Example projects MSc thesis

Here you can find examples of MSc thesis projects of the Operations Research and Logistics group from the recent past. These example projects are to be regarded in terms of themes within ORL in general; you can also check these examples in case you are looking for inspiration for a BSc thesis.

The aim of the examples is to give inspiration about possible thesis projects at ORL, and about future work with a specialization in ORL. Unfold the items below to read more about the presented titles.

If you want to discuss your options for a thesis or internship at ORL, you can contact us via: education.orl@wur.nl. For more information about prerequisites and procedures, see the ORL thesis and internship page.

MSc thesis projects

Sustainability measurement tools for soy and beef supply chains in South America (SALSA)

The objective of the thesis is the investigation of practices aiming to measure the sustainability performance of beef and soy supply chains, and consequently the elaboration of a sustainability measurement tool which would support the decision making process. The thesis includes a desk research which gives an overview of the contemporary approaches and already existing tools for measuring and monitoring sustainability results in the supply chains. The desk research as well as the data provided from the SALSA project (a EU project at the chairgroup related to the sustainability of  soy and beef chains) is the basis for the elaboration of the Excel Tool. Four perspectives are taken into consideration: global warming, energy consumption, water consumption and costs (profit). The four components are constituted by various parameters. All stages of the soy and beef supply chains are measured and the results are represented numerically and graphically. The Excel Tool supports the development of three scenarios: basic, secondary and target. The results of the three scenarios are compared and evaluated. Finally, case studies are applied in order to test the validity of the Excel Tool.

Location Allocation of a Biomass Fermentation Plant – A Case Study in The Netherlands

There is an increasing worldwide demand for raw materials and energy. However, the availability of these resources is decreasing. Instead of reusing the raw materials, there is a trend to use new raw materials for new products. An innovative waste collecting company in the Netherlands has the goal to recycle 100% of the collected waste and use it for other purposes. The aim of the company is to set up a new biomass fermentation plant to make renewable energy. Using GIS data and MIP modelling, a decision support system is set up for selecting the optimal location for a new biomass fermentation plant of this waste collecting company. In order to find the optimal location, economic and social constraints and criteria are taken into account. The model that is used is a location-allocation model, with a cost minimisation objective. In this thesis three different options for a potential location are elaborated.

Minimising internal supply chain costs by optimising inventory location and dock allocation over two neighbouring warehouses

Main focus of this research lies in the field of warehouse management. It is investigated how (un)loading allocation and inventory allocation are optimised within and between two neighbouring warehouses in which fresh fruits and vegetables are distributed. Heuristic approaches of experienced decision makers for efficiently using available floor area are supported by a MILP optimisation model.

The wild boar Supply Chain - The Supply Chain design in which wild boars are processed and sold in Amsterdam

How do we feed the 9 to 10 billion mouths in 2050? Feeding the world's rapidly growing population is one of the greatest challenges of this century. A source of natural animal protein roams Europe's countryside for many years. A professional way of valorising wild boars in order to use this natural resource responsibly has not yet been developed for the German federal state of Rheinland-Pfalz. The purpose of this report is to address the current problem of valorising wild boar meat obtained through hunting. The outcome of this research provides insight into the possibilities, profitability and environmental impact for hunters in Europe who wish to work towards a joint solution for the valorisation of wild boars. The report is structured in three parts. First, the existing and potential Supply Chain stages needed for the design of the Supply Chain are outlined. Second, the relevant Supply Chain strategies and production processes are defined. Third, the Supply Chain stages are modelled using the number of processed wild boars as variable. The simulation model provides insight in the economic and environmental performance of the wild boar Supply Chain. Recommendations are made regarding the management strategy, production process and the mode of transport to be deployed.

Optimizing the commercial production planning in plant micropropagation

Plant micropropagation is the most successful commercial application of plant tissue culture technique. Optimizing the mid-term production planning by matching the production with the given demand poses a challenge both for industry and academia. An optimization-based mixed integer linear programming approach has been developed to solve this problem. It can be concluded with confidence that all the models offer exact optimal solutions for the quantity of plants materials to start with at the best required timing in order to match the given demand and under certain model assumptions. To further enhance the reliability and robustness of these models, it is highly recommended to further validate the model with case studies.

Solving the inventory control problem of perishable food products using Reinforcement Learning

This thesis investigates the possibilities of using machine learning techniques for controlling the inventory of fresh food products, through literature research and simulation modelling. The literature study gave insights into relevant aspects of both inventory control and machine learning. The problem setting was modelled as an Markov Decision process (MDP) and multiple inventory control algorithms were constructed using the Reinforcement learning technique Deep Q-learning (Watkins, 1989). These algorithms varied in the applied solving method and type of artificial neural networks used. The performances of the algorithms were compared to one another as well as to the performance of an order policy(BSP) commonly used by retailers. The algorithm applying double Q-learning and using feedforward neural networks performed best in settings with order costs, also outperforming the BSP. In settings without order costs, none of the algorithms performed better than the BSP. The results look promising however more research is needed to increase performance and make it more stable. The advantage of using reinforcement learning over other methods is that it can more easily be expanded to larger problem settings. Therefore, future research could investigate the use of reinforcement learning to control the inventory of multiple products at once (e.g. product category).

