The relationship between energy consumption and baggage handling systems (BHSs) has not been widely studied. In this study, this relationship is explored by considering BHS configurations - the designed arrangement of devices. The BHS consists of seven main processes: drop-off, transportation from drop-off to screening, hold baggage screening (HBS), transportation from screening to sortation, sortation, early baggage storage (EBS), and make-up. Transportation appears to cause a high share of the system's energy consumption, as it is a part of and connects several processes. An equation-based model is developed to estimate the energy consumption of BHSs with varying configurations. The model only contains parameters available at the BHS's conceptual design phase and is based on the BHS of a midsize airport in Scandinavia. It includes a series of formulas for each process, which depend on the structure of the process elements. The study proved a twofold effect of an airport’s BHS configuration on the overall energy consumption. Firstly, more process elements structured in series result in bags travelling a longer distance on the conveyor within the BHS, thereby increasing energy usage. Secondly, the parameter values for transportation impact energy consumption notably. The research suggested that more process elements structured in series decrease the overlap of device usage, which in turn reduces energy consumption.

The drilling industry faces the challenge of moving away from conventional maintenance strategies to maximise operational efficiency. This thesis introduces a holistic approach to land drilling rig Predictive Maintenance (PdM), integrating component maintenance into system-level decision-making. By taking the operations of the whole system into account, PdM can be applied more effectively, and the uptime of a drilling rig can be improved. A hierarchical Multi-Agent System (MAS) framework is proposed, employing two layers of agents for the core equipment of a drilling rig and one centralised system agent. Middle-level agents perform component diagnostics, prognostics, and operational state classification. The system agent uses expert reasoning for integrated maintenance scheduling. As a proof of concept, the proposed model is partially developed and validated, using readily available data and additional condition monitoring.
In the field research phase of this thesis, valve vibration signals were gathered from a mud pump during a geothermal drilling project in Delft. From these signals, dimensionless features were extracted and used in a Fuzzy Logic (FL) classifier combined with Weibull Accelerated Failure Time (WAFT) modelling for dynamic Remaining Useful Life (RUL) prediction. In the case study, where synthetic valve failure is induced in historical operational data, the model demonstrated its ability to predict these failures on time. Maintenance actions were suggested by the model when components reached 93% of their theoretical lifetime, preventing excessive maintenance and minimising disruption to drilling operations through integrated scheduling.

This thesis presents a data-driven design approach for emission reduction using bunker delivery notes (BDNs) to help support the revised IMO strategy to achieve net-zero greenhouse gas emissions by international shipping close to 2050. This research supports the Horizon Europe’s Digital Twin for Green Shipping (DT4GS) project which focuses on the development of digital twins (DTs). Part of this project involves the development of a DT-supported method for the design and retrofit of ships. DTs are a promising approach for supporting maritime decarbonization efforts due to their simulation and big data handling capability. Despite the abundance of shipping data and growing digitalization, the potential of using ship operational data for decarbonization efforts remains not fully exploited. A data-driven method such as a DT could fill this gap. However, as DTs, by definition, require real-time connection between a physical entity and the digital representation, developing a true DT for new-build alternatively fueled ship designs remains a challenge. This research thus starts by looking into retrofitting using data from existing ships.

A design framework is proposed to construct digital models to support a DT for retrofitting purpose. The proposed framework is tested on a case-study using a 300-meter bulk carrier. Since January 2019, operational ship data is collected through BDNs, a mandatory data collection method for ships of 5000 GT and above, adopted by the IMO. Constructing a DT based on BDNs is considered to be convenient as it provides a solid source of operational data in the future.

First, the available data from the BDNs is preprocessed using an adopted framework based on data science literature. The resulting 5,678 data points are used for the construction of a model representing the bulk carrier and a model representing the green ship technologies part. A fuel consumption model is constructed to represent the bulk carrier. It utilizes a gray-box modeling approach, consisting of a white-box resistance model and a black-box artificial neural network. Both models incorporate environmental-dependent inputs. The investigated green ship technologies for the potential retrofit are represented by various wind-assisted ship propulsion (WASP) systems, namely a towing kite, a DynaRig sail, and a Flettner rotor. These systems are modeled using a white-box modeling approach, together with available wind data. Using an adopted integration framework, based on the propeller-engine matching procedure, both representations are combined into one green ship digital model.

