Analysis of Product Assembly Quality Using Regression Recovery Methods

To ensure high product reliability, it is necessary to provide for various assembly methods: full interchangeability method; partial interchangeability method; group (selective) interchangeability method; adjustment method using compensators; fitting method. An effective method for ensuring a specified accuracy can be selected through error calculations, including the calculation of dimensional chains for the assembled product. Assembly error calculations are performed to determine the resulting error and identify deviations that have the greatest impact on the output parameters. The calculation can be performed analytically using statistical data. The following errors may occur during product assembly: deviations in size, shape, and relative positioning of parts, poor-quality mating, and part deformation. These factors affect the accuracy of the output parameters and their reliability. Assembly errors can be systematic, random, methodological, and instrumental. Standard methods for processing empirical data are widely known. However, they do not fully refl ect the specific features of production and technological processes. In practice, it is usually necessary to work with large data sets, and the complexity of calculations becomes very high. In this paper, an automated system for analyzing the actual accuracy of product assembly based on regression reconstruction methods is developed. The practical significance of this work lies in the possibility of using the developed program to analyze the actual accuracy and distribution laws of the geometric parameters of assembled components. The actual distribution of vector errors obtained in production conditions can be described by a variety of different distribution functions. In most cases, vector errors determining the geometric accuracy of a surface are described by a modified Rayleigh law. This law is convenient for computer interpretation when calculating vector and scalar-vector dimensional chains using statistical data obtained in production.