Simulation-Driven Development for Increased System Reliability

Sammanfattning: The product development process at industrial companies has traditionally focused on hardware-oriented solutions. However, a current trend is to extend business offers to include more service-oriented solutions e.g., functional product business models. A functional product typically includes the four main constituent’s hardware, software, service support system and management of operation. Availability is one critical property for functional product businesses which the customer and provider must agree upon. Hence, performing simulations for functional product development and operation enables the possibility for functional product availability prediction andincreases the possibilities for business offerings. Functional product availability is a function of reliability and maintainability. Today, many companies measure the strength of the hardware product in terms of durability or life length. However,measuring strength in terms of reliability, which is based on a statistically significant basis, supports the functional product development process. Simulations have generally been used to verify existing solutions; therefore, simulation-driven design strategies have been proposed to more rapidly converge on optimal solutions i.e., simulations are used to drive the development rather than simply verify suggested solutions.Today, measured data or estimated data are often used as input to reliabilityprediction methods e.g., fault tree analysis or Bayesian network. When designing new hardware systems, concepts need to be evaluated by means of reliability, but such data may not exist and prototypes often need to be manufactured, which is expensive and time-consuming. A combination of deterministic and probabilistic simulations can be used to derive needed input data for reliability prediction methods i.e., concepts can be evaluated by means of reliability in early stages of the functional product development process, even before a physical prototype is manufactured. The objective of the work presented in this thesis is to develop a simulation-drivenreliability prediction methodology for product development processes.This methodology shall be applied in early phases of existing product development processes, especially to generate, evaluate and select concepts in terms of reliability, to be used as a decision basis for systems solutions. Further, this methodology takes important variations into account and includes a combination of deterministic and probabilistic simulations.The conducted research is based on theories regarding the functional productdevelopment process, product development methodologies (e.g., stage-gated process), reliability prediction methods (e.g., fault tree analysis), deterministic simulation methods (e.g. rigid body dynamics) and reliability engineering (e.g., probability and reliability theory, distribution functions and Monte Carlo method). The research presented in this thesis followed a 5-step procedure including as-is study, to-be scenario development (simulation-driven methodology and a future functional business situation are described), methodology development, methodology verification (through case studies) and methodology validation (through system testing).Existing reliability methods were evaluated according to suitability in differentdevelopment phases during the as-is study. Fault tree analysis and probabilistic variation mode and effect analysis were found to give accurate results due to use of objective input data. It can be concluded that deterministic simulations (such as rigid body dynamics or welding sequences) can be used to derive input data to be used for probabilistic reliability prediction methods. Probabilistic variation mode and effect analysis can be used to derive a safety factor when designing systems, based on components variation contribution. Hence, a method has been developed based on probabilistic variation mode and effect analysis to derive system reliability. The method uses a distribution function (Normal distribution) and the variation contribution from included components. A method based on deterministic simulations to derive component reliability information has further been developed. This method takes different variations intoaccount and through a series of simulations, input data for system reliability (such as fault tree analysis and Bayesian network) can be derived.A simulation-driven hardware system reliability prediction methodology wasdeveloped, including both deterministic and probabilistic simulation models and methods. The methodology is used to predict hardware system reliability in early phases of the functional product development process and to partly exclude hardware system testing. The methodology takes into account component variations when a limited amount of information exists. The simulation-driven methodology will further be implemented in the company product development process. A case scenario was developed for the simulation-driven mobility function methodology. The methodology manages an increased amount of multidisciplinary interactions to combine deterministic and probabilistic simulation, both in parallel and coupled, within and between the four main constituents. The simulation-driven methodology is primarily focused on mobility functions but is more general and canpartly be used in the simulation-driven methodology for hardware system reliability prediction.