Choose an application

• Reference Manager
• EndNote
• RefWorks (Direct export to RefWorks)

This master dissertation introduces the capacitated dual index policy which extends the traditional dual sourcing literature by including capacity constraints of the transport modes into the sourcing cost function. Based on the dual index policy, a prominent policy in the dual sourcing literature, the capacitated dual index policy introduces an extra step in the form of rounding the orders up or down to achieve full truckloads. This work demonstrates the excellent performance of the capacitated dual index policy, because the rounding makes sure that the inventory cost is significantly lower than the dual index policy. The sensitivity analysis of this policy points out that the holding cost that a company faces is the parameter that has the most influence on the total cost, when the capacitated dual index policy is applied. This is because the holding cost impacts the inventory cost, the cost of ordering, and the cost of transportation all in a negative way- meaning that the different costs grow bigger when the holding cost grows.

Choose an application

• Reference Manager
• EndNote
• RefWorks (Direct export to RefWorks)

This thesis provides research regarding predictive analytics in service maintenance. More specific, it investigates the impact of model misspecification on performance of the condition-based maintenance policy. It analyses the impact of using different stochastic models to model component degradation. A simulation model is constructed and statistical analyses are performed to find the impact of model misspecification on performance of the condition-based maintenance policy.

Choose an application

• Reference Manager
• EndNote
• RefWorks (Direct export to RefWorks)

Environmental concerns raise the need for more efficiency and sustainability in the freight transportation sector. For this purpose, the Physical Internet is introduced, which aims to connect logistics networks into one hyperconnected supernetwork. To transport freight over such an integrated network, the innovative concept of synchromodality is presented. Synchromodality is defined by the usage of multiple modalities when planning shipments, where real-time switching between transportation modes is possible. In this thesis, a synchromodal planning model is developed that constructs optimal transportation routes in a multimodal network with stochastic transit times, formulated as a mixed-integer linear programming problem. To cope with this transit time stochasticity, transportation routes are adapted in accordance to real-time information about the transit time outcomes. Hence, the model determines the optimal decision to be taken in a terminal depending on the realized transit time. A numerical study demonstrates the potential advantages of real-time planning adaptation in terms of costs, service quality and environmental impact.

Choose an application

• Reference Manager
• EndNote
• RefWorks (Direct export to RefWorks)

Artificial intelligence, robots, virtual reality, ... objects and technological applications are evolving at high speed in many sectors including leisure, transport, education and industry. Machine learning has already shown high development potential with self-driving vehicles, object recognition, search engines and personalized interactions between machines and humans, but what is the current scope of machine learning in the industry? In the age of industry 4.0, this master thesis focuses on an overview of the impact of machine learning on supply chain management. First, a general explanation of the concepts of supply chain management and machine learning is given, intended for an audience that is not necessarily familiar with the subject. Then, several existing machine learning applications in supply chains are presented. A general description of these implementations, their benefits and concrete examples of companies and start-ups that are developing them are given.

Choose an application

• Reference Manager
• EndNote
• RefWorks (Direct export to RefWorks)

This thesis addresses the problem of the IEEE 2015 PHM Data Challenge, where one must predict plant breakdown of unknown plants in advance. This paper proposes three supervised machine learning algorithms based on classification approaches, namely decision trees, random forests and artificial neural networks as possible solutions. Two cost-sensitive variants of the decision tree approach and one costsensitive variant of the random forest approach and the artificial neural network approach are also modeled. These variants apply a similar methodology compared to the standard classifiers but are optimized using a cost function. This results in predictive maintenance models that attempt to avoid premature failure misclassifications. The forecasts, which are based on historical input data, support managers in the decision-making process to optimize the planning of maintenance actions. The structure of the thesis can be divided into four stages. The first stage consists of a literature review on the existing research that has been done on the use of machine learning in predictive maintenance. This is followed by the preprocessing of raw, unordered data of the problem under consideration into useful information. Descriptive statistics are performed on the cleaned data in the third stage. In the last stage, the predictive maintenance models are built. A combination of RapidMiner, Microsoft Excel and Groovy software is used to preprocess the data and to model the classifiers. Results show that the number of machines within a plant does not affect the prediction accuracy of the models and that the decision tree approach provides the most accurate results.