions. One way to verify the corrective action is by being able to turn the problem on without the corrective action in place and turn it off with the corrective action in place. If actions are implemented that fail to solve the problem at the root cause, time money and effort are wasted and the problem will recur. Conducting tests prior to implementation, when possible, is more cost effective than waiting to see if the corrective actions worked after they are implemented. However, followup verification is also required.

THE CORRECTIVE ACTION MUST ADDRESS THE ROOT CAUSE Steps to follow in this activity: Develop an action plan for conducting the verification Establish the verification criteria Choose the indicators and establish a base line Apply appropriate statistical data to assess the corrective action Monitor the corrective action using reports and ongoing data

OTHER STATISTICAL TOOLS TO USE IN THIS STEP OF THE PROCESS Scatter Diagram Histograms Check Sheets Control Charts Process Capability

STEP 6 IMPLEMENT PERMANENT CORRECTIVE ACTIONS


OBJECTIVE: Define and implement the best permanent corrective action. Choose ongoing controls to ensure the root cause is eliminated and the corrective actions remain permanently in place. Once in production, monitor the long tem effects and if necessary, implement contingency actions.

In this step take the BEST solution the one that you tested and verified during the process of determining the root cause and verification in step 5 and put this solution in effect. The key element to look for in this step is to be sure that the corrective action is fully implemented. THE CORRECTIVE ACTION MUST ADDRESS THE ROOT CAUSE. This must be done with the interim containment actions being removed. In this step you MUST develop an Implementation Action Plan that defines WHO does WHAT and WHEN is to be done. This plan should include actions for putting controls in place to ensure the corrective actions are maintained and are effective and for contingencies. Some guide lines to use for this step are: Conduct test of the proposed solution Anticipate problems that may occur and develop contingency plans Remove interim containment actions Use statistical methods to measure effectiveness over time. Update the process flow diagrams, and any other related documents Test the proposed solution as much as possible before implementing it. Use a decision making process to evaluate and balance the benefits against the risks.

Controls refer to processes used to make sure that the corrective action solutions are working the way they were expected to work and that the solutions continue working even after the solution was verified. It is important to monitor the effects of the corrective actions for a long period of time to ensure the root cause was eliminated.

IMPLEMENT PERMANENT CORRECTIVE ACTIONS FLOW CHART SELECT THE BEST CORRECTIVE ACTION

DEVELOP THE IMPLEMENTATION ACTION PLAN WHO does WHAT and WHEN is it to be done

REMOVE INTERIM CONTAINMENT ACTIONS

IMPLEMENT PERMANENT CORRECTIVE ACTION

MONITOR THE LONG TERM EFFECTS

OTHER STATISTICAL TOOLS TO USE IN THIS STEP OF THE PROCESS Priority Matrix Activity Network Process Flow Chart Action Plan

Most problem solving stops at this point. The corrective action is in place and is expected to work. However, Step 7 Preventive Actions, is the key step to stop the problem from happening again. So take one more step to permanently resolve the problem.

STEP 7: PREVENT RECURRENCE


OBJECTIVE: Modify the management systems, operating systems, practices and procedures that allowed the problem to happen in order to prevent recurrence of this problem and/or of similar problems.

Problems happen because the system or process allowed the problem to occur. It may be from excessive variation in the process, lack of maintenance, lack of training or any one of many different system deficiencies. This step in the process addresses the deficiencies in the management system that allowed the problem to happen in the first place. Prevention involves modifying any management systems, operating systems, practices, and procedures to prevent the problem from recurring. Recurrence of a problem is very costly in time, effort, money, resources and quality control. Preventive measures help anticipate possible problems and recurrence and ensure that problems are avoided before they happen. In this step it is important to analyze and understand the underlying cause and effect relationships. Often, what happens in problem solving is that once a permanent corrective action is in place and verified, it does not stay in place for the duration. Old ways may creep back into the process because they are easier or out of habit, people are moved, there is little or no follow-up, etc. Part of the prevention activity is to implement controls to ensure the corrective action remains in place and becomes part of the standard operating procedures. Some guide lines to use to accomplish this are: Establish controls such as SPC, charting, gauging, audits, checklists etc. to ensure the corrective actions remain in place Provide needed training for all involved Have contingency plans for emergencies Update related documents Document all changes Identify potential things that might go wrong and jeopardize the corrective action Examine the management system. Is it outdated, unclear, need to be revised, training needed, documentation needed, etc. List and evaluate high risk areas Standardize the corrective actions Coordinate and document required actions Identify methods to prevent similar problems from happening in related areas Change the system that allowed the problem to happen

OTHER STATISTICAL TOOLS TO USE IN THIS STEP OF THE PROCESS Interrelationship Diagraph Process Flow Chart Statistical Process Control Prioritization Diagram Tree Diagram PDPC diagram

STEP 8: CONGRATULATE YOUR TEAM


OBJECTIVE: Recognize the collective efforts of the team.

This is an important part of the process in that the team has put forth an extra effort with no pay increase. Recognition becomes the extra pay for the extra work done. Based on the resources of the company and the degree of the problem solved, some type of financial reward would be appropriate. However, if the resources are not available, then public recognition for the work done will suffice. Non monetary things such as an afternoon off or luncheon should be considered. Either way, the team should be publicly applauded by top management for the job they did. This step boosts the morale not only of the team, but also of all employees when they see that management cares and recognizes extra work. Doing this will ensure the success of future team projects.

SUMMARY
This eight step problem solving process is a disciplined approach to avoiding the Quick Fix to problems. Although it may appear to be a long, drawn-out process, it can be done very quickly. It is a problem solving mind set that every manager must have in place in their method of operation. This process is applicable to problem solving in any situation whether at work, home or wherever it happens. I strongly encourage you to make this process a part of your problem solving methodology. IT WORKS.

PROBLEM SOLVING
In 1968, J. P. Guilford stated To live is to have problems and to solve problems is to grow creatively. What recruiters and company managers expect from college graduates: Original and visionary thinking Analytical and problem solving skills Be prepared to step in and make decisions and solve problems not to learn how to do it Strategic thinking Adaptability ability to deal with ambiguity

1. CREATIVITY the ability to modify self imposed constraints 2. DIVERGENT THINKING brainstorming; quantity; no judgment; novel is good

3. CONVERGENT THINKING select and evaluate the options; use judgment; check objectives; be deliberate

PROBLEM SOLVING PITFALLS 1. 2. 3. 4. 5. Not understanding the problem Not considering all of the causes Not identifying the real root cause(s) Failing to follow through Jumping to conclusions

FALSE NON VALUE ADDED ACTIONS 1. 2. 3. 4. 5. Runaway tests and analysis Just remove and replace the defective item Returning the defective part without analysis Tear systems apart without a plan shotgun approach Silver bullet theory

You might also like