An environmental assessment is performed using the IMO's EEXI and CII assessment tools, respectively evaluating the design and operational aspects of the potential retrofit. Additionally, a financial assessment is conducted using the payback period. Results showed the design implications and emissions reduction potential of implementing such systems which will guide the retrofit decision by the ship's owner.

This research focuses on the complexities of integrating routing decisions, inventory management and pressure control in the context of the cryogenic logistics of Liquefied Natural Gas (LNG). This research aims to address these complexities by developing a planning method that considers the transportation and nitrogen cooling costs. The study evaluates the developed planning method through a case study at Rolande, which is a company that operates LNG and Bio-LNG refuelling stations in the Netherlands, Belgium and Germany.

This study considers three refrigeration methods for controlling the pressure level at refuelling stations. These are nitrogen cooling, offload cooling and logistic trailer cooling. The identified key performance indicators that are used to evaluate the develop planning method are: Total Costs, Total Transportation Costs, Total Nitrogen Cooling Costs and Cost per Kilogram. Furthermore, the key constraints and variables for the integrated control of the LNG logistics are identified.

The planning methodology was developed through three steps. First a Full Mixed-Integer Linear Programming (MILP) model was developed. Second, the Full-MILP model was refined to a Simplified MILP model. This Simplified MILP model was improved with the rolling horizon approach and a pre-solve process. The rolling horizon approach divided the planning horizon into smaller manageable time blocks. The pre-solve process identified the critical stations for each day to reduce the number of nodes in the network.

The proposed planning methodology was experimented in the case study that involved a network of 19 refuelling stations in the Netherlands and Belgium. The sensitivity analysis indicated that the model was sensitive to the vehicle capacity. Therefore, the pre-solve process was extended with determining the supply of LNG. The case study results revealed that the model is able to solve a seven-day planning horizon, while maintain the inventory and pressure levels within the specified bounds. However, the model can become computationally complex for days with high number of critical stations and cool vehicles.

This research contributes to inland LNG logistics by addressing the integration of routing, inventory and pressure management. Future studies should focus on a comparative analysis with heuristics, experimenting the feasibility and computational time with soft inventory and pressure bounds, and model a non-linear offload cooling effect to improve the realism of the model.

With Schiphol Airport's flight numbers growing, working assets are essential to ensure on-time processes. The Passenger Boarding Bridge (PBB) is a critical asset in the airport's turnaround process. By ensuring that the asset is working properly, the operational processes can run efficiently. Currently, improving the reliability of the PBB when in use happens after the fault has occurred. With this maintenance strategy, the PBB data is not used to predict the future health state of the PBB. Literature shows that the PBB can be classified as a multi-component system. Research in the predictive maintenance strategy of multi-component systems is still in an early phase. Research until now is more theoretical than practical, and an investigation into applying theoretical knowledge in practice is needed. With the upcoming developments of Industry 4.0, a Cyber-Physical System (CPS) architecture is proposed for a multi-component system. This architecture has been applied to the PBB to develop and use a predictive maintenance strategy for this system. Based on the implementation of a simulation model, the output showed that the proposed CPS architecture enabled the development of a predictive maintenance strategy for the PBB. With this strategy, proactive maintenance is planned while the system's reliability is held on a preset level to ensure a working asset during in-time use.

