- Lehrstuhl
- News
- Lehre
- Forschung
- Team
- Stellenangebote
Für mehr Informationen zur Forschung gehen sie bitte zur englischsprachigen Version dieser Seite.
Robert Bosch GmbH is one of the world’s leading semiconductor manufacturers for sensors used, e.g., in smartphones and the automotive industry. The production processes are subject to stochastic and dynamic effects. In particular, yield and process availability change over time due to process improvements. Digitalized production systems increase the availability of data and allow for a systematic analysis of different types of variability.
The student is expected to gather and analyze available shop floor data of different production steps. The data serves as a basis for an in-depth analysis of different sources of variability. The student has to apply statistical methods to gain insights into the magnitude and interdependencies between the different sources of variability. Therefore, knowledge in statistics is recommended (e.g. CC 502 Applied Econometrics). Knowledge in MySQL is advantageous for the collection of data.
Students are encouraged to start their master thesis at Robert Bosch GmbH with an internship,
one to two month prior to the thesis. In this period, the master student will be employed as an intern at Robert Bosch GmbH in Reutlingen. The start date of the thesis is flexible.
Please contact Dr. Justus Arne Schwarz (schwarz(at)bwl.uni-mannheim.de) for further information.
The effectiveness of the integration between maintenance and production scheduling has already been shown in numerous studies in the literature. All the studies though, focus only on the minimization of scheduling objectives in the form of make-span, total completion time, maximum lateness… This makes sense under the assumption that the costs of producing the jobs on the machine are constant throughout the scheduling horizon. In practice though, some production costs (e.g. electricity costs) may highly varies over the scheduling horizon and can have great impact on the quality of the schedule and therefore on the performance of the plant. In these cases then, a tradeoff between the two objectives (scheduling objective and production costs) have to be found; maintenance related decision can play in these situations an even higher role in the effective operations of the plant, e.g. if maintenance activities are planned during high production costs peaks.
The objective of the thesis is multiple: first, the candidate will formulate the problem as a MILP and perform a sensitivity analysis to highlight the impact of the variable production costs on the optimal solution. Afterwards, few heuristics solution will be developed to effectively address the integrated scheduling problem.
Prerequisites: Knowledge of a modeling language for Mixed Integer Programming (e.g., OPM 662).
Literature:
Many business processes are analyzed with queueing models. Before such queueing systems reach a steady state, they pass a transient phase. In this phase, the performance of the queueing system changes over time even though all parameters are constant. This phase can be observed in many real-world applications in which the jobs waiting in the queue are frequently cleared, e.g., at airport check-in counters, supermarkets, or production systems starting without inventory. A solid understanding of the transient phase is of managerial importance because the performance may substantially deviate from its steady-state behavior. Hence, the analysis of the transient phase serves as a basis for the support of design and control decisions in queueing systems before they reach a steady state. Moreover, a sound understanding of the length of the transient period is also a prerequisite for a justified application of steady-state models, which dominate in the scientific literature.
The goal of the thesis is to provide a literature review of business processes and decision models related to the transient phase. Based on this a numerical investigation of drivers for the length of the transient period, e.g., the size of the system and the degree of variability, has to be conducted. Existing implementation of analytical transient results and a discrete-event simulation can be used. A comparison with existing approximations of the relaxation time and the development of an own approximation based on regression analysis are expected.
Requirements: sound knowledge of queueing theory, e.g., from OPM 661.
Literature:
Personaleinsatzplanung in Wohnheimen für Menschen mit Behinderung: Optimierungsmodelle und heuristische Lösungsansätze
Die LEBENSHILFE Dillenburg e.V. ist Träger verschiedener Einrichtungen der Behindertenhilfe im nördlichen Lahn-Dill-Kreis. Neben stationären und ambulanten Wohnangeboten unterhalten wir mehrere Werkstätten für Menschen mit Behinderungen, einen familienentlastenden Dienst sowie ein Kinderzentrum. Insgesamt sind bei uns ca. 400 Mitarbeiterinnen und Mitarbeiter beschäftigt, die etwa 1000 Menschen betreuen.
Unsere Wohnheime für Menschen mit Behinderungen bieten ein Zuhause für Menschen mit zumeist hohem Hilfebedarf. Menschen die Unterstützung in der Bewältigung des Alltags in Form von Anleitung, Begleitung und Entwicklungsförderung benötigen. In unseren Wohnheimen leben mehrere Menschen in einem Haus in unterschiedlichen Wohngruppen zusammen. Sie arbeiten zumeist in einem unserer Betriebe, auf einem Außenarbeitsplatz des ersten Arbeitsmarktes, sind Rentner oder sind aufgrund des hohen Hilfebedarfs in einer Tagesbetreuung.