Floricultural network design with process and inventory allocation (DaVinc3i)

Improving the efficiency of a supply chain with network design is being used in a variety of fields. MILP is one of the methods that is frequently applied to network design problems. This research specifically focuses on multiple hub locations with inventory and process allocation problem for the floricultural supply chain, a problem that originates from the DaVinc3i project (running at the chairgroup). The formulation of the model is based on MILP and aims to minimize total costs. A case study is included to prove the applicability of the formulation. The model is programmed in  Excel, using the Solver as well as an artificial analysis. The model needs to be able to find the optimal location of hubs with suitable process and inventory nodes in the floricultural supply chain. The optimal solution in the example case study is given, but in practice, the optimal plan is highly vulnerable to different situations and contexts, such as different transport costs and fixed costs. The scientific contribution of this research can be used in more complex floricultural hub allocation problems as a basic model; it also is applicable for other type of supply chains.


Towards a more responsive supply chain from China to HEMA, by implementing a more frequent ordering strategy

This research took place at the international retailer HEMA. The aim of the research was to find out how the order frequency can be increased, while the costs are minimized, to create a more responsive supply chain. The research focused on a small part of the supply chain, from suppliers in China to the port of loading where orders are loaded on a vessel to be transported to HEMA’s distribution centre in Utrecht. 

By means of literature research and/or expert interviews information was gathered about responsiveness of a supply chain, about HEMA’s supply chain, about the implications of more frequent ordering on various supply chain actors and about transport network scenarios. Three transport network scenarios were calculated and analysed thoroughly. For one of the scenarios a heuristic approach has been developed, partly based on a set partitioning model. The results showed interesting opportunities for HEMA.

Analysis and disaggregation of the forecasting and planning model of De Winter Logistics

In this research is examined in which way the current forecasting- and planning model of De Winter Logistics (DWL) can be expanded, to determine the required capacity in a disaggregated way. DWL is a logistics service provider in the horticultural sector and provides services for the Dutch growers and traders, directly or via the auction. Because of uncertainties about the orders that are placed, it is hard for DWL to make efficient use of vehicles and determine where and when the vehicles need to be deployed.

A disaggregated planning model is designed, by determining the amount of trolleys per transportation service and per time period, to be able to expand the current planning model of DWL in a disaggregated way. The disaggregated planning model translates the forecasted total amount of trolleys into required trucks and trailers per location of DWL by disaggregation of the forecasted total amount of trolleys, respectively into trolleys per trajectory, trolleys per shift, trolleys per time period, required hours, required trips, required vehicles and required trucks and trailers per location of DWL.

Decision Support Modelling for A Local Food System

Global supply chains are nowadays argued to contribute to climate change, waste, environmental degradation. Driven by this concern, Flevoland is considering to localize its food supply chain. Therefore, the aim of this research is to analyse and evaluate local food system performance by developing a decision support model that can address economic and environmental concerns. A literature study was executed to investigate stakeholders’ attitudes toward local food, benefits of local food system (LFS) and distinctive characteristics of LFS. A quantitative model was developed to support decision making in a food distribution system. Besides, a case study was executed to find out optimal solutions for tomato distribution in Amsterdam metropolitan area, and a sensitivity analysis was conducted to better understand the model and gain insights. It can be concluded that the performance of local food distribution system is dependent on context and the extent of localness. Context such as production cost and emission difference, unit transportation cost and alternative transportation modes can affect its performance. As for the extent of localness, transportation distance, consolidation with other products at intermediaries and demand level also have impacts on network performance.

Shared Pickup and Delivery for Last-mile Logistics in Urban Areas

Shared logistics is a phenomenon more and more common in the retail sector in urban areas. Providing pickup service for external customers from other retailers can reduce driver waiting time and CO2 emission, lead to higher efficiency in resource utilization, but this comes at the expense of extra costs. Whether the shared logistics service is profitable for the service provider remains to be quantified. In this thesis, a mathematic model is formulated to study the impact of providing pickup service for external customers on the service provider. The model is solved by a mathematical programming-based heuristic called Relax and Fix (R&F). Computational experiments result on Solomon benchmark and real-life instances show that R&F is generally an effective approach for solving the mixed delivery and pickup problem. Results on real-life instances suggest that shared logistics is promising in improving last-mile logistics’ efficiency in urban areas.

CO2 Emission Reduction in Supply Chains: Case of a Multi-Objective Intermodal Transportation Network Design

Modelling in supply chain management is mainly focused on two aspects: Minimize costs and maximize service. Growing environmental, governmental and economic pressure stimulate the implementation of sustainable development in companies worldwide. Green supply chain management aims to extend the traditional supply chains with activities that minimize environmental impact. This study aims to support companies in developing green supply chain management by including a third aspect in modelling: minimize CO2 emissions.

To support decision making about the implementation of green supply chain management, an environmental assessment framework is constructed. The framework consists of a longlist with CO2 reduction measures, structured by the categories: alternative fuels, fuel efficiency, intermodal transport, logistic network, and logistic alignment. Intermodal transport is quantified in multi-period multi-objective network design. The three objectives consider minimizing transportation costs, transportation time and CO2 emissions. A case study helps to test the model, and results are obtained regarding the influence of an additional transportation mode, barge, on the objectives. The results also consider the required investment in transportation time and transportation costs to achieve a reduction in CO2 emissions. Sensitivity analyses indicate the influence of the distance between the production site and consumer and the demand pattern of those consumers.