The parcel transport industry utilizes automated parcel sorting systems to effectively manage incoming parcels. These systems typically comprise multiple infeed conveyors that converge onto a single merge conveyor for parcel merging. However, the conventional First-Come-First-Serve merging method results in reduced utilization of the merge conveyor and throughput. To address this, a novel approach is proposed in this study, combining a dynamic programming-based control algorithm with a maximum heap algorithm. This approach aims to enhance the utilization of the merge conveyor and achieve a balanced load among the infeeds. To optimize utilization, an additional imaginary segment is introduced ahead of the merge conveyor, synchronously moving with the merge conveyor's speed. This segment is intelligently populated with parcels using dynamic programming, eliminating unnecessary empty spaces. Results demonstrate that the dynamic programming approach outperforms other techniques in terms of utilization. This research presents an initial investigation into implementing dynamic programming as a control algorithm to improve merge conveyor utilization in the parcel transport industry.

The demand for accurate and effective cool chains has been expected to increase, especially for pharmaceuticals in the air freight industry. However, several problems and challenges remain such as cool chain breaks and cool storage capacity constraints while there are rarely routine systems in place for consistent insight into the operational quality of such systems. Besides, the concept of a DT has received increasing attention in the literature, while a multitude of applications have been found in fresh cool chains. Therefore, the DT concept has been applied to a pharmaceutical cool chain for operational quality improvement. Firstly, a novel operational quality metric has been proposed: the OCCE. Consequently, a cool chain at KLM Cargo has been studied and modelled by means of the DES technique in order to derive a virtual representation. Consequently, the DT concept has been applied through the implementation of a decision support module for cool storage decision-making. The model implementation has shown that the cool chain operational quality has been improved while the average exposure of freight has decreased.

Ship-to-Ship (STS) cargo transfers can significantly improve the efficiency of offshore installations while reducing their associated costs. However, the added complexity of cargo transfers involving ship-mounted cranes at sea poses significant challenges. To address this, crane manufacturers have introduced Relative Heave Compensation (RHC) systems to assist crane operators. Nevertheless, concerns have emerged among installation vessel companies regarding the relative heave measurement systems, which rely on Motion Reference Unit (MRU) sensors placed on both ships. Given that many supplier vessels lack the required sensors and the communication protocols pose bureaucratic hurdles, this research focuses on proposing novel, feeder-independent relative heave measuring solutions. This work introduces two unexplored sensor units, namely the LiDAR-MRU and RADAR-MRU. In the absence of actual sensor data, a simulation workflow using Simulink and Unreal Engine (UE) is presented to generate synthetic data. After, four distinct measurement solutions are developed, implemented, and evaluated using the simulated data in MATLAB. The implemented solutions estimate relative heave distance and speed with a Mean Absolute Error (MAE) ranging from 0.3-5.2% and 0.8-4.3% of the reference's maximum amplitude, and processing times from 1.8-91.4 [ms]. The results underscore their potential for field implementation. However, future validation is needed with actual sensor data.

Automation in tomato greenhouses is lacking compared with some other crops. The efficiency and consistency of greenhouse operations can be boosted by implementing automation. Therefore, a new concept that facilitates the robotisation of crop-handling tasks is investigated. This concept combines the existing technology of a movable-gutter system with the cultivation concept of a limited-truss tomato. The combination of these two techniques has not been previously explored. For this paper, a generic conceptual design framework was developed which contains a simulation model and an evaluation protocol. In the evaluation protocol a sensitivity analysis and an optimal dimension analysis is performed. Case studies are conducted to evaluate the simulated performance, to examine the capacity of the transport system and crop-handling tasks and to assess the robustness. These case studies prove that the developed framework is generic and all-encompassing. The framework has proven its worth by eliminating erroneous design choices before significant resources were spent. The new concept is promising, as it facilitates automation and has a competitive simulated production. Therefore, it is advised to move on to the next design phase, which should mainly focus on validating the crop-growth model and the preliminary design of subsystems as the transport system and components as the movable-gutter itself. Furthermore, a cost analysis and economic assessment should be conducted.