Für viele Menschen mit Behinderung ist es sinnvoll, in einem Wohnheim zu wohnen und nicht in ambulante Wohnformen überzugehen. Grund dafür ist vor allem ein hoher Hilfebedarf. Das Betreuungspersonal ist ständig vor Ort und kann begleiten und anleiten. Eine besondere Bedeutung hat deshalb das Betreuungspersonal und damit die Personaleinsatzplanung. Sie muss den hochgradig volatilen Personalbedarf decken, Mitarbeiter mit verschiedenen Arbeitsverträgen verplanen, individuelle Einsatzwünsche berücksichtigen und komplexe rechtliche Nebenbedingungen beachten.
Die Masterarbeit soll die problemspezifischen Annahmen und Ziele herausarbeiten und in die wissenschaftliche Literatur einordnen. Ein ganzzahliges Optimierungsmodell soll aufgestellt und in GAMS (o.ä.) implementiert werden. Basierend auf zu erhebenden Realdaten, sollen Testrechnungen zur Lösbarkeit mit Standardsolvern durchgeführt und geeignete heuristische Lösungsverfahren entwickelt, sowie getestet werden. Sensitivitätsanalysen unterstützen die Ausarbeitung von generellen Einsichten in das Optimierungsproblem.
Voraussetzungen: Kenntnis einer Modellierungssprache für ganzzahlige Probleme (z.B. GAMS) wie sie z.B. in OPM 662 erlernt wird.
Performance evaluation of time-dependent queueing systems with the Surrogate Distribution Approximation: A numerical comparison
Several real world queueing systems feature time-dependent parameters. Examples are time-dependent traffic volume at call centers and IT systems and time-dependent truck arrival rates at sea ports.
Explicit analytical solutions for the performance evaluation exist only for special cases. Hence, several approximation approaches have been developed. One of them is the Surrogate Distribution Approximation (SDA). The basic idea of the approach is to numerically solve the moment differential equations (MDEs). A surrogate distribution is used to close MDEs. Based on this idea different versions of the SDA have been developed. They relax assumptions on the characteristics of the queuing system and improve the approximation quality.
The goal of the thesis is to provide a comprehensive numerical comparison of approximation results by the different versions of the SDA, other analytical approximations and discrete-event simulation. The implementation of the SDA is a key part of the thesis whereas existing implementations of other approximation methods and a discrete-event simulation can be used. As an initial step, meaningful test instance have to be created which capture the variety of operating environments of time-dependent queues. Based on the numerical study the strength and weakness of the SDA approach have to be identified.
Prerequisites: Sound knowledge of queueing theory, e.g., from OPM 661; programming skills or willingness to acquire basic programming skills, e.g., C++ or Java.
Literature:
Production scheduling with flexible maintenance
Classical scheduling problems assume that machines are always available. Although this can be frequently considered a valid assumption, in some cases machines need to be stopped every now and then to perform a maintenance activity and to be restored to a fully working condition. During a maintenance activity machines are taken oline and cannot be used for production; this has clearly a disruptive impact on the plant productivity, in particular if due dates are considered as some production jobs has to be postponed to leave space for the required maintenance activity. The problem of scheduling production and maintenance simultaneously can be modeled as a MILP.
The objective of the thesis is to study the problem of simultaneous production and maintenance scheduling on a single machine in which jobs are required to be completed within a given due date. The machine is assumed to be capable of working for a given amount of time after which a maintenance activity has to be planned. The candidate will start with a brief literature review about the problem, the MILP formulations and the solution approaches presented in literature. The MILP formulations will then have to be implemented in GAMS and critically compared by mean of an extended numerical analysis. For big problem instances, the limits of the standard softwares will have to be analyzed numerically. To solve such problems, advanced optimization methods and/
Analysis and optimization of queuing models with virtual waiting
Waiting on hold influences the costumer’s perceived service. Service centres need to decide whether they increase the number of agents to improve the waiting time or to accept longer waiting times. The start-up virtualQ offers a new service that maintains a caller’s position in the line without letting the costumer wait in the line. The call centre notifies the caller to call in, when it is his turn. Different strategies to influence when users are notified can be implemented and thus effect the service quality.
Such a virtual queue allows the service centre to control when costumers will retry. In phases of high load, the call centre can shift calls to low load periods to improve the waiting time for costumers in the high load phase. However, uncertainties due to the costumer behaviour have to be considered in this queuing model (e.g. how quickly after the notification a costumer will call in).