Optimizing production planning of mushroom cultivation: towards more sustainable food production chains (TIFN)

This research combines an MSc thesis and an internship project. It is part of a larger project funded by the Top Institute of Food and Nutrition (TIFN) which focuses on valorization of raw material and process efficiency in food supply chains. The objective is to develop a conceptual model that optimizes the production planning of mushroom cultivation in order to eliminate inefficiencies and improve sustainability. Multi-Criteria Decision Making techniques were used to quantify trade-offs between economic (e.g. profit) and environmental (e.g. exergy) indicators of sustainability and to identify inefficiencies in mushroom production. The model was developed and parameterized in close collaboration with the industry. Decisions related to when and how to product in order to meet demand were optimized while sensitivity analysis was conducted to identify opportunities for improvement.

We found out that exergy is a promising indicator of the environmental performance for mushroom production because it accounts for both quantity and quality of energy flows and losses. The cost of compost is a determinant factor for the re-use of compost and the quantities of produced wastes. Re use of compost becomes interesting if alternative technology that focuses on minimizing the costs related to pest and disease becomes available.

New approximations for parameters of replenishment policies of perishable products, by SDP-simulation and regression analysis (TIFN)

FAO reports food waste to be over 30% of the food production. To explore ways of reducing food waste, a DSS will be developed as part of a project of Top Institute Food and Nutrition. Within this DSS a simulation program simulates a food supply chain. One of the factors to simulate is the ordering by retailers. In this thesis approximations are derived for the control parameters in the (S,q,Q) policy and for the performance measures of outdating and the fill rate. A lost sales inventory system of perishable products for a retail environment is investigated, with periodic review, positive lead time, FIFO and LIFO withdrawal policies and a fixed shelf life. Demand is stationary, discrete and stochastic. Objective is minimising average outdating and shortage costs. For a broad Design of Experiments, by means of SDP and simulation these costs are minimised, and the optimal levels for the replenishment parameters and performance measures are derived. Using regression, formulas are estimated for order-up-to level S, minimum order quantity q, maximum order quantity Q, relative outdating, fill rate and determine the cost ratio by a certain fill rate. These formulas are also tested for a policy without q and Q (the order-up-to S policy), and by means of inter- and extrapolation from the DoE. Also for a realistic case from the TIFN project the regression models are tested on their performance. The final approximations are fast and perform very well. These fast and simple approximations can be used in practice by inventory managers, but also in other simulation studies to quickly determine values of replenishment parameters.

Dynamic programming for a path planning purpose; routing an egg gathering robot

Gathering floor eggs by hand in a poultry house is very time consuming and requires a lot of physical effort. Thereby the project “Automation for poultry production” aims for the development of an autonomous egg gathering robot.

This MSc. thesis research contributes to this project by developing a path planning algorithm for the robot. The algorithm should produce an optimal path for gathering as much eggs as possible in the given time frame. Previous research on this topic resulted in a heuristic approach. Comparing the optimal solution with the heuristic enables benchmarking prior results.

The method used for this optimization is called Dynamic Programming (DP) which is widely used in various optimization disciplines, including Operation Research. Its use for path planning purposes is limited.  

After investigating the computational complexity, it is concluded that the optimal solution cannot be computed for a realistic problem size. Due to this restriction another heuristic has been developed which will be compared by simulation with results from prior research.

Research into possible logistical scenarios of the plant auction of FloraHolland; An evaluation of three logistical scenarios of the plant auction of FloraHolland, in terms of cost and throughput time (DaVinc3i)

Three possible logistical scenarios of the supply chain of potted plants of FloraHolland were evaluated based on costs and throughput times. One scenario represents the current practice of the supply chain; the plants are transported from grower to the FloraHolland locations based on sales expectations. The plants are only sold at the location where they are physically present. In the second scenario the plants are transported from the grower to the geographically closest FloraHolland location. The plants can be bought at all locations and therefore interauction transport, after the auction, is necessary. In the third scenario the plants are transported to the FloraHolland locations based on sales expectations and the plants can be bought at all FloraHolland locations, which leads to interauction transport. A conceptual model of the supply chain was created and used as a blueprint for a computer simulation model that was used to simulate the scenarios. This research shows that the first scenario leads to medium costs and the lowest throughput times. The second scenario leads to the lowest costs and the highest throughput times. The third scenario leads to the highest costs and medium throughput times.

Game theory: Coalition viability in a multiple coalition game (TNO)

This thesis is performed in cooperation with TNO and the WUR. TNO reveals research questions on strategic horizontal co-operation in supply chains. Within this concept, a practical case of potential co-operation of five terminals at Maasvlakte 2 (Port of Rotterdam) is seen as a case study. The research question is how concepts from Game Theory can be used to study horizontal co-operation in terms of coalition formation of similar suppliers.  Game theory is a mathematical approach to study strategic behaviour of interdependent stakeholders in a decision making problem. To answer the question, we elaborate the literature on this topic and provide easy to grasp examples that reveal the necessary information to analyse the co-operation question. The case that will help to demonstrate how Game Theory can be used is that of the terminal co-operation. These terminals might consider to form a coalition to bundle their containers for transport per train in order to cope with distribution inefficiencies. The conclusion and recommendations of this study are related to how Game Theory can be used in similar coalition formation problems.