Bulk handling grabs are often used for unloading bulk materials such as iron ore and coal from bulk carrying ships. To improve and compare virtual simulations of new grabs, accurate real life kinematic data of grabs is required. Moreover kinematic data can potentially provide insight in current use of grabs, which can be interesting information for both manufacturer and customer of the grab. The possibilities towards implementation of a sensor system on bulk handling grabs are investigated by selecting a suitable sensor and measuring kinematics and operations of imitated grab movements. This research focuses on investigating kinematics of multi rope grabs with dual scoops. A sensor selection procedure is followed resulting in the selection of an Inertial Measurement Unit (IMU) sensor with Global Navigation Satellite System (GNSS) receiver. In several experiments the sensor data is compared with reference data of imitated grab movements assumed as ground truth. When using multiple IMU’s (one on each grab scoop) providing orientation data, the opening angle can be determined. The results of the experiments indicate that the position data provided by the sensor can be used for classification of activities, but less for high precision movement trajectory reconstruction. Position and orientation data of imitated grab movements can be interpreted to generate performance indicators that describe grab operations. Experiments showed that performance indicators of imitated grab movements can be identified accurate up to an error of two seconds when comparing known performance indicators and sensor data analysis. When a high level of accuracy of the kinematic data is reached, performance indicators describing grab operations can be determined more accurately.

Container traffic worldwide has shown a growing trend during the last years. Also, the number of containers handled per call has increased for many terminals. This poses challenges to existing terminals for more efficient operations and higher productivity. Automation is a way some terminals have tried to achieve this. Most existing automated container terminals were greenfield projects, but brownfield automation is gaining momentum.
Literature on brownfield automation of container terminals is limited. The PIANC MarCom WG Report n° 208 – 2021, Planning for Automation of Container Terminals is the most recent and complete guideline for automation projects, but it is mainly focused on greenfield projects. Challenges for automating brownfield terminals are different than for greenfield ones. All of this translates into little guidance being available for future brownfield automation projects. Therefore, the experiences and lessons learnt from historical brownfield automation of container terminals are a valuable source of information to guide future projects. The aim of this thesis is then to present a characterisation of historic container terminals converted, or in the process of being converted, to some degree of automated operations.
An empirical research methodology was defined to gather data on drivers for automation, challenges, observed benefits, drawbacks, and solutions adopted by brownfield automated terminals. A questionnaire was used for this purpose. Also, satellite images of the conversion process were analysed to observe implementation strategies. Additionally, terminal throughput, equipment, yard size and quay length were determined through a desk study. Data was analysed through categorization. Trends regarding the type of automation, i.e., semi-automation (automated yard equipment) or full automation (automated yard + automated transport between the stacks and the quay), and regarding the type of yard equipment adopted were determined.
For the historic conversions analysed, it was concluded that the main drivers for automation are operational cost reductions, labour shortage, productivity increase, reliability/safety, and capacity. But these drivers are case specific. No benefits can be assumed a priori from automation. Objectives for automating should be defined by each terminal and solutions to achieve them proposed. The main challenges were continuity of operations during the conversion, adaptation to new operations, labour relations, and communication systems. Most terminals adopted semi-automated solutions, due to fewer labour issues expected and improved vessel productivity compared to full automation, among other reasons. It was observed that terminals with yards smaller than 10 [ha] chose aRTGs as their stacking equipment, while terminals with larger yards chose either an aRMG or the automated version of their yard equipment before automating.
Three implementation strategies were observed: greenfield-like, phased, and big bang approaches. The first one refers to developing the automated yard in a new location, with little disruption of the operations of the old yard. A phased approach means that the implementation is divided in phases, and operations of the old yard must be disrupted from phase 1. The big bang approach was observed only for conversions from straddle carriers (SC) to auto SC. In these cases, a small site is used to test the technology. When the testing is finished, the terminal is closed for 2-3 days, and the technology is rolled out onto the entire yard or a large portion of it.

In this thesis, a simulation cost model is presented that can be used to assess the cost-effectiveness of a selection of PdM programs. The proposed model combines a discrete event simulation (DES) model with a cost model. It is specifically designed to analyse the PdM programs for the vacuum compressor pumps at Duyvis. All the PdM programs proposed will work with vibration analysis to detect bearing failure. The simulation model will focus on the sample frequency and detect-ability of a PdM program. The sample frequency is related to maturity levels, and the detect-ability is related to detecting the different stages of bearing damage. The goal is to use the proposed simulation cost model to assess a selection of PdM programs on their overall cost and reliability.

