Statistical analysis of mechanical test results for samples of pipes from trunk oil pipelines after long-term operation

Abstract

CONDUCTING ANY SCIENTIFIC RESEARCH includes the process of acquiring, processing, and analysing data which are obtained as a result of one or a series of experiments. Over the prolonged period for which science has existed, a multitude of methods have been created for obtaining experimental data. As part of studying the properties of materials, standard methods are applied for tensile, impact toughness, and hardness tests, as well as fatigue tests. This list of tests can be taken as a base which is used to establish the level and character of loads the given material is capable to sustain during operation. This paper examines approaches to processing experiment results which are obtained using standard (basic) methods when testing pipe steels for trunk oil and oil product pipelines. At the present time, the degree of automation in experimental equipment is fairly high. However, there is still some human participation when preparing samples and setting-up the experiments, which may lead to errors occurring when performing the tests (the human factor). This article presents some approaches which make it possible to perform initial processing and verification of experimental data for errors and reliability, and against hypotheses and distribution laws for the results obtained. Methods are also suggested for determining the minimum quantity of samples to be tested, based on evaluation of the mechanical property variations in pipe steels. Several approaches may be applied to solve the task of identifying links and relationships between the analysed parameters. The first approach is deterministic and requires long-term study. When one takes into account the considerable number of factors being studied, it is also fairly expensive and involves rather complicated experiments. The second approach is probabilistic and statistical, which allows implicit dependences to be established between series of measurements (correlation analysis) and linear (pair regression) and non-linear dependences (multiple regression) to be revealed between the parameters being examined. This study applies the statistical approach to analysing data from mechanical tests of pipe steels in trunk oil and oil product pipelines.