Multicriteria decision making for optimal blending in human nutrition: fuzzy programming approach (WU)

In this project, fuzzy programming approaches have been used to obtain optimal diets for humans. In nutrition, recommended nutrients intakes are given as intervals, but these are not always very strict. For instance, exceeding the upper level for the intake of calcium by one or two milligrams is not forbidden, it is only less desired. This is modelled with a fuzzy programming approach. After reviewing the literature three models are chosen. These models are applied to a small scale problem and analysed. The analysis provides more insight into models and their features. Later on, one of the models is chosen and applied to the nutritional problem. At the end, the results of the Fuzzy Programming model are compared with the outcomes of a previously developed Goal Programming model and the analysis is followed by a discussion and concluding remarks.

Allocation of VALs activities in a Metro Model network in the European floricultural market (DaVinci)

The focus of this research is on the development of a European logistics network for the Dutch floricultural market, based on the concept of the metro model, which connects all supply and demand locations via inbound and outbound hubs. This research evaluates the allocation of a number of VALs (Value Adding Logistics) activities along the supply chain, by optimizing the routes to be taken in these hubs. This thesis is composed by a first part, covered by a literature review, in order to get insights in the Dutch floricultural cluster and in the value adding activities currently performed; the second part is about the allocation of these activities into the metro model, by using Linear Programming, applied to a number of scenarios, which are created according to a case study of the Benelux representing a potential future transportation network. The results provide insights in how a certain VALs allocation impacts CO2 emissions, responsiveness, and transportation costs when products are routed optimally according to minimizing CO2 emissions, or minimizing costs, minimizing total response time. These insights can be used in exploring a suitable design of the network.

A critical assessment of distribution strategies for supplying Dutch florists (FloraHolland/DaVinc3i)

This thesis is written as part of the DaVinc3i project to fill the research gap about the position of the Dutch florist within the floricultural network. The main objective of the research is to analyse the current and “new” distribution strategies for supplying the Dutch florist. The main research question is: “Which distribution strategy is most suitable for supplying Dutch florists?” First the current state is analysed and distribution strategies described in literature are presented. By conducting interviews the perspective of florists on the evaluation of the current supply and requirements for distribution were obtained. KPIs are defined in order to evaluate the defined distribution strategies for various context factors. Four distribution strategies are defined that improve the supply of Dutch florists. The choice depends namely on the purchasing strategy adapted, distance between supplier and florist, and the order volumes. It is concluded that no ‘most suitable’ distribution strategy exists for the supply of florists but for various context factors opportunities arise in reducing transportation costs, the emissions of CO2 and the lead time. In order to define the optimal strategy for a specific set of context factors extensive quantitative research is required.

Pooling Inventory: From a Cash-and-Carry to Pick-Up Strategy (Van der Plas/Davinc3i)

Van der Plas, one of the largest cut-flower exporters in the Netherlands, is constantly looking for new ways to improve business. The goal of this research is to get insight in the effects of pooling inventory on the performance of Van der Plas. The research focuses on Van der Plas’ locations in France. Several key performance indicators are determined and evaluated with respect to Van der Plas’ locations in France. The effect of pooling on the main performance indicator (cost) is analyzed and the optimal strategy for different scenarios is determined. A Literature Study, In-Depth Interviews and a Scenario Analysis are used as methods for this thesis. This research has shown that there are strong indications that pooling inventory can have a positive effect on the performance (reduction of cost) of Van der Plas.

Simple Approximation of Non-stationary Order-up-to-level

This master thesis deals with the quick approximation of determining optimal order-up-to levels for a fixed-life time perishable products under the assumption that the demand is non-stationary. A lost sales inventory system is investigated, with periodic (R, S) order policy. After the analysis of the regression based approximation formulations for perishables as derived for stationary demand in an earlier MSc thesis,  new approximations for non-stationary demand are developed. These new approximations are tested in a simulation model in Matlab. It appears that setting order-up-to levels for non-stationary demand is not straightforward.

Logistics collaboration concepts for Superunie; Case of the AGF chain, focussing on supplier Fossa Eugenia

This thesis is done in combination with purchase organisation Superunie about the rethinking of logistical structures in the retail market. The goal of this MSc thesis is to determine key logistics collaboration concepts (LCCs) for Superunie in such a way that the benefits of collaboration in the supply chain will be clear for the members. This is done with a case on the AGF chain where supplier Fossa Eugenia has provided the analysed data and insight in the chain. The research question that will be answered is ‘

. First, the organisation Superunie is introduced and several theories about collaboration are discussed. The current situation of the AGF chain is sketched, and by having interviews problems and possible bottlenecks in the situation are formulated. The problems are based on not reaching the set norms of key performance indicators (KPIs). The used KPIs in this thesis are: delivery reliability, freshness and logistics costs. For the identifying of the main problems, cause-effect models are built for each of the KPIs. To exclude or improve the main problems, the most preferred scenario for Superunie is made which consists of a combination of the LCCs transport bundling and cross docking.  The combination of that two LCCs is also the answer on the research question of this thesis. To ‘test’ the scenario, a case study is done which focuses on supplier Fossa Eugenia that also has provided the data for the case study.

Assessing opportunities to shorten the order lead time in a demand driven floricultural chain - The case of RoseLife

RoseLife orchestrates the supply chain of horticultural products for several supermarkets. Since they want to reduce the current order lead time, insight into the supply chain is required and opportunities to reduce the order lead time need to be assessed.  By the creation of Event-Process Chain diagrams, insight in the supply chain is gained and waiting times are identified. Based on these, different scenarios are formulated and simulated using (terminating) discrete event simulation, to investigate opportunities to optimize the supply chain, after which the impact of these scenarios on pre-defined performance indicators is measured. It appeared from the diagrams that most waiting time was due to sub-optimal planning, and the results of the simulations indicated that easy lead time reduction could be achieved by postponing the current ordering moments. Sharing preliminary information and adapting the timing of different processes could potentially lead to tremendous order lead time reduction, but this required significant extra effort from several chain actors. Due to the exploratory character of this study, the results should be interpreted carefully and used as a starting point for further in-depth research, since several assumptions were made due to time limitations and demarcation purposes which could have influenced the results.

