Modelling and controlling variation propagation in mechanical assembly of high speed rotating machines

Assembly plays a vital role in the quality of a final product and has a great impact on the manufacturing cost. The mechanical assemblies consist of parts that inevitably have variations from their ideal dimensions. These variations propagate and accumulate as parts are assembled together. Excessive amount of variations in an assembly may cause improper functionality of the product being assembled. Improving assembly quality and reducing the assembly time and cost are the main objectives of this thesis. The quality of an assembly is determined in terms of variations in critical assembly dimensions, also known as Key Characteristics (KCs). Key Characteristics are designated to indicate where excess variation will affect product quality and what product features and tolerances require special attention. In order to improve assembly quality and reduce assembly time and cost, it is necessary to: (1) model non-ideal parts based on tolerances defined in design standards or current industrial practice of component inspection, (2) model assemblies and their associated assembly processes to analyse tolerance stack-up in the assembly, (3) develop probabilistic model to predict assembly variation after product assembly, and (4) implement control strategies for minimising assembly variation propagations to find optimum configuration of the assembly. Two assembly models have been developed, a linear model and a fully non-linear model for calculating assembly variation propagations. The assembly models presented in this thesis also allows for inclusion of geometric feature variation of each assembly component. Methods of incorporating geometric feature variations into an assembly variation model are described and analysis techniques are explained. The assembly variation model and the geometric variation models have been developed for 20 and 3D assemblies. Modelling techniques for incorporating process and measurement noise are also developed and described for the nonlinear assembly model and results are given to demonstrate the calculation of assembly variations while considering part, process and measurement errors. Two assembly case studies originating in sub-assemblies of aero-engines have been studied: Case Study 1, representing the rotating part (rotor) of an aero-engine, and Case Study 2, representing non-rotating part (stator) of an aero-engine. A probabilistic method based on the linear model is presented as a general analytical method for analysis of 3D mechanical assemblies. Probability density functions are derived for assembly position errors to analyse a general mechanical assembly, and separate probability functions are derived for the Key Characteristics (KCs) for assembly in Case Studies 1 and 2. The derived probability functions are validated by using the Monte Carlo simulation method based on the exact (full non-linear) model. Results showed that the proposed probabilistic method of estimating tolerance accumulation in mechanical assemblies is very efficient and accurate when compared to the Monte Carlo simulation method, particularly if large variations at the tails of the distributions are considered. Separate control strategies have been implemented for each case study. Four methods are proposed to minimise assembly variations for Case Study 1, and one error minimisation method is suggested for assemblies of Case Study 2. Based on the developed methods to optimise assembly quality, the two case studies were investigated, and it was found that the proposed optimisation methods can significantly improve assembly quality. The developed optimisation methods do not require any special tooling (such as fixtures) and can easily be implemented in practice