The goal of this thesis is a structured analysis of existing queuing models with virtual waiting. The student is expected to provide a description of the decision model and to analyse the performance of service centres using real-world data. Relevant levers to optimize the model shall be identified in a sensitivity analysis. If applicable the model shall be developed further and can be evaluated in a live environment.
Prerequisites: Basic knowledge of queueing theory (e.g., OPM 661).
Appointment rules and robust optimization for outpatient appointment systems
Appointment systems are used in many service systems (e.g. healthcare systems) to manage access to service providers. They improve productivity and match demand with capacity by smoothing demand. Appointment scheduling problems arise in such systems, which are to set appointed arrival times to minimize the expected performance measures of the system (e.g. patients' waiting time, practitioner's idle time, and overtime).
Developing such an appointment schedule for outpatient clinics is challenging because it requires considering several environmental factors, e.g., patients’ no-show and walk-in patients. No-shows affect the total performance of the system by increasing the practitioner’s idle times. On the other hand, walk-ins increases the patients' waiting time and possibly also the overtime of the practitioner. In practice, several appointment rules are used to design an appointment system. An appointment rule determines the basic template of the appointment system by specifying the number of patients scheduled to each appointment slot (i.e., block size) and the length of appointment intervals. Apart from appointment rules, scenario-based stochastic programming (i.e., robust optimization) is another possible way to deal with the appointment scheduling problem.
The student is expected to give an overview of appointment rules for appointment systems with no-shows. Based on this overview, a simulation model that evaluates several of those appointment rules shall be developed. A numerical study shall investigate the effect of the environmental factors on the performance of the appointment rules. In addition, a robust optimization method should be developed and tested. The performance shall be compared to that of the appointment rules.
Prerequisites: Basic knowledge of stochastic models (e.g., OPM 661), knowledge of a programming language or willingness to acquire basic programming skills (e.g., Java, C#).
Operational robustness for planned and unforeseen changes under stochastic and time-dependent conditions
Corning is one of the world’s leading innovators in materials science. Its Environmental Technologies Division is a leading supplier of advanced cellular ceramic substrates and particulate filters for the world’s major manufacturers of gasoline and diesel engines. Due to the introduction of Euro 6c and China 6 emission standards, Corning has announced its $100 million investment in DuraTrap® GC Gasoline-Particulate Filter (GPF) in its Kaiserslautern and Shanghai facilities to address growing demand from automakers for filters to comply with new standards.
Due to the introduction of the GPF products in 2015/2016 in addition to the existing products in Corning GmbH, there is a need to analyze and restructure the work flow process within the Department Lab/
The master thesis is expected to provide a structred analysis and deciption of the current processes, including the identification of sources of variability, and occuring decision problems at the Department Labs/
Students are encouraged to start their stay at Corning with an internship, two to four month prior to the thesis. In this period, the master student will be employed as an intern in Corning GmbH. The option of an internship is offered only if the student agrees on writing a thesis at Corning GmbH. The intended start date of the thesis is 1st of March, 2017.
Integration of RTN and heuristics for production scheduling
The problem of short-term scheduling of batch plants has been extensively studied in recent years. The most common process network representations are the Resource-Task-Network (RTN) and the State-Task-Network (STN) representations. In addition, there has been growing effort toward the alignment of the industrial sector and the power grid through demand side management (DSM). When integrating DSM into RTN formulations there are many objective functions that could be considered, for example: minimization of makespan, minimization of tardiness, minimization of penalties on violating load commitment or minimization of electricity or total cost. This project will explore the use of multi-objective optimization in RTN-based scheduling formulations considering energy aspects and will explore the tradeoffs and relationships between different objective functions. Regardless of which cost-based or combined objective function is chosen, integrating DSM into RTN formulations increases problem size considerably to the point where the resulting model quickly becomes intractable. That being said if the resulting model is tractable, it is able to provide a globally optimal solution. Conversely, heuristic-based methods have very fast solution times and scale well with increasing problem size, however, these methods provide no indication of optimality. This project will also look into the integration of heuristic-based and mathematical-programming-based scheduling algorithms in a way that greatly improves the solution time of RTN based formulations while making minimal sacrifices to optimality.
Numerical analysis of the impact of variability in time-dependent service systems
Balancing customer satisfaction, costs and stakeholders' interests is a major challenge for contemporary businesses. This has provoked extensive research dealing with the operational challenges in services in a post-industrial economy.
