Six Sigma: The new tool in your maintenance toolbox

David Berger says maintenance departments can benefit from Six Sigma programs.

By David Berger, P.Eng., contributing editor

1 of 2 < 1 | 2 View on one page

A popular methodology that companies of all sizes and industries are using is Six Sigma. The technique is best known for its ability to reduce product and service quality problems. Although Six Sigma shares objectives with Lean (improved processes, waste reduction, increased productivity and greater customer satisfaction), the methodology is more data-driven, quantitative and statistically based than Lean. Be assured that maintenance departments can benefit from Six Sigma programs.

Statistics reflect reality

Six Sigma provides systematic problem-solving using a variety of statistical tools and analysis. The term “six sigma” comes from the statistics measure of deviation from the mean. For normal distributions, 68% of the population should fall within one standard deviation — one sigma — from the mean. Similarly, 95% and 99.7% falls within two and three sigma, respectively.

Assume a specification calls for a part 1.00 inch in length, with 3 sigma being equal to 0.01 inch. With 3-sigma quality, you’d expect parts to be within spec 99.73% of the time (a defect rate of 2.7 per 1,000 parts). This was the accepted quality benchmark in manufacturing before the emergence of Six Sigma.

However, some companies felt that a 3-sigma standard wasn’t good enough. Motorola, for example, observed that a process could drift by about 1.5 sigma over time. In the example above, this would cause the process mean to range from 0.995 to 1.005, which might represent a significant shift for some customers.

Thus, keeping the data points within an acceptable range required a counterbalancing tightening of tolerance. For a 1.5-sigma drift in the mean (half of 3 sigma), the sigma level tolerance would need to be half, or plus/minus 6 sigma.

The acceptable Six Sigma tolerance level is 3.4 defects per million opportunities (DPMO), that is to say at least 99.9996599% of data points should fall within plus or minus 6 sigma from the mean. Although this really represents 4.5 sigma in a normal distribution in which the mean doesn’t drift, it’s considered Six Sigma because of the expected 1.5-sigma process shift.

View more Six Sigma content on PlantServices.com

Some experts argue that the 1.5-sigma process shift is more empirical than theoretical. In fact, any process mean that changes as much as 1.5 sigma should be considered statistically out of control, unpredictable and, therefore, at risk of producing defects, regardless of the customer’s specification limits. The good news is that with modern technology such as condition-based monitoring and control software, trends can be tracked for detection and correction of any significant process drift.

Even with countless variations on the theme, there are really two fundamental Six Sigma methodologies: DMAIC and DMADV. Both strive to achieve predictable, defect-free performance, and are similar to Deming’s “Plan-Do-Check-Act” approach. Although the two frameworks have similarities, there are significant differences.

DMAIC

This version — define, measure, analyze, improve and control - is applied to existing substandard business processes. Define establishes goals for improvement in line with customer demands and overall business strategy. This might be hierarchical, such as improved return on capital employed (ROCE) at the overall business strategy level, increased asset performance at the Operations and Maintenance departmental level, and reduced defects at the improvement project level.

Measure refers to tracking data related to the process, using reliable metrics relevant to the goals established in the first step. Analyze involves using statistical tools and root cause analyses to identify ways to minimize the gap between current metrics and the desired goal. Improve means optimizing processes based on analysis, using project management and change-management techniques to ensure effectiveness. Control refers to process monitoring and control to correct variances before defects appear. It requires adjusting policies and procedures, budgets, compensation and incentives, information systems, organizational structure and so on to ensure results sustainability.

DMADV

This version — define, measure, analyze, design and verify — applies to designing products or processes, or to an existing process that requires more than an incremental improvement. Define determines the design goals in light of customer demands and deliverables. Measure helps determine product, service and process characteristics, as defined by the internal or external customer. It uses actionable and quantifiable business specifications such as design failure mode effects analysis (DFMEA) as part of a reliability-centered maintenance program. It’s a risk assessment.

1 of 2 < 1 | 2 View on one page
Show Comments
Hide Comments

Join the discussion

We welcome your thoughtful comments.
All comments will display your user name.

Want to participate in the discussion?

Register for free

Log in for complete access.

Comments

No one has commented on this page yet.

RSS feed for comments on this page | RSS feed for all comments