Project name: GenMech
Start date: September 1997 Finish date: March 1999
Funding
EPSRC Engineering in Manufacturing Grant £105,993
Staff
Chris Price, Patrick Olivier
Collaborating organizations
Jaguar Cars Limited, Pilkington Optronics, Integral Solutions Limited
Introduction
The aim of the project is to investigate the feasibility of automating mechanical failure mode and effects analysis (FMEA) for automotive subsystems employing qualitative reasoning techniques on representations of mechanical systems constructed from existing information sources such as CAD models and functional specifications. This will also crucially involve studying both the process and the information used by practicing automotive engineers in performing mechanical design FMEA.
The increasing complexity of design in automotive systems has been paralleled by increased demands for analysis of the safety and reliability aspects of those designs. The increased level of analysis must also be performed in a shorter time frame in order to meet reduced product lead times. Such demands can place a great burden on the engineers charged with developing a new design. Increasing the level of computerized assistance available can improve the quality of the analysis done while reducing the amount of effort needed to do it.
The qualitative reasoning research group at Aberystwyth has developed a system which can combine an electrical CAD design with a reusable description of the function of the design, and by simulation produce a consistent, human-readable FMEA report. The level of automation provided is significantly in advance of any other tool, and speeds up the electrical FMEA generation task from a matter of weeks to a matter of hours, thereby changing the design process [5].
An important component of this project is the development of techniques for automatically generating descriptions of device behaviour from existing design information such as functional specifications and CAD models. By using such representations, in combination with compiled knowledge of the functional roles of components and inferred structural knowledge (e.g. qualitative spatial location), failure modes could be introduced to the models, and classes of significant failure behaviour generated. If this can be achieved, this will serve as a good basis for an automated mechanical FMEA system, or at least a tool to assist in the construction of mechanical FMEAs.
We only expect this investigatory grant to be able to lay the foundation for further research that could be carried out in close liaison with automotive companies in order to finally produce a mechanical FMEA system with similar performance to that of our electrical FMEA system.
Aim and objectives of the Project
The main aim of the project is to investigate the feasibility of automating mechanical failure mode and effects analysis. This will be accomplished through investigation of existing practices, available design information and knowledge of the behaviour and failure modes of mechanical components. The aim of the project can be broken down into four main objectives:
- Understanding expert performance of FMEA. Despite its importance in the domain of safety-related systems, and in sharp contrast to system diagnosis which is understood from both a computational and a cognitive perspective, there are no existing studies of expert reasoning in the performance of mechanical FMEAs. Experience has proven that any proposed automatic assistant must be sensitive to both the cognitive context and formal and informal practice.
- Developing representations and integrations of domain models Mechanical FMEA can encompass a much wider set of problem domains than the electrical case, including dynamics, kinematics, structures, thermodynamics, magnetics, fluids, tribology and even chemical response (e.g. mechanical property change in the presence of corrosion). Unlike the electrical domain, components do not have unique functions, and in practice the function set of a mechanical component is distributed across the boundaries of conventional modelling approaches. For example, mechanical FMEA for a subsystem such as a fuel-injection pump must include concurrent representation and reasoning for dynamic, kinematic and hydraulic behaviour. Though it is often stated as being highly desirable, there is virtually no existing work in multi-modelling relevant to such mechanical systems.
- Construct and acquire compiled functional models of components and failure models. It is not possible to extract device functionality from a CAD specification alone. Detailed information regarding the geometry of a mechanism is typically decoupled from both design engineers' intent as regards the mechanism's role in the subsystem, and their tacit knowledge as to the mechanical context of its operation (e.g. what the desired responses to characteristic loadings are). By drawing exclusively on existing information sources from our industrial collaborators, and in particular CAD models and functional specifications, the project aims to integrate structural and functional knowledge, and in doing so to ground itself in the realm of genuine engineering problems. Past FMEA reports will also be available as a metric by which system performance can be meaningfully established.
- Mechanical FMEA report generation. The studies carried out during the project will determine the degree to which mechanical FMEA can be automated. The final objective will be to construct an experimental mechanical FMEA generation system using the information identified and the representations developed earlier in the project. The efficacy of this FMEA generation system will be evaluated against the case studies detailed earlier in the project.