Queuing theory is often used in the performance evaluation and optimization of call centers. They are an important example of service systems not only because their nature allows for significant data collection for research, but also because of their role in the economy as a major employer. In the master’s thesis call centers are modeled as a time-dependent queuing network with blocking, abandonments, and generally distributed processing and abandonment times. The time dependency is an important assumption, as call center operations often incur significant fluctuations over the day, which stationary models fail to account for. The student has to investigate the sensitivity of staffing requirements to changes in the coefficient of variation of the service and abandonment processes. To do so, the DIS-MOL approach of Liu and Whitt 2012 is implemented in Java, including necessary extensions to an existent queuing tool.
Truck arrival management at production plants: Assigning trucks to multiple truck docks
(in cooperation with INFORM GmbH)
The supply of various production processes is delivered by trucks. This Master’s thesis considers the related real-world problem, in which each arriving truck needs to be scheduled to visit one or more truck docks around the production facility in order to replenish materials required for the production processes inside. Examples for multiple truck dock visits are deliveries of the same kind of parts that are used in different (variant) production processes or deliveries that comprise different kinds of parts.
For the considered planning problem, it shall be assumed that the sequence of required truck dock visits is predefined and that each truck features a preferred arrival time. The decision is to assign each truck to one or multiple particular truck docks in specific time slots such that the truck’s visiting sequence is fulfilled. The objective is to minimize the total times that the trucks have to spend on-site as well as the deviation of the actual start of the truck’s first truck dock visit from the preferred arrival time. This deterministic optimization problem is referred to as Truck Dock Assignment Problem (TDAP). In addition, traffic conditions and capacity shortages at the truck docks may lead to deviations from the scheduled visit times at the truck docks. Therefore, the robustness of the deterministic decision is a key issue in a stochastic environment.
The student models and solves the TDAP with GAMS. Furthermore, a simulation study analyzes the impact of stochastic arrivals and capacities on the robustness of the deterministic solution.
Allocation of check-in counters considering time-varying passenger arrivals
Check-in counters are the first point of contact for many passengers at airports. Airline managers and airport authorities face the decision how many counters should be opened at which time to account for the time-varying passenger arrival process. Optimal plans have to balance the trade-off between personnel costs or long waiting times and customers that cannot check-in on time. Parlar and Sharafali (2008) propose an analytical model to dynamical allocate check-in counters based on a stochastic dynamic programming formulation. The goal of the thesis is to gain insights regarding the model of Parlar and Sharafali (2008) and its limitations. Therefore, an implementation of the approach is required which allows conducting a comprehensive numerical study. Sensitivity analyses with respect to key parameters and an extension of the approach to a generalized queueing system are expected to build the basis for a critical discussion of the model assumptions.Requirements for this topic are a sound knowledge of queueing theory, e.g., from OPM 661 and programming skills, e.g., C++ or Java.
Literature:
Numerical Analysis of Optimal Product Rollover Strategies
The relevance of product rollovers, i.e., the period in which an old product variant is phased out and a new product variant is introduced to the market, has gained importance within recent years. The main reason is, that product life cycles have shortened and new product variants are entering the market more frequently. Thus, the transition period becomes more relevant. An analytical model for the profit maximization by optimization of pricing and timing decisions for the old and new product is provided by Lim and Tang (2016).
The goal of this thesis is to provide insights about the model of Lim and Tang (2016) and its limitations. Therefore, in a first step, an implementation of the model is required. A sensitivity analysis with respect to key parameters of the model has to be conducted. A critical evaluation of the modeling assumptions builds the basis for extensions to the model. These extensions have to be incorporated in the existing model. One direction for such an extension is the consideration of contingency plans to react to uncertain events that cause changes in parameters.
Optimization of a finite number of service rates using Markov chains
Many queueing systems in real life show state-dependent service rates. Wireless controllers for example adapt their service rates to the number of arriving data packets. If only few data packets are waiting to be transmitted, the transmission speed is decreased and thus energy consumption is reduced. If many data packets need to be transferred, the service rate and thus transmission speed is increased. Other examples can be found in call centers that perform up-selling.
In such queueing systems, the decision on the service rate depends on the present queue length. The goal is to balance the cost of inventory and cost of service effort. Service costs depend on the chosen service rate and holding costs are incurred as a function of the queue length in front of the system. If the number of possible service rates is finite, all possible policies for the decision problem at hand can be enumerated. A policy defines the chosen service rate depending on the state of the system. For each policy, a Markov process can be developed which allows to determine the state probabilities and performance measures by solving a system of linear equations.