Due to the short shelf-life of flowers (circa 7 days) good inventory management is required to minimize disposals. The trade-off between supplying enough to meet demand yet not too much to prevent expiration is a delicate matter. However, in the current situation uncertainty about inventory levels due to inadequate information and conventional replenishment strategies lead to a situation that does not perform efficiently. This paper demonstrates how an RFID overhead system can be designed that has 98\% reading performance for inventory. Furthermore is showed how this access to better information can already lead to a reduction between 30-50\%, depending on the type of distribution the demand follows. If there are obvious seasonal trends, then the better information in combination with optimal control can have a reduction of 90\% in disposed products with the same availability of products. These reductions are significantly sufficient to conclude that the benefits comfortably break-even with the investments from the inventory system.

Although e-groceries is nothing new, the scale at which it is currently growing in the Netherlands has not been seen before. Over the past four years, the market share of online grocery shopping showed an average growth of over 20% per year. While it was already challenging to keep up with this high demand, it became increasingly more difficult due to COVID-19. In fear of getting infected, some consumers no longer visit the supermarkets, and instead, order their groceries online, further increasing the demand for e-groceries. One of the challenges caused by the high demand is handling enough customer orders in the online grocery fulfilment centres. These fulfilment centres can be described as “supermarkets” in which groceries are collected by operators instead of the consumer. After the groceries have been collected, they are either delivered to the consumer, or they can be collected at a pick-up point. While multiple strategies are described in literature that allow for the handling of more customer orders, most of them have only been applied to warehouses and distribution centres. Due to the significantly different operations of fulfilment centres, it is unknown whether the same strategies can be applied here. This thesis examines which strategies are appliable to online grocery fulfilment centres and how much additional customer orders can be handled. The identification of strategies was done through a case study at AH Online’s fulfilment centre in Rotterdam. A discrete-event simulation model was then used to evaluate the strategies. The output of this thesis is an overview of four strategies including their performance on timesaving, order picking productivity, and of course, the additional number of customer orders that can be handled.

This report presents insights in the feasibility of large-scale multi-AGV based parcel sortation systems. A simulation model is created to look at the influence of different aspects of such an Automated Guided Vehicles system. The model uses several design parameters and KPIs to determine the level of influence. Important aspects of this research are: the charging process of the AGVs, influence on pick-up points (parcel collection) and drop-off points (parcel destinations) and multiple types of AGVs on the same map. The AGVs will be modelled by an A* algorithm and uses grid-routing (either X or Y -movement at the same time). These kind of AGVs are controlled by a centralized robot control system. The task allocation the AGVs is done by a market- auction-based algorithm.

The first test set consists of a map of 1000m2 to operate at a throughput of at least 1600 parcels/hour. Here the placement of the pick-up points on all sides of the map is more efficient than on either one or two sides. Though the placement of two sides is further used as it is a more probable solution due to the use of large truck-bays and smaller vans in terms of a parcel sortation center. A linearity is found when adding more AGVs to the system. This is in between 5-50 AGVs for the 6 and 8 pick-up points with a ratio of 44.62 parcels/hour per additional AGV. Linearity is decreased due to more queueing and re-routing by congestions to 39.86 parcels/hour per additional AGV for 10 and 12 pick-up points. The charging process reduces the throughput with at least 7.54 % on a 16 hour simulation run. This is with a ratio of 5.71 AGVs per charging point. Using a higher ratio leads to AGVs shutting down due to low battery charge states.

Another test case is analysed with a 3000m2 area and the demand of achieving a throughput of at least 6000 parcels/hour. This is achieved by using 170 AGVs on a width of 75m and a length of 40m area. Even a small analysis is done in the use of 6 kg and 32 kg AGVs on the same working area. From which, a throughput of 1000 parcels/hour is gained by using 15 6 kg AGVs and 25 32 kg AGVs. The throughput is lower compared to using 40 32 kg AGVs: 1400 parcels/hour. Loss in efficiency could be considered compensated by the reduction in terms of AGV purchase costs.