Decision support modelling for product portfolio optimization and process synthesis in agri-food industry

The goal of this MSc thesis is to get an overview of the mango process network and the corresponding supply chain network. An LP model is created to identify the relation between different tasks within the process network and the corresponding product portfolio. The effect of revenue prices of end products, energy use of processes, waste production and demand restrictions on the final outcome of the model in terms of quantities of produced end products is investigated. Both energy use and profit generation have a large influence on the product portfolio. Furthermore, an MILP model is created depicting the supply chain network and to study the effect of different production locations, transport cost variations and seasonal influences on the division of tasks over the available locations. Especially transport costs have a large influence on the product portfolio and facility openings. No difference in product portfolio was found between separate and integrated process network and supply chain optimization.

 

Assessing the forecasting performance of a Multi-objective calibration method for a bio-economic farm model

Bio-economic farm models aim to simulate farmer’s  behaviour and to evaluate the consequences of future changes. The capacity of the model to reproduce historical decisions is crucial for increasing confidence to model’s results. Existing calibration techniques recover unknown parameters of a non-linear cost functions based on limited datasets of historical decisions. They implicitly assume that farmers make decisions based only on maximizing profitability. In many cases,  farmers decision making involve multiple conflicting objectives. To ensure a good forecasting performance multiple objectives should be taken into account. Non-interactive multi-objective calibration has been proposed to calibrate on historical farm decisions. Even though calibration is guaranteed,  the forecasting capacity of the method must be evaluated before application. Our objective is to assess the forecasting capacity of the multi-objective calibration method. An arable farm model was developed and calibrated based on historical data (year 1999). The Calibrated model was used to forecast changes that occurred in the past ( year 2003). The difference between observed activity levels in 2003 and model’s predictions are used as  indicator of forecasting performance. We conclude that multi-objective calibration of bio-economic farm models can improve forecasting capacity and eliminate arbitrary assumptions associated with existing single-objective calibration techniques.

Towards a more sustainable aquaculture by marine input reduction

In 2014, for the first time in history, the output of aquaculture surpassed that of wild fisheries, showing that fish farming can afford us the capacity to ease the distress off the oceans’ wildlife. However, the environmental pressures that originate from aquaculture’s demand for marine inputs, may have deep ecosystem implications that have to be taken into consideration for the industry to thrive and be sustainable. Marine inputs are important to aquaculture, as well as marine products are important for direct human consumption as they contain the highest concentrations of Essential Fatty Acids (EFA’s), which have a fundamental role in human nutrition. Organizations such as UN and NOAA recently acknowledged the urge to replace marine inputs with environmentally friendlier alternatives to supply these EFA’s. Some alternatives have already been identified, and this study aimed precisely to find out if these alternatives would suffice to replace the marine ingredients commonly found in aquaculture feeds. The study shows that the necessity to include ingredients of marine origin can be reduced significantly by introducing EFA-rich alternatives like insect meal, algae, hempseeds and linseeds. However some EFA’s predominantly found in larger fish, become nutritional bottlenecks that deserve special consideration. Results and analysis are based on a linear programming model. 

Optimization Models in Wind Farm Design

Wind energy can lead to a great socio-economic impact in Europe nowadays, while the installed capacity of wind power is estimated to be doubled by 2030.In order to generate power from the wind, a set of wind turbines is needed. A set of turbines is called a wind farm. The turbines of a wind farm are all connected via cables to the main power station.

For an optimal design of wind farms, accurate models are needed. Computational models on wind farm design vary in the existing literature and they have been developed by engineers and scientists with different backgrounds in order to meet all requirements. This thesis investigates the contribution of Operations Research scientists to these mathematical models. More specifically, an Integer Linear Model from literature has been taken as a starting point. This model aims to minimize the cabling costs for offshore farms. In order to understand the existing model, the structure has been modified and more requirements are considered. Besides the objective to optimize the cable routing, a sub-problem arose to facilitate the model by relaxing a specific type of constraints (i.e. the planarity constraint). The addition of a node in the existing grid turns out to be beneficial for specific cases, taking into account the fixed costs arising from the installation and maintenance.

Reducing the environmental impact of the Dutch diet based on LCA; Replacing meat while retaining its nutritional value

This thesis identifies products of the Dutch diet that have a large environmental impact in terms of Life Cycle Assessment (LCA) indicators and investigates how replacements for these products can be composed with a lower environmental impact, while retaining the nutritional value. The environmental impact is assessed by using the LCA indicators climate change, land use, water use and fossil depletion. Meat products have a major impact in terms of these indicators. Beef and chicken, which respectively have the highest and lowest impact among meat products, are chosen as references. A model is built to compose meat-replacing burgers for both references that have a lower environmental impact and have the same nutritional value. A vegetarian, vegan, insect, and fortification free burger are composed. The computational analysis shows that the environmental impact of the burgers is lower than those of the references for most impact indicators. The burgers have at least the same nutritional value as the references. The vegan burger, which has soy as main ingredient, has the lowest environmental impact of all burgers.

