A CASE FOR SHARING SIMULATION DATA

REFERENCES 141 OEM worked concurrently to engineer both the product and the process. Simulations were conducted by the OEM to analyze the stresses within the component during operation. The supplier used computer modeling to develop a gate and runner system. The re- sults of the simulations were not exchanged due to ‘‘proprietary’’ reasons. Conclusions drawn from the results, however, were shared freely, resulting in minor changes to the design to improve function and manufacturability. At release, the partners believed both the product and process had been optimized. Prototypes were fabricated from production, such as tooling, for product verification tests. The initial prototypes failed. Al- though simulations were conducted concurrently, the simulation results were never examined side by side. Comparison of the sim- ulation results showed that two major metal fronts in the die con- verged at the point in the structure that bore the highest stresses during operation. The process tooling was redesigned such that the two metal fronts converged at a low stress point within the structure. Had the results of the product and process simulation been compared earlier, redesign of the process would have been avoided. REFERENCES 1. Robust Design Using Taguchi Methods Workshop Manual, American Supplier Institute, Livonia, MI, 1998. 2. Solutions, American Supplier Institute, Livonia, MI, 1999. 3. Clausing, D., Total Quality Development, ASME Press, New York, 1994. 4. Vinarcik, E., ‘‘Minimizing Cost Through Part Integration,’’ Engineered Cast- ing Solutions, Winter 1999, p. 56. CONTROLLING QUALITY IN HIGH INTEGRITY DIE CASTING PROCESSES 145 10 APPLYING STATISTICAL PROCESS CONTROL TO HIGH INTEGRITY DIE CASTING PROCESSES

10.1 INTRODUCTION TO STATISTICAL PROCESS

CONTROL Before the industrial revolution, craftsmen controlled the quality of their products. As the industrial revolution eased into its second century, the mass production of products by unskilled labor de- manded a means to control quality. By bringing together the dis- ciplines of engineering, statistics, and economics, statistical process control SPC theory was first developed in the 1920s by Walter Shewhart while working for the Western Electric Com- pany. 1 At the time, Western Electric was a leading manufacturer of telephones in the United States, with over 40,000 people working at the Hawthorne Works in Chicago, Illinois. Product reliability was a major problem. Manufacturing inconsistencies caused ex- cessive failures of handsets, switching units, and amplifiers. One out of eight employees at the Hawthorne facility was an inspector whose role was to scrap defective product. Western Electric’s management realized that they had a bigger problem than scrap- ping defective products. They were paying 12 of their workers to do nothing that added value to their products. Western Electric’s management looked to minimize the sources of defective produc- tion and selected Shewhart to lead the effort. 146 APPLYING STATISTICAL PROCESS CONTROL Common-Cause Variation Special-Cause Signal a b Figure 10.1 Dart board comparison of a common-cause variation and b special-cause variation. Shewhart began studying several manufacturing processes in depth. As a practical engineer, he understood that in real life sit- uations, laws and theories are not exact. Shewhart concluded that all processes exhibit variation that can be classified into two dis- tinct types: 1. inherent or common-cause variation and 2. intermittent or special-cause variation. To concisely demonstrate the difference between these types of variation, a dartboard may be used as an example, as shown in Figure 10.1. The goal of each dart throw is to hit the center of the board. Common-cause variation is subject to chance with un- discoverable random causes. This is illustrated in Figure 10.1a by a random distribution of hits clustered around the center of the dartboard. Special-cause variation, however, does not fit into this predictable random variation, as shown in Figure 10.1b. Special- cause variation can be assigned directly to some event or phenom- enon. Shewhart believed that these assignable causes could be discovered and removed with an economic benefit. Once sources