Đang chuẩn bị liên kết để tải về tài liệu:
High Cycle Fatigue: A Mechanics of Materials Perspective part 15

High Cycle Fatigue: A Mechanics of Materials Perspective part 15. The nomenclature used in this book may differ somewhat from what is considered standard or common usage. In such instances, this has been noted in a footnote. Additionally, units of measurement are not standard in many cases. While technical publications typically adhere to SI units these days, much of the work published by the engine manufacturers in the United States is presented using English units (pounds, inches, for example), because these are the units used as standard practice in that industry. The graphs and calculations came in those units and no attempt was made to convert. | 126 Introduction and Background and of mean log-lives for a material having variable scatter provided that the precision at the highest stress level is markedly lower than the others. Their results also showed that results obtained from test programs with different number of specimens per level are better than those of plans with uniform replication. For the particular set of data used from which experimental results were chosen at random for the statistical analysis a confidence level of 90 could be obtained over the range ct using the following test programs 10-3-5-6 specimens a total of 24 specimens tested at levels 550-520-490-460 MPa for the exponential Nelson model Equation 3.22 and 10-3-6-10 specimens a total of 29 specimens tested at the same stress levels for the linear model Equation 3.23 . 3.8.1. Run-outs and maximum likelihood ML methods A unique feature of testing in the HCF regime is the occurrence of run-out tests where the test is terminated after a certain large number of cycles. Alternately a test that fails after N cycles can be used in an analysis describing the behavior for fewer than N cycles so it can then be treated as a run-out for the fewer cycles being considered. Run-outs in statistics terms are considered to be censored data that is they fall outside the range of behavior lives being considered. In fitting models to data containing run-outs the run-outs cannot be used in least-square fits because the exact location of the data point is unknown. All that is known is the fatigue life was greater than some quantity. Since least-squares methods cannot use censored data because the data points do not really exist other methods have been employed. As mentioned in the discussion of the RFL model previously maximum likelihood ML estimation schemes are able to deal with censored data from a statistical point of view. The virtue of the ML method is that it applies to virtually any assumed statistical distribution of lives or fatigue strengths as