德国汽车工业质量标准VDA Kapitel 4-05_en
VDA-Volume 4: Design of Experiments 1 We would like to thank all the companies and their staff who have contri- buted to this workgroup: AUDI AG, Ingolstadt BMW AG, Munich Robert Bosch GmbH, Stuttgart Continental AG, Hannover DGQ (Deutsche Gesellschaft für Qualität), Frankfurt Fichtel the more unfavorable the assignment to the evaluation criterion, the lower the ranking. For example: Effort for setting the input variable 1 large . 10 small Assumed factor influence 1 small . 10 large Costs for changing the setting of the input variable 1 high . 10 low The multiplication of the means determined for each evaluation criterion yields a statistical parameter for the priority of an input variable. The input variables are ordered according to their priority. 1.3.3 Input Variable Weighting All team members work together to evaluate and weight the input variables. The initial result is an average ranking for each of an input variable's evaluation criterion. Subsequently, all of an input variable's average rankings are multiplied together. The result is a statistical parameter for this input variable's priority. Finally, the input variables are ordered according to their priority and thereby yield the weighting. 1.3.4 Impact Matrix (according to Scheffler) An impact matrix is a tabular presentation of the changes in a target quantity or target quantities when input variables are varied. The wave form and symbols characterize the assumed or known change. Effect known Effect known and non-linear Effect suspected Effect unknown X1 - Xn = input variables Y1 . Yn = target quantities Figure 4 Impact Matrix According to Scheffler 1.3.5 Interactions In selecting a suitable and economical experiment design, it is particularly important to acquire preliminary information on possible interactions between the input variables. If interactions that actually exist are unintentionally not taken into account when the experiment design is defined, the experiment results may lead to incorrect statements. 1.3.6 Factor Levels Those input variables that are considered in the experiment design are called factors. In the simplest case, two factor levels are assumed. The two levels of the separate factors are selected with “suitable“ separation from each other, on the basis of technical considerations, boundary conditions and experimental feasibility. For qualitative factors, the imprecision of the setting must be negligibly small compared to the difference between the intervals. VDA-Volume 4: Design of Experiments 10 1.3.7 Factor Selection Summary in a Flow Chart no Input variable evaluation, evaluation criteria and evaluation scale Input variable weighting mathe- matical combinations from 2 Set up factor impact matrix to target quantities Obtain preliminary information on possible interactions among the factors Define factor levels Check compatibility Unsuitable for experiment design Input variable reproducible independent ? yes 6 5 4 3 2 1 Figure 5 Flow Chart for Factor Selection and Factor Levels VDA-Volume 4: Design of Experiments 11 VDA-Volume 4: Design of Experiments 12 1.3.8 Example There is sufficient basic knowledge about the principle effects of the diverse input variables for the machining process. With the help of this existing knowledge, the following eight significant input variables are taken from Figure 3 and defined as factors for the examination: A = Cooling lubricant S = Cutting speed C = Cut depth D = Feed rate E = Material F = Cutting edge angle G = Chip format H = Cutting radius To obtain the simplest possible experiment plan, each of these factors was examined at two levels. With the help of existing experience with the process, the team defines the following factor levels: Factor levels Factor - + A B C D E F G H No 100 m/min 1 mm 0.2 mm/h Material 2 45° Small 0.8 mm Yes 150 m/min 2 mm 0.3 mm/h Material 1 75° Large 1.2 mm Figure 6 Factors and Factor Levels in the Machining Process To the extent that knowledge or assumptions are available regarding the effects of these input variables on the target quantities, they are shown in the impact matrix shown in Figure 7. Influencing factors Target quantity A Cooling lubricant - + B Cutting speed - + C Cut depth - + D Feed rate - + E Material, work piece - + F Cutting edge angle - + G Chip from groove - + H Cutting radius - + large Surface roughness small unfavorable Chip form favorable Figure 8 Impact Matrix for Known and Assumed Effects of the Eight Selected Factors for the Target Quantity 1.4. Selection of an Experiment Strategy Several especially frequently applied experiment designs will be described in brief. They represent only a small extract from the known experiment strategies. More comprehensive information is given by Juran |7|, for example. 1.4.1 Single-Factor Experiment A single-factor experiment is an examination of the effect that one quantitative or qualitative factor has on one or more target quantities. The factor is set to t