This paper reports on an investigation into the modes of change propagation in complex products based on a series of case studies in an aerospace and an automotive company. The paper identifies three main conditions under which change tends to propagate and highlights the heterogeneous nature of factors that contribute to propagation. It was found that change tends to propagate between components when there are insufficient margins between coupled components, when there is an encroachment of design constraints or as a result of shortcomings of tools and processes used to support change processes. The paper concludes with a discussion on the implications of the nature of change propagation mechanisms on predicting the knock-on effects of change.@en

When using connectivity models to assess the potential impacts of component changes on other parts of a product, plausible inferences are readily assessed when such products are represented at the appropriate level of granularity to support specific queries. In this paper, we describe the development of a prediction algorithm, which enables coherent computations of the likelihoods of change propagating on several levels of detail of product description given component level change probabilities. The results show that a multilevel approach to change prediction supports an increased range of design queries beyond that achievable with a single level model. Such change prediction capability is useful when carrying out a comprehensive change impact assessment.@en

It is essential to assess the effects of change rigorously before a change is implemented, yet design engineers are faced with the problem that an exhaustive analysis of all product parameters affected by change proposals is simply impractical. In this paper, we describe a criticality-based approach to ordering predicted risks. This enables design engineers to concentrate their efforts on the components which are critically disposed to the effects of proposed changes. Using a diesel engine as an example, we show that components which have the most noteworthy risk of causing significant redesign if affected by a change are not always of the highest priority when evaluating effects of change.@en

Unforeseen side effects of engineering changes can affect other parts of a product, which are expected to remain relatively unchanged. As a result, known component relations are used as cues for assessing probable change propagation paths. In this paper, we assess the use of functions, behaviour and structural relations as cues for predicting engineering change propagation. It was observed from the analysis that estimates of possible propagation paths varied according to the type of the query as well as the type of cue used support prediction. The result indicates that in order to have credible estimates of change propagation paths, it is important that all three product-relations are considered when analysing change instances.@en

In many published design texts and standards, it is often advocated that hierarchical descriptions should be structured into successive levels of systems, assemblies, components and so. However, there is no one single hierarchy in a complex product. In practice, sub-units within a hierarchy are defined purely in an arbitrary manner. The lack of a structured approach to product decomposition leaves a chance of creating hierarchical decompositions with limited practical use. As a way of creating meaningful hierarchical product descriptions consistently, this paper suggests precautionary steps to aid the product decomposition process. These guidelines are based on insights gained from an assessment on decomposability of a motorcycle.@en

Connectivity models are useful aids to support design reviews but building the models is an extremely effort intensive process. Connectivity models help to minimise incidents of unexpected rework by drawing attention to vital component interfaces and dependencies. The uptake of connectivity models can be enhanced if effort burden associated model building is significantly reduced. This paper describes an investigation into the use of a distributed approach to building connectivity models as a means of effort reduction. The Design Structure Matrix is broken down into sub-sections which are L-shaped or cross-shaped. These sub-sections are referred to as modules of the connectivity model. The effort required for building a model is compartmentalised within the modules. In other words, the minimisation of model building effort is attained by distributing the modules such that each individual's effort contribution is limited to within a single module.@en