Promotie Erwin van Wingerden: “System-focused spare parts management for capital goods”
In het kader van ‘successen zijn er om gevierd te worden’ hebben wij het genoegen u hiermee te informeren over de promotie van Erwin van Wingerden, onderzoeker binnen het TKI Dinalog project Service Logistics for Advanced Capital Goods . De promotie vond plaats op dinsdag 24 januari 2019 aan de Technische Universiteit Eindhoven.
When a capital good, such as the lithography machine of ASML, used in a 24 hour production process is out of order for a period of time, this has a big impact on the entire production process and will cost the production company a large amount of money. Also for many other capital goods, the costs of not being able to operate are often considerable. A proper management of the maintenance of capital goods is important to make sure downtime of these capital goods is kept to a minimum. Cap ital goods generally consist of many different critical parts which cause the machine to stop working when a part has broken down. In case this happens, it is generally not easy nor the fastest option to repair the broken part on site, and therefore the part is preferably replaced by a spare part to bring the capital good back up and running as fast as possible. However, simply stocking an ample amount of stock is also not desirable as this results in high holding costs as parts can be very expensive. Therefore, one generally prefers to have the highest availability of capital goods while at the same time minimizing the inventory holding costs of spare parts. In this thesis, we focus on developing models with this aim in mind. We consider models that are applicable for the initial phase of a capital good as well as models for the stationary phase of a capital good.
To make stocking decisions, a company generally has estimates on the expected life times of critical components, and hence on the demand rates, hut these demand rates may be highly uncertain. This holds in particular during the initial phase of a newly designed system. This uncertainty in the demand rates leads to a double demand uncertainty: the common uncertainty that one also has under known demand rates and the additional uncertainty because of the uncertainty in the demand rates. When the rate is higher than expected, this involves high downtime costs whereas if the rate is lower than expected this involves unnecessary holding costs. In Chapter 2 we pro vide a model that is able to take the double demand uncertainty into consideration for the decision making. We show that, depending on the setting and distribution used to model the uncertainty, costs can increase to over 18 times the costs of the solution under known demand rates. More importantly, when one decides not to take the double demand uncertainty into consideration when deciding upon the base stock levels it is likely that this will result in a lot of downtime.
Having shown the benefit of more reliable demand rate information, we investigated how the decisions should change if there is a possibility to invest in more reliable demand rate information in Chapter 3. We consider a finite horizon that models the first years of the exploitation phase of a capital good, where decisions are taken at the start of the horizon on the initial base stock level and the investment in more reliable demand information and on the adjustment of the base stock level at a later time point. The problem is modeled as a three-stage dynamic program and we derive ana lytica! results to obtain the optimal base stock levels. We then numerically determine the optimal investment in more reliable demand rate information. Based on these results, we show that the price of a part and the level of uncertainty have the largest influence on how much can potentially be saved by investing in more reliable demand rate information. Another interesting observation we made is that the optimal base stock level at the start of the time horizon, when there is still a large amount of double demand uncertainty, is often rather high in order to avoid large downtime costs.
In Chapter 4 we introduced a model for two-echelon spare parts networks where we included the use of an emergency warehouse. We included the use of an emergency warehouse to distinguish between a regular replenishment order and emergency re quest at the central warehouse. The emergency warehouse is used to keep a number of spare parts separate which are only used in case of an emergency request. We developed an approximate evaluation procedure to evaluate the performance of this network. This approximate evaluation procedure has not only shown to be accurate but also applicable to many different network structures. Based on the approximate evaluation procedure we investigated the benefit of an emergency warehouse. We show that in particular when local warehouses are far apart from each other, and thus lateral transshipments are not very attractive, there is more benefit in using an emergency warehouse. When local warehouses are located closely to each other and there is plenty of opportunity for lateral transshipments, there is hardly any benefit for the use of such an emergency warehouse.
Finally, in Chapter 5 we developed a generalized approach to determine a classifi cation scheme that enables companies to manage spare parts inventory in a simple way while still getting close-to-optimal. We show that getting close to system-focused inventory control is possible by taking the following four aspects into consideration: the class sizes, number of classes, ranking criteria, and the service level targets per class. We also show that if one of these aspects is not chosen properly, this may lead to poor results.
Bekijk hier het proefschrift.Geef de eerste reactie