Branch and Bound Algorithms in Greenhouse Climate Control

The horticultural sector has become an increasingly important sector of food production, for which greenhouse climate control plays a vital role in improving its sustainability. One of the methods with which the climate can be controlled is Model Predictive Control, which can be optimized through a branch and bound algorithm. The application of the algorithm in literature is examined and analyzed through small examples, and later extended to greenhouse climate simulation. A comparison is made of different alternative objective functions available inliterature. Subsequently, a modified version of the B&B algorithm is presented, which reduces the number of node evaluations required for optimization. Finally, two alternative algorithms are developed and compared to consider the optimization problem from a discrete to a continuous control space. 

Valorisation of biscuit waste to close the loop in the biscuit supply chain

Due to large proportion of food waste generated along the supply chain, higher priority should be given to an effort in waste reduction. Valorisation of food waste is one of the suggested solutions to benefit the existing waste. Recent data shows that biscuit is one of the top ten food wasted in the Netherlands. Valorising biscuit waste has the potential to improve economic and environmental performance of a supply chain. This study aims to assess different valorisation alternatives found in literatures for biscuit waste. A literature review is performed to study recent literatures that concerns with food chains, economic and environmental indicators, logistical decisions, mathematical programming, and waste valorisation. Three valorisation alternatives are identified i.e. valorisation to produce sugar, succinic acid, and poultry feed compound. A mathematical model is formulated to compare the identified alternatives. The optimization is used to find the best valorisation alternative, and the best location to process biscuit waste. The considered locations include Zaandam, Wageningen and Dordrecht. Results show that poultry feed compound production in Wagenigen gives the best performance. Processing 1000 tons of biscuit waste can generate a gross margin of €157,764.

Treating uncertainty in dynamic sustainable food supply chains

Nowadays, an incredible lack of efficiency is present in Food Supply Chains (FSCs) and consequently one third of all food does not reach the final consumer. Many uncertainty factors influence the performance of FSCs. To tackle this problem a dynamic model is developed in this master thesis in order to enhance the economic and environmental performance of the decision maker. Specifically,  multi-stage stochastic programming concepts are applied to a case study concerning a mushrooms supply chain in the Netherlands. The literature review shows the need for specific research in the field of sustainable FSCs, i.e. the development of dynamic multi-criteria decision making approaches. These types of approaches are hardly covered in literature. The developed dynamic model decomposes decision making into multiple stages and enables to take into account uncertainty factors in the mushrooms supply chain. A comparison of the developed multi-stage model with the deterministic model is discussed. It is concluded that the proposed approach enables to increase the efficiency of the supply chain  by improving both the economic and environmental performance.

Optimization of the location of odor sensors in the neighborhood of pig farms

Odor nuisance by animal production did not receive much attention in the past. However, intensification of livestock farming in the Netherlands causes an increase in odor emission. Therefore, there is legislation to control the amount of odor emission. The expected odor emission is based on measurement studies in which groups of panelists determine the odor level by nose for different types of livestock and farming system. With odor dispersion model V-Stacks and those odor emission factors the expected odor level in the environment of farms is calculated. However, this is an expensive system, and trust of surrounding houses might be affected negatively when calculations are not accurate. To overcome those problems, a pilot has been executed to test the alternative of real time monitoring by sensors. In the ideal situation each house contains such an odor sensor which is too costly. Therefore, First, the conventional method and the aspects of sensor monitoring are explained. Next, a model is proposed to optimize the location of sensors. The behavior of the model is tested by the application of illustrative examples and a case study.

Treating Uncertainty in Dynamic Sustainable Food Supply Chains; A Three-Stage Stochastic Programming approach for the production planning of mushrooms under demand and yield uncertainty

Nowadays, an incredible lack of efficiency is present in Food Supply Chains (FSCs) and consequently one third of all food does not reach the final consumer. Many uncertainty factors influence the performance of FSCs. To tackle this problem a dynamic model is developed in this master thesis in order to enhance the economic and environmental performance of the decision maker. Specifically,  multi-stage stochastic programming concepts are applied to a case study concerning a mushroom supply chain in the Netherlands. The literature review shows the need for specific research in the field of sustainable FSCs, i.e. the development of dynamic multi-criteria decision making approaches. These types of approaches are hardly covered in literature. The developed dynamic model decomposes decision making into multiple stages and enables to take into account uncertainty factors (demand and yields) in the mushrooms supply chain. A comparison of the developed multi-stage models with the deterministic model is discussed. It is concluded that the proposed approach enables to increase the efficiency of the supply chain by improving both the economic and environmental performance.

Modelling and testing FrieslandCampina’s supply chain using a new ES&OP tool

The major goal of this study is to contribute to the testing and implementation of FrieslandCampina’s (FC) new linear programming ES&OP tool. Also, the study investigates the practical and financial feasibility of including price elasticity within the ES&OP tool. In order to test the new tool, the necessary functionalities were listed and tested by executing structured test scripts. Price elasticity, i.e. the impact of produced volumes on the market price, was investigated using literature research in order to define the concept and modelling in order to insert it within the model. Based on of the results of this research, it can be concluded that all necessary functionalities of FC’s ES&OP process are present and properly functioning within the new tool. Concerning price elasticity, this study indicates that the practical feasibility of including price elasticity is high. The financial impact of including price elasticity is also high, but needs further research. More research should be done on the role of competitors when including price elasticity. Next to the research questions, also 2 assignments were commissioned by FC. These assignments contain a description of ES&OP in general and for FC specific including the knowledge regarding the ES&OP model based on manuals.

