Sponsored By

November 28, 1999

7 Min Read
Applying Six Sigma to design

Using the principles of Six Sigma has helped many manufacturing operations reach higher productivity with fewer defects. Gavin Finn, president and CEO of Prescient Technologies Inc., believes designers can implement the same process for faster product development and greater productivity. Presented here are excerpts from Finn’s paper “Implementing Six Sigma in Engineering Design,” along with comments made during an exclusive IMM interview.

Over the past 15 years, manufacturing experts have focused on quality programs for the shop floor with some amazing results. Six Sigma standards at companies such as Motorola and General Electric have increased quality and profits while reducing defects and downtime. Transformations on the shop floor, however, have not included the design function because managers, and even engineers, tend to view design engineering as a creative process that can’t be quantified.

We’ve found that as design departments incorporate 3-D CAD, product data management (PDM) tools, and team engineering principles, the activities that designers perform from one product to the next have become more repetitive. In other words, the product changes but the system does not.

New work methods and tools for the engineering design process have resulted in the creation of an engineering product—the digital product model. Designers can begin to improve engineering quality by measuring the digital product for errors.

Unnecessary design errors cost manufacturers millions of dollars annually, and costs increase dramatically if the error is detected late in the design-to-manufacturing cycle. Roughly half of the changes made in a product development cycle are the result of errors and omissions. Designers spend only about 10 percent of their day doing creative tasks, with a much larger chunk of time spent cleaning up and repairing digital models. Adopting a Six Sigma approach offers designers a way to eliminate wasted time by early error detection and prevention.

Analyzing Digital Models
In general, the Six Sigma method applies statistical analysis to the measurement of end-product traits against specifications. This helps determine how well the process for making those products is working.

Digital models have many of the same features as the actual end product, especially when it comes to geometry. Nongeometric properties are represented in the digital version in symbolic form. For example, a hole diameter is the same in both the actual part and digital model, while material properties inherent in the physical part are represented as attributes, parameters, or even formulas in the digital model.

The typical design engineering process begins as engineers work on a set of part and assembly models using a CAD/CAM system. Designs are then released into another system that makes the data available for analysis, moldfilling simulation, and mold design. Until recently, the only way to track engineering quality was to count the number of engineering change orders after the fact. Now it is possible to automate analysis of design errors and to correct them prior to manufacture.

To apply Six Sigma principles to the design process, data must be captured and measured using one of the commercial design quality software packages now available. Software can then be used to determine how data deviate from defined standards. Analyses can be run at several levels—assemblies, subassemblies, individual parts, and features of each part—analogous to checking a text document against established standards for spelling, grammar, and punctuation.

Setting Standards
How a CAD model is built can be just as important to Six Sigma quality levels as geometric accuracy. By adopting standards such as how to make a hole, which layers should contain surfaces, and how thick ribs should be relative to wall thickness, all of the digital product models within an organization can be easily modified. Without these standards, each designer’s model may be too unique for general use.

To demonstrate this, Prescient Technologies conducted a test at a major aerospace company. Seven engineers were asked to build a solid model of the same part using the same CAD system. Each engineer was then given another person’s model and asked to modify it. Because they were unfamiliar with the way the model was built, not one of them could make the changes. After receiving his own model back, each engineer made the modifications easily.

Consistent, or best, design practices can be built into a design department’s Six Sigma process as the standard by which models are evaluated. This type of system benefits not only existing designers, but also contract engineers as well as newly hired designers.

Measurable Features
Digital models contain a number of features that can be measured for statistical purposes. Geometric representation is the most identifiable of these traits, and consists of a set of features at a lower-level mathematical basis or a higher-level symbolic one. Lower-level examples include boundary surfaces, curves, and other geometric elements. Each element requires a discrete number of operations to create it, and each operation can be considered an opportunity for error.

Symbolic representations of features are a collection of both geometric and dimensional elements grouped to form more meaningful entities. For example, a hole feature would consist of a set of surfaces as well as some other basic mathematical entities, and would be described by its x-, y-, and z-axis location, radius, and length. Each of these descriptions is, again, measurable as an opportunity for error.

Nongeometric entities, such as material properties, can be strings, integers, floating numbers, lists, or any combination. Opportunities for error here are defined as any required field in the attribute list. For instance, if a digital model requires definition of the part’s material, having no value or an incorrect value would represent an error.

Finally, the form and format of the model can also be measured in a Six Sigma quality program. Defining a digital model format might include naming conventions for elements within the model, layer structure for model construction, accepted or best modeling practices, and placement and content of text notes. Every item in this format description is a measurable opportunity.

Editor’s note: Prescient Technologies offers a family of engineering quality software that supports the Six Sigma process in engineering. For more information, check the contact box below.

Five steps to Six SigmaIn essence, Six Sigma methods characterize quality by measuringactual errors vs. the number of opportunities for making sucherrors. The resulting statistic is listed as the number of defectsper million opportunities (dpmo). Each sigma (s)level represents one standard deviation from the average, andas a process moves away from the average, its defect rate fallsexponentially. As a result, processes that fall within the 1s range (one standard deviation from theaverage) have the highest dpmo rating, while those at 6shave the least. Arriving at the quality factor, or sigma level,is as simple as comparing the actual dpmo to the appropriate valuein the commonly used table shown here.

For high-volume applications, looking at quality in this wayis much more meaningful than using quality percentages. For example,a process that is 99.73 percent quality compliant sounds high.In fact, this figure means defects occur .27 percent of the time.For commercial airline flights, this translates to 2,700 crashesout of every 1 million flights, a clearly unacceptable level.


 Defects per million opportunities [dpmo]







The Six Sigma quality assurance method has several importantsteps:

  1. Define. Specific characteristics of the process are detailed, and key quality measures (CTQs—critical-to-quality characteristics) are identified in detail.

  2. Measure. Every step in which an error or defect could occur in a CTQ is identified. This creates a quantified metric describing the number of opportunities for defects.

  3. Analyze. Process is measured to obtain a baseline, or datum, sigma level against which future measurements will be compared. Analysis includes counting the actual defects, errors, or failures, identifying opportunities for improvement, and setting quantitative objectives for these opportunities.

  4. Improve. Steps are taken to improve the quality performance of the process as measured in step two. The top measures are isolated for a focused approach to process improvement, and ideas are generated for changing the process to reduce quality problems.

  5. Control. Methods to control the process are set. Process performance (variation) is constantly measured, and those parts of the process within quality limits are maintained. Control charts are used to identify out-of-control elements of the process and to take corrective measures.

Sign up for the PlasticsToday NewsFeed newsletter.

You May Also Like