Desktop version

Home arrow Marketing

  • Increase font
  • Decrease font


<<   CONTENTS   >>
Table of Contents:

Control Phase

As we enter the control phase, we have identified the major root causes of poor KPOV performance. We know how to maintain optimum performance levels to the original targets by modifying the KPIVs, either because we have a regression model from analysis of historical data or experimentation or we have a non-mathematical solution derived using Lean tools and methods. Controls need to be placed on these KPIVs to ensure the solutions are sustainable. The controls will be incorporated into a control plan and the project will be transitioned to the process owner. Important questions will need to be answered at this time: Which variables need to be controlled? What is the control strategy (e.g., control charts, FMEA, audits, checklists, others)? How will controls be implemented? Who is accountable for monitoring and taking corrective actions? When will a control activity start and finish? How will we know if the process is not under control?

The important actions to ensure effective process control include listing the process steps in the updated future state map and identifying KPIVs from the analyze and improve phases on the map. If the project was focused on a product, the KPIVs will be features, functions, and

FIGURE 9.38

Response surface methods.

attributes. Table 9.20 describes common solutions to eliminate root causes and controls that help sustain the solutions. In the third column, a note is made relative to how easy it will be to sustain the solution. Eliminating root causes for a problem may require using a combination of solutions. As an example, perhaps updated procedures, training, and audits are all needed to sustain a solution. We want to avoid situations where only the less sustainable solutions and controls are used, instead using robust

TABLE 9.20

Controls on Process Improvements

Solution/Improvement Controls

Ability to Sustain

Verbal instructions

Audit, checklist, written procedure, testing

Difficult

Written instructions

Audit, visual, web-link to ensure compliance, testing

Training

Audit of frequency, content, web-link to ensure compliance, testing

Audits

Set schedule, skills, templates, feedback, and corrective action

Requires frequent audits

7-S methods

Visual control, roles and responsibilities, audit, feedback, and corrective actions

Statistical process control

Visual control, operator involvement, reaction plans (other control charts)

New roles and responsibilities

Monitor effectiveness through audits and measurements, incorporate thorough training and procedures

Easier

Total productive maintenance

Set schedule, skills, templates, feedback, and corrective action

Policy changes

Monitor effectiveness through audits and measurements, incorporate thorough training and procedures

Mistake-proofing

Limited or no controls (FMEA)

Eliminate operations

No controls needed

Design change

No controls needed

solutions and controls. If we can eliminate a process step or design feature that causes poor performance, the improvement is more easily sustained.

The combination of solutions is incorporated into the formal quality control plan shown in Figure 9.39. The two versions differ in their level of detail. Important questions are, Which outputs and inputs are important from the root-cause analysis and need controls on their solutions? How should the inputs be controlled and to what level? How should we measure each input and output to ensure it remains at its target level? What is the frequency of measurement, including inspection and testing? Who is responsible for control of the outputs and inputs, including training and work instructions? Everything needs to be documented in the control plan and integrated into the organization’s quality management systems. The control plan integrates all control actions and ensures process outputs and key inputs are under control. The control plan is a formal document

FIGURE 9.39

Control plans.

with supporting information that shows the KPIVs and how to control them so the KPOV stays on target with minimum variation. In addition, reaction plans are created and incorporated within the control plan to bring a process back under control if KPIV levels change dramatically. It is used throughout product or process life cycle and is a living document to ensure quality control and continuous improvement.

FMEA is used in the control phase of a DMAIC project to reduce the probability that KPIVs will move from their optimized levels. An example of an FMEA is shown Figure 9.40. At this point in the project, the FMEA is used as a risk management tool as opposed to helping identify causes for poor performance as in the earlier DMAIC phases. The risk assessment is used to evaluate the effectiveness of the current controls, the likelihood of their failure, and reaction plans to eliminate the failure. There will usually

FIGURE 9.40

Why is an FMEA useful?

be extensive supporting documentation added to the FMEA, such as process, testing, and maintenance instructions, as well as roles and responsibilities and other information, to quickly bring performance back on target. The FMEA is integrated into the quality control plan. There may be versions for both design and support processes. These enable tracking and prioritization of control risks and mistake-proofing the process based on risk.

Control charts are time series graphs that have been modified using control limits to aid in evaluation of the non-random patterns. Control limits are commonly set as ±3 standard deviations from the mean of the variable being charted. A normal (symmetrical) probability distribution will encompass 99.73% of the variation within ±3 standard deviations from the mean. Table 9.21 summarizes the most common charts in terms

TABLE 9.21

Common Control Chart Summary

Attributes

Individual Moving Range Charts

X-Bar and Range Charts

P-Chart/NP-Chart

C-Chart/U-Chart

Data format

Continuous

Continuous

Percentage or proportion

Counts

Use these charts for stated data format

We are measuring cycle time, hours, incident rates, temperature, weight, costs, or other continuous numbers.

We are measuring cycle time, hours, incident rates, temperature, weight, costs, or other continuous numbers that are subgroups (e.g., if we measure delivery time each working day, the sub-group is “week,” which has 5 days).

We take a sample of things and classify them as pass or fail based on a standard or evaluation criterion (e.g., if we take 100 pumps per day and classify them as clean or dirty, the percentage of clean pumps is plotted every day; if we take 200 invoices and classify them as accurate or not, we plot the percentage of accurate invoices per day).