The Application of Multi-Attribute Decision Making in Ranking Diets

The Dutch National Institute for Public Health and the Environment (RIVM) would like to give an advice for a weekly menu in terms of meat, fish and meat replacers when different weights for safety, health and sustainability are applied. These weights represent the relative importance of the criteria (safety, health and sustainability). Seven different weekly menus were composed by RIVM. The goal of this research is to find the most suitable method for ranking these menus. The methods applied were analytic hierarchy process (AHP), best-worst method (BWM), weighted sum model (WSM), weighted product model (WPM), complex proportional assessment (COPRAS) and  technique for order of preference by similarity to ideal solution (TOPSIS). The results are very similar for all applied methods. In order to obtain the weights for the criteria, the AHP method would be most suitable, since only a few pairwise comparisons had to be made. For the evaluation of the alternatives a method that would avoid rank reversal would be most suitable. Therefore the ideal COPRAS method would be most appropriate for evaluating the alternatives, since this method requires the lowest number of calculation steps and is most reliable.

Designing a framework for the supplier evaluation and selection at Pactics (Cambodia)

In this thesis a framework is designed for the supplier evaluation and selection at Pactics. A multi-criteria decision-making technique is used, with elements of the Analytic Hierarchy Process to calculate weights and elements of the PROMETHEE method to score supplier performances. The framework is applied to evaluate Pactics’ current and potential RF0026 fabric suppliers.

Firstly, different divisions of Pactics have indicated which criteria they found important for the supplier evaluation and selection at Pactics, based on a longlist of criteria retrieved from the literature. Price, Delivery & Service, Quantity Flexibility and Suppliers’ SCM Capabilities are selected as economic sub-criteria, Environmental Documents, Environmental Friendly Products and Environmental Management Capabilities as environmental sub-criteria and Meeting Customer Requirements, Long-term Relationship Potential and Certification as social sub-criteria. Also, prerequisites are determined which the suppliers must meet in order to participate in the evaluation.

Thereafter, the relative importance of the (sub-)criteria are calculated with a pairwise comparison. Subsequently, with the use of indicators a questionnaire is designed to measure the performance of the suppliers on these (sub-)criteria. Their answers are scored and entered into the model. Finally, the model has provided a ranking of the suppliers with a score for each supplier between 0% and 100%.

Exploring the Impacts of Using Different Sets of Nutrients in the DEA Diet Models: Towards a Healthier and Acceptable Diet

Incorporating acceptability is a challenge for diet modeling. Data Envelopment Analysis (DEA) is proposed as an alternative approach to calculate healthier diets without explicitly introducing acceptability constraints. In the DEA diet models, evaluated diets are benchmarked based on a set of less-is-better (undesirable) and more-is-better (desirable) nutrients, and the choice of nutrients can affect the results of the analysis. This study explored the impacts of using different nutrient sets obtained from four selected Diet Quality indexes (i.e., the NRF6.3, NRD9.3, NRF11.3, and NRD11.4) on the model solution. Besides, the difference amongthe DEA diets models (i.e., the IO-DEA, OO-DEA, and ADD-DEA) was analyzed. These DEA diet models were used to improve the diets for the Czech Republic.

Generally, the more nutrients included in the DEA diet models, the more efficient diets, and less room forimprovement. The type of nutrients involved also influenced this association. The comparison results addressed the importance of selecting a suitable set of nutrients, which selection should be on a case-by-case base.The finding implicated that the DEA diet models can be used to design specific diets and the model output can provide policymakers with a direction to promote public health. 

Gain Allocation in Biomass Supply Chains by Using Collaborative Game Theory ; A case study of sugar beet supply chain

Biomass Supply Chains (BSCs) have been proposed as a promising alternative energy source. Current tools for optimizing decisions in BSC mainly focus on maximizing the total profit of the chain, while that may result in an unfair allocation of cooperative profit. Plenty of Cooperative Game Theory (CGT) methods are applied to fairly distribute the profit in the Supply Chain (SC), although which method is appropriate to apply to BSC is disputed. This study aims to find the most appropriate CGT methods to fairly allocate profit in the biomass supply chain. According to the literature study on BSC and CGT methods, four CGT methods (Nash Bargaining (NB), Shapley Value (SV), Least Core (LC), and Separable & Non-Separable Method (SM)) are selected for further analysis in this research. A Dutch sugar beet BSC is addressed as a case study for testing the applicability of those methods. The result shows that NB, SV, LC and SM methods can all be applied in BSC under certain conditions. In conclusion, NB can be used for two-actors BSC, while SV, SM and LC methods are appropriate to apply in BSC with only three to five actors. In this research, we did not find any suitable CGT method to directly allocate profit for a large and complex BSC which has more than five actors and massive players are included per actor.

Machine learning to reduce food waste - Can neural networks improve food demand forecasting?

