Variation Control in Virtual Product Realization - A Statistical Approach

Doctoral thesis, 2011

All manufacturing processes are afflicted by variation, which leads to products that may not fulfill assembly, functional or esthetical requirements. The variation in the final product originates from variation in individual parts and assembly processes. In this work, methods and tools aimed to reduce the amount of variation and, more importantly, the effects of the variation in the final product are treated. Four main areas are treated in this thesis: • Statistical process control: When a deviation or increased variation arises in a process, it is important to discover this as soon as possible. Different methods for statistical process control, based on inspection data, are suggested and compared. When an increased variation or deviation is obtained, it is also important to find the cause. Methods for the diagnosis of assembly fixtures are suggested, compared and evaluated. • Inspection: In the verification phase of product development, a product is normally inspected at a large number of points in order to learn as much as possible about the product and processes. When full production starts, the main purpose of inspection shifts toward statistical process control, which normally requires fewer inspection points. Today, the reduction in demand is usually done manually. This puts great demands on experience and craftsmanship. A method based on cluster analysis is presented and applied to inspection data in order to achieve a systematic reduction. By the clustering of the data, a representative for each cluster or group of points with correlated inspection values can be selected. By inspecting only those representatives, the number of points needed to monitor the product can be reduced. This method leads to larger reductions than manual procedures. The information in a set of inspection points is quantified. • Spot welding sequence optimization: By altering the joining sequence when joining two parts together, using, for example, spot welding, the level of variation in the final assembly is affected. Different strategies to find optimal spot welding sequences (in other words, sequences that minimize the variation in the final assembly) are suggested and compared. • Simulation accuracy: Variation simulation is used to predict variation in the final product or in sub-assemblies. In that manner, the requested number of physical verifications and tests can be reduced, leading to saved resources and increased sustainability. However, if the simulation is to be able to replace physical tests, it is vitally important that the simulation results are accurate and agree with real outcome. Factors affecting this agreement for non-rigid variation simulation are investigated. Four of the factors (spot welding sequence, contact modeling, fixture repeatability and the influence of heat during welding) are further examined and methods to include them in variation simulation are suggested.