We count the number of defects per sampling unit (i.e., the number of inaccurate data fields per invoice, the number of accidents per month).

We could also classify an invoice as accurate or not if any data field has an error; this gives us a percentage.

Sample size

25-125 observations in the total sample with sub-group size = 1

4-6 observations per sub-group and 20-25 sub-groups in the initial baseline sample

50 or more observations per sub-group; p-charts are constructed using sub-groups of equal or unequal size, whereas np-charts are used in situations where sub-group sizes do not vary.

Convenient sampling unit (1,000 square feet, every 1,100 people, incidents per month, etc.) that is constant from one sample to another; U-charts are used in situations where the size of the sub-group varies.

Assumed

distribution

Normal/moving range chart in control

Normal/moving range chart in control

Binomial

Poisson

of the data format, sample size, and assumed distribution. There are many other specialized control charts based on other assumptions such as short production runs or other distributions. Control charts are applied to the analysis of process variation and are used to monitor a process over time. If the process is stable, its behavior can be predicted and statistical conclusions can be drawn from the analysis. These charts differentiate process variation due to common versus special causes (i.e., trends, cycles, shifts, changes in variation, outliers, etc.). This allows identification and removal of special causes (e.g., outliers and non-random patterns) and prevents excessive tweaking of a process so that the process operates with less overall variation.

This is possible because a representative sample from the process is used as the reference distribution to set up the control charts statistical limits. The theory is that subsequent samples taken from the same process should match the reference distribution if the process has not changed over time. Data points exhibiting common-cause variation remain within a control chart’s upper and lower control limits without exhibiting non-random patterns, whereas special cause variation data exhibit non-random patterns such as outliers, trends, cycles, or other patterns. Control charts are constructed by taking sequential samples of size n from a process. Over time, additional samples are taken from the same process workflow of size n and compared to the reference distribution. If the process has not changed relative to its mean or variation, then the two patterns should be similar.

The control charts shown in Table 9.21 have as their basis different underlying assumption and practical uses. The most common difference is relative to the distribution of the variable being charted. As an example, if a variable is continuously distributed and sub-groups are taken from a process, then the resultant distribution of sub-groups will most likely be normally distributed (central limit theorem). This is the basis of the X-bar and R control charts. If a variable is measured as pass or fail, however, then the resultant probability distribution will be binomially distributed and discrete control charts such as p or np charts will be constructed. The fourth common control chart is based on counting the number of defects. In this application, the C control chart will be based on the Poisson distribution.

A p-Chart example is shown in Figure 9.41. It is constructed using sample proportions based on data collected in groups over time and classified as pass or fail. An application would be gathering 50 parts every hour from a process and classifying them as pass or fail based on evaluation

FIGURE 9.41

Control chart example.

criteria. In Figure 9.41, we see that the process is stable within the control limits, except for one outlier that is above the upper control limit. This data point should be investigated. It may have occurred by chance or it might be an assignable cause of variation whose root cause must be eliminated from the process.

Table 9.22 shows control and capability are different but both important. A process will need to be under control and stable before we can

TABLE 9.22

Control versus Capability

Capable?

In Statistical Control?

No

Yes

No

Bring process into statistical control and determine entitlement: then make improvements

Create an

improvement project (DMAIC Projects)

Yes

Bring process back into statistical control

OK

rely on sampling statistics such as capability indices, which are calculated using samples from the process. When we begin the DMAIC project, the first actions are to construct the baseline of the KPOV metrics using the appropriate control chart and to ensure it is stable before estimating performance to target. If the process is stable but not capable, we start the DMAIC project because this indicates a chronic process issue.

At the end of the control phase, we should be able to demonstrate that the process is both stable at the target performance level and under control as shown in Figure 9.42. Each project is different relative to its root cause analysis and solutions. Some are highly quantitative whereas others focus on Lean tools and methods. In each of these examples, the KPOV metric was significantly improved. In the process map, the percent of

FIGURE 9.42

Project metric baselines — are we capable? NVA = non-value-add; UCL = upper confidence limit; LCL = lower confidence limit.

non-value-add operation was reduced by 50%. Yields were improved and complaints were reduced in the other three examples.

Summary

In this chapter, we discussed the most common tools and methods used in the Six Sigma program from a management perspective. The Six Sigma initiative became popular in the mid-1990s as hundreds of organizations embraced its breakthrough methodology. The initiative revolutionized the way that quality improvement was applied not just to manufacturing but also to services and supply chains. It showed that quality management and improvement are critical for improving an organization’s competitiveness across diverse industries. In this chapter we continued the decision that it is important to align quality assurance and control activities with the concurrent engineering team to ensure products and services are designed to have high capability to meet customer requirements under a variety of actual use conditions. High quality products and services create competive advantages by simplification, standardization and doing things right the first time. In addition to breakthrough projects, an organization must also embrace continuous improvement and other quality initiatives. A quality program should be integrated to include continuous improvement. Six Sigma breakthrough and other initiatives that enhance customer experience. Associate training in basic quality tools, methods, and concepts to continuously improve their process helps change a culture and make it more customer centric, productive and competitive.

10

 
<<   CONTENTS   >>

Related topics