STATISTICAL METHODS IN CHEMICAL PRODUCTION Introduction

Cuthbert. Daniel. Ind. Eng. Chem. , 1951, 43 (9), pp 2053–2053. DOI: 10.1021/ie50501a025. Publication Date: September 1951. ACS Legacy Archive...
0 downloads 0 Views 109KB Size
STATISTICAL iETHODS Iw CHEWICAL PRODUCT101 Presented before t h e Division of Industrial a n d Engineering Ch.emistry at t h e 119th Meeting of t h e American Chemical Society, Boston, Mass.

The control of quality by statistical methods is now being applied extensively in the process industries. The genius of Walter Shewhart, the high competence of a large number of other statisticians, and the urgencies of wartime production have combined to produce sets of relatively simple tables and rules for statistical quality control. The continual need of the production engineer t o know if this pressure or that flow rate is too low or too high can often be answered by quality control methods. But we are now moving, and with considerable speed, into a new phase. In development and production divisions, we are finding that statistical methods often give the means of testing a large number of hypotheses a t the same time and with extraordinary sensitivity. It is mainly with these methods that our symposium is concerned. The analysis of variance, multiple egression, and tests for randomness, are three of these methods. These were all developed to answer questions in other fields. Their successful application to problems of the chemical industry speak well for their generality and we may expect that the peculiar attributes o f the process industries, especially those connected with continuity of flow, will generate new problems to challenge the research statisticians. CUTHBERT DANIEL

clience of Trouble w i t h the advent of statistical techniques in industry, the ark of production trouble shooting is fast becoming a science. Use of these more rational diagnostic methods on process ailments imposes the obligation to plan experiments, perform tests, and collect data so that results tell the truth. Costly wrong decisions have been traced to inadequate, ill-planned, or prejudiced experimental data. Whether or not statistical analysis is used, scientific trouble shooting must employ rational procedures, such as separation of a multistream process into unit streams, and dissection of variability into within-hatch, batch-to-

batch, and time-to-time components. In plant scale experimentation several variables can be studied simultaneously, but all extraneous variables capable of influencing the test must be standardized or randomized. In a typical problem several methods of attack are critically evaluated.

T

industry. The statistical approach not only alters the form and manner of carrying out these functions, but it induces a profound metamorphosis in the brain cells of those individuals who embrace its methodology. They find in this new ideology n freshness and vigor with which t o attack the stodgy and ritualistic thinking habits t h a t exist around them. The generation of such a force inevitably leads to a re-examination of long-used methods and procedures, to the discarding 0.f some and reinforcement of others, ultimately to the achievement of a synthesis between the old and the new. It is precisely through this mechanism that many new sciences have been developed.

HE advent of statistical techniques a t the operating level in industry is fast becoming a reality. The growing use of quality control charts in manufacturing and the adoption of uniform sampling acceptance procedures in inspection attest to this fact. Add t o this the xyidening circle of engineering design, research, and development activities that have embraced the principle of statistical examination of engineering data and it is clear that the penetration is deep, indeed. What is more important than the ostensible evidences of the greater use of statistical techniques, however, is the far-reaching impact of the statistical philosdphy on the many functions of

LEONARD A. SEDER Gillette Safety Rasor Co., Boston 6, Mass.

2753