This research has no optimization process, more potential is possible if an optimization is done with a more elaborate/complete (design) plan. This optimization could lead to more efficiency and more insights of an AGV parcel sortation system. Finally the possibilities of multiple types of AGVs on one map can be further investigated, with additional task allocation schemes or a more elaborate research in the pick-up points process.

Currently a new small and automated dairy factory for the onsite production of cheese at farms is under devel-opment. This new factory aims at the production of cheese without any human interaction, based on the (raw) milk from cows who are milked with a milking robot. At the moment only a small test factory exists, in which human interaction is still present. Of the new factory it was unknown what configuration of processing stations and transportation vehicles would be required to produce cheese. To find the optimal configuration of this new automated dairy factory, first a data analysis was used to determine the characteristics of the milk input of the factory. Next the present test factory was analyzed and a model representing the automated dairy factory was developed. The model was used to identify the opti-mal configuration of the dairy factory. Four different farm groups were researched, per farm group one opti-mal configuration of the dairy factory was identified.

Diesel-powered Automated Guided Vehicles (AGV) are currently deployed for container transport at terminal’s water side. Looking at the trends in formal legislation and market developments towards sustainability, these diesel AGVs are most likely to become outdated. Although battery-electric AGVs are an emerging, zero-emission alternative, there are serious technical, operational and financial questions regarding their implementation at brownfield terminals operating 24/7. Taking the northern side of the ECT Delta Terminal in the port of Rotterdam as a case, the operational and financial feasibility of replacing diesel AGVs by a battery-electric AGV fleet have been evaluated by means of a functional design of the AGV charging process, simulation study and total costs of ownership analysis. The results indicate that battery-electric AGVs opportunity plug-in charged at the automated stacking cranes’ transfer points prove to be an operationally and financially feasible alternative to diesel AGVs: operationally under the condition that a sufficient amount of AGVs, charging power and plug-in chargers are installed and financially under the condition that the increasing trend in diesel price and decreasing trend in electricity price will continue in the near future. As environmental legislation for heavy-duty vehicles becomes more stringent while there is a decreasing trend in electricity prices, battery-electric AGVs are most likely to become profitable for deployment at brownfield container terminals. Therefore, this study’s findings could pave the way for terminal operators to replace their environmentally unfriendly diesel AGVs by zero-emission vehicles, potentially becoming the key force increasing the global penetration rate of electric vehicles in heavy-duty industry.

In recent years, technologies such as artificial intelligence have enabled computer systems to provide decision support for maintenance planning. Using advanced analytics to determine optimal maintenance actions and moments is called prescriptive maintenance. Multiple prescriptive maintenance models have been proposed for different industries. However, none of these models or frameworks
were designed for use in the FMCG industry. This industry seems to be lagging compared to the oil & gas, process and transport industries for example. The goal of this research is to define a framework that can achieve prescriptive maintenance at a FMCG production plant. This research focuses on the decision making for maintenance on multi-component systems using technologies such as machine learning. The proposed framework uses an intelligent agent-based approach to optimize maintenance planning on a component and system level. The agents act on both levels and negotiate to find an optimal system level result. A life-cycle cost approach was used, where all impact of maintenance
was translated to costs and thereafter minimized. As prescriptive maintenance aims to increase operational efficiency and reduce costs by artificial decision making, the effect of the framework on these KPIs was simulated and compared in a FMCG case study. This case study was done using a discretetime simulation of a cooling compressor system in a beverage production facility. It was found that, given a high-quality prediction of the remaining useful life, this prescriptive maintenance framework can increase a systems operational efficiency and reduce maintenance costs. However, the largest improvement could be seen in the overall production time of the system, as maintenance was planned in such way that loss of functionality was prevented.