A Structured Methodology for Process Improvement


Batson, Robert G.   (1990, ASQC)   The University of Alabama, Tuscaloosa, AL

Annual Quality Congress, San Francisco, CA    Vol. 44    No. 0
QICID: 9582    May 1990    pp. 1000-1005
List $10.00
Member $5.00

This article is not available online. Contact us to receive a scan of the archive, in PDF format.

Article Abstract

Moving from inspection and charting of product quality characteristics to the establishment of process control is a primary goal in quality improvement. Yet few authors have revealed how to make this transition and without a structured methodology, many may waste time and resources on fruitless approaches. Ishikawa states that process control activities endeavor to find technology that can engage in preventive control, i.e. control of process variables that results in high quality output along typically multiple dimensions of quality. Whether such control is obtained via automatic equipment or well-coordinated human actions, it is critical to establish the respective ranges within which process variables must remain in order to assure high quality output. These "acceptable operating ranges" are referred to as process specifications, and each centerline is called a process variable target. This paper describes the available methodology to determine these target values. Once the targets are known, further experimental work can determine the acceptable ranges about them, and control charting may begin to determine (1) is the variable in statistical control, and (2) is the current control technology capable of meeting the process specification?

Our methodology of process improvement consists of three phases: (I) Process Observation; (2) Process Characterization; and (3) Process Optimization. Process observation is critical in our methodology of process improvement because we are restricting the alternatives in phases 2 and 3 to methods that characterize and optimize real production processes, not laboratory experiments. Process characterization consists of defining a family of mathematical and statistical functions that relate each product quality characteristic back to a set of process variables. The alternatives for process characterization are mechanistic models, response surface models, and regression models, in decreasing order of preference. We claim that, with proper precautions, the use of regression models on time-indexed product and process data can be useful at the earliest stages of process improvement, as an aid to learning.

Process optimization utilizes the family of functions developed in process characterization in some sort optimization scheme. Because we are dealing with multiple quality characteristics (responses), compromise will be necessary in selecting the process variable targets. We present three such alternative schemes: (1) Optimize individual quality characteristics, then compromise; (2) Treat all but one quality characteristic as a constraint, and optimize the remaining characteristic -- repeat for each characteristic, then compromise; (3) Set targets for each quality characteristic, then employ goal programming to find the settings of the process variables that minimize deviations from product quality targets. Examples of successful use of each approach in the process industries are discussed.



Browse QIC Articles Chronologically:     Previous Article     Next Article

New Search

Featured advertisers

ASQ is a global community of people passionate about quality, who use the tools, their ideas and expertise to make our world work better. ASQ: The Global Voice of Quality.