This thesis is an exploratory research on the potential of machine learning based forecasting techniques to forecast demand more accurately to reduce food waste. The problem considered is a supervised regression problem. Based on the results of a literature study, artificial neural networks is selected as an interesting algorithm to research the potential of machine learning based food demand forecasting. After performing a literature study, different features potentially having an impact on consumer food demand were found. Considered in this thesis are events, weather, seasonality, price, promotions, and previous sales of the same product. The neural network is trained and optimised based on, respectively, the Backpropagation and a gradient descent optimiser with momentum (RMSprop) algorithms. The optimal number of hidden layers and whether regularisation (dropout) should be applied is determined based on trial and error. The optimal number of neurons in each hidden layer is determined based on a Random search. Both a feedforward neural network with only Dense layers and a neural network of which the first layer is a recurrent Long Short-Term Memory layer are considered. Multiple linear regression is the traditional forecasting method used as a benchmark to compare the results of the neural networks. From the results, it can be seen that feedforward neural networks with only Dense layers outperformed the benchmark in all cases. Adding a recurrent Long Short-Term Memory layer improved forecasting accuracy of some products, outperforming both the benchmark and the feedforward neural networks with only Dense layers. Based on the results it can be concluded that neural networks, and machine learning, have potential when it comes to improving demand forecasting to reduce food waste (and improve service level) at the retailer.

Gain Allocation in Biomass Supply Chains by Using Collaborative Game Theory ; A case study of sugar beet supply chain

Biomass Supply Chains (BSCs) have been proposed as a promising alternative energy source. Current tools for optimizing decisions in BSC mainly focus on maximizing the total profit of the chain, while that may result in an unfair allocation of cooperative profit. Plenty of Cooperative Game Theory (CGT) methods are applied to fairly distribute the profit in the Supply Chain (SC), although which method is appropriate to apply to BSC is disputed. This study aims to find the most appropriate CGT methods to fairly allocate profit in the biomass supply chain. According to the literature study on BSC and CGT methods, four CGT methods (Nash Bargaining (NB), Shapley Value (SV), Least Core (LC), and Separable & Non-Separable Method (SM)) are selected for further analysis in this research. A Dutch sugar beet BSC is addressed as a case study for testing the applicability of those methods. The result shows that NB, SV, LC and SM methods can all be applied in BSC under certain conditions. In conclusion, NB can be used for two-actors BSC, while SV, SM and LC methods are appropriate to apply in BSC with only three to five actors. In this research, we did not find any suitable CGT method to directly allocate profit for a large and complex BSC which has more than five actors and massive players are included per actor.

Utilizing heathland topsoil - Logistic implications for the Dutch province Gelderland

Heathlands are habitats of high cultural and natural value, but their management is cost intensive. Valorizing by-products bears the potential to partially outbalance those costs. One by-product from heathland management is topsoil. In the province Gelderland, a maximum quantity of 26,000 tons of heathlands topsoil could be available in one year. This by-product can be utilized in circular agriculture.
A model was developed to get insight in important cost drivers and the effects of several scenarios. Considered performance indicators were quality of use, transport distance, disposal costs, income, and uncertainty. It was concluded, that applying topsoil on local farmland should be the preferred way to utilize the by-product. This utilization could be stimulated by increasing the permitted radius for the application of topsoil on farmland from one to five kilometers. Topsoil processed into compost can be used as natural fertilizer. Currently, the high gate fee forms an obstacle. It is one of the main cost factors for heathland topsoil disposal. Omitting the gate fee at composting facilities could help to decrease those costs. Including tree nurseries does not bare the expected potential. Main advantages could be the reduction of costs and transport distances, and more options of disposal.

Identification of logistical cost reduction opportunities based on analysis of customer order data - A case in the fast-moving consumer goods market

Logistics can be a very expensive part of business operations. Especially nowadays, where the increased strain on infrastructure and the environment has sparked many businesses to re-evaluate their logistics to improve the cost and resource efficiency of it. This thesis aims to identify opportunities for logistical cost reduction on order picking and tier-2 logistics. Because of the heterogeneity of the customers, it is not practical to have a single definition of optimal, since it is known upfront that for many customers this optimal situation will never be achieved. In this thesis, a framework was developed that shows how optimality for both cost components can be defined while keeping each customer’s unique requirements in mind. Customer orders are analyzed do calculate the efficiency of the current ordering patterns and an improved (optimal) situation that a customer should be able to transition to easily. This framework was then implemented in a case in the fast-moving consumer goods market. Some initial results from implementing the optimization in the case central to this thesis are inconclusive yet seem to indicate that some positive effect became apparent shortly after implementation.

Bottom-up modelling of urban food-systems and their environmental impacts

Case study of the environmental impacts of Almere's food consumption

Delivery time optimization at Agrifirm Feed B.V. ; Design of a logistical concept to reduce the delivery time

Agrifirm wants to reduce this delivery time for cattle feed to one day. They want to reduce this delivery time because of customer satisfaction, efficiency and competitive advantage. The objective of this research is to find out how the workload variation of the incoming orders can be reduced so Agrifirm can optimize the delivery time (with as starting point total costs and efficiency).

Determining order-up-to level by approximating the waste ratio

This research aims to determine the order-up-to level (S) under a waste constraint, which will be implemented by developing a fast and good approximation to estimate the waste ratio. We study the two approximations from literature and drive two of our approximations. To further improve the results, we include those four approximations and other inputs and variables as independent variables into the linear regression model. Besides, feature engineering transforms the variables to become linear. Furthermore, we split our database into 15 subsets by three clustering methods. The approximations and models are tested by inventory simulation database. The final approximation is very fast and performs very well, with an average R2 and mean square error (MSE) of 0.85 and 0.0021. This approximation has limitations on input parameters but we have other approximations for a wide range of input parameters.