آزمون های زمانبندی در پروژه های "تحقیق و توسعه" صنعت خودرو

In automotive R&D projects a major part of development cost is caused by tests which utilize expensive experimental vehicles. In this paper, we introduce an approach for scheduling the individual tests such that the number of required experimental vehicles is minimized. The proposed approach is based on a new type of multi-mode resource-constrained project scheduling model with minimum and maximum time lags as well as renewable and cumulative resources. We propose a MILP formulation, which is solvable for small problem instances, as well as several variants of a priority-rule based method that serve to solve large problem instances. The developed solution methods are examined in a comprehensive computational study. For a real-world problem instance it is shown that the introduced approach may enhance the current methods applied in practice.

During the last decades the automotive industry has been confronted with a shortening product life cycle, as its market changed from a seller’s to a buyer’s market (cf. e.g. Henseler, 2003). Consequently, the time-to-market has been reduced in order to support the integration of current needs from customers in the development of new car models. Furthermore, the number of cars produced throughout the life-span of a model cycle has decreased, which has led to an increase in the portion of indirect costs. While production divisions were able to counteract this trend, for instance by enlarging the utilization of non-variable parts among different car models (cf. Stake, 2001), it is still a challenge to significantly reduce development costs in interaction with a decreasing time-to-market (cf. Gembrys, 1998 and Risse, 2002). The product development process in the automotive industry generally consists of two alternating stages. First, new components are constructed using computer aided engineering techniques. Subsequently, these components are tested with the help of experimental vehicles that have to be built up by the prototype section. Testing is necessary to reveal further demand for engineering in order to reach the level of quality customers expect. Additional tests verifying certain product attributes are prescribed by law. While engineering costs have decreased throughout the last few years due to the successful implementation of platform strategies, testing costs have risen because of increasing product complexity and variety (cf. Risse, 2002). Since the construction of one experimental vehicle costs up to one million Euros, the majority of testing costs is caused by the prototype section. All tests that have to be carried out are specified in advance such that the demand for experimental vehicles depends only on the schedule of these tests. Thus, we consider a scheduling problem where we have to determine a start time for each test such that the number of required experimental vehicles is minimized and several constraints are met.

In this paper, we proposed a formulation for the problem of scheduling tests in automotive R&D projects as an appropriate multi-mode resource-constrained project scheduling model with minimum and maximum time lags as well as renewable and cumulative resources. A novel challenge in the context of project scheduling is that the key resources must be created by certain project activities in order to enable the execution of further activities, but in advance it is unknown how many of these resources are required. Moreover, the aspect of partially ordered destructive relations between certain project activities has not been considered before. For small instances a MILP formulation of the problem can usually be solved to optimality within 10 seconds. In order to solve large instances, both priority-rule based single-pass as well as multi-pass heuristics are devised. A comprehensive computational study shows that (compared to the single-pass heuristics) the longer computation time of the multi-pass heuristics results in an improvement on the generated solution of up to 3.4%, where a so-called “slack optimized” scheduling approach performs best. For a real-world problem instance this approach outperforms the manual planning techniques currently applied in practice by 26% regarding the solution quality.