The student is expected to build an evaluation method that analyzes a given policy using the theory of Markov processes. This evaluation method should then be integrated in a heuristic that optimizes a state-dependent queueing model by balancing service and holding cost.
Requirements for this topic are a basic knowledge of queueing models and queueing theory (e.g., OPM 661) and basic knowledge of a programming language (e.g. Java or C++).
Literature:
Literature overview on Markov decision processes to optimize state-dependent service rates
Many queueing systems in real life show state-dependent service rates. Wireless controllers for example adapt their service rates to the number of arriving data packets. If only few data packets are waiting to be transmitted, the transmission speed is decreased and thus energy consumption is reduced. If many data packets need to be transferred, the service rate and thus transmission speed is increased. Other examples can be found in call centers that perform up-selling.
In such queueing systems, the decision on the service rate depends on the present queue length. The goal is to balance the cost of inventory and cost of service effort. Service costs depend on the chosen service rate and holding costs are incurred as a function of the queue length in front of the system. In literature such queueing models are frequently analyzed with Markov decision processes. Markov decision processes are Markov processes in which a decision is made in each stage. They are used to model decision making in situations where the outcome is both random and depends on a decision.
The student is expected to present a literature review on Markov decision processes for the analysis of queueing systems with state-dependent service rates. Therefore, the general modeling framework of Markov decision processes should be presented. Using this framework, relevant Markov decision processes from literature should be discussed and compared in detail.
Requirements for this topic are a basic knowledge of queueing models and queueing theory (e.g., OPM 661).
Literature:
Optimizing queueing systems with state-dependent service rates and a method of truncated holding costs
In many situations queueing systems have state-dependent service rates. An example are wireless controllers that adapt the transmission speed to the number of arriving data packets: If only few data packets arrive, the transmission speed is decreased and thus energy consumption is reduced. Other examples can be found in call centers that perform up-selling.
In such queueing systems, the decision on the service rate depends on the present queue length. The goal is to balance the cost of inventory and cost of service effort. Service costs depend on the chosen service rate and holding costs are incurred as a function of the queue length in front of the system. In most papers, the service rate is chosen from an infinite interval of possible service rates. In George & Harrison (2001), a dynamic programming approach is presented that uses truncated holding costs to compute an optimal state-dependent policy.
The student is expected to critically assess the approach of George and Harrison. Therefore, the decision problem at hand should be described properly and a brief literature review should highlight the differences to other solution approaches in literature. The key part of the thesis is the implementation of the approach. A sensitivity analysis should investigate the impact of different choices of the cost functions.
A simulation-optimization approach for the shift scheduling problem in a call center with abandonments and retrials
The shift scheduling problem assigns a number of personnel to pre-defined shifts in order to minimize staffing costs and satisfy a service level constraint. Especially in service industries e.g. call centers, staffing costs are a main aspect of the overall cost structure and thus economically highly relevant.
In order to evaluate performance measures and service levels in call centers with abandonments and retrials the method of simulation is rising in popularity. Explicit analytic queuing formulations and approximations do only exist for either simplified or very specific cases. The combination of simulation and optimization offers the possibility to combine the higher accuracy of stochastic models with the aim to produce a cost-optimal staffing schedule. With simulation being very computationally expensive, heuristic solutions to limit the feasible schedules to a manageable number have been developed.
A heuristic solution algorithm is developed and implemented in Java, including necessary extensions to an existent simulation tool. The performance of the algorithm is compared to a scenario generation approach of Helber and Henken (2010). Possible approaches are a cutting plane algorithm, local search or other improvement heuristics.
Scheduling Problems in Finished Vehicle Processing – A Comparison of MIP-based Solution Approaches
(In cooperation with INFORM GmbH) Car terminals are the linking elements in automotive distribution chains via which transports between plants, ports and dealerships are commonly organized. Apart from connecting two transportation steps, these terminals typically offer maintenance, repair or modification services for the vehicles that are passing through.
The master’s thesis considers the real-world problem of scheduling service orders to be done either at the terminal or at external sites. The basic scheduling problem can be considered as a variant of the job shop scheduling problem, where a job consists of all service operations to be done on a vehicle. A job may require multiple locations visits, where each location can only process a limited number of operations at the same time. The main objective is to minimize the total tardiness costs with respect to job priorities.
Against the backdrop of the SyncroTessVDS software system developed by the INFORM GmbH, the master’s thesis describes the considered scheduling problem by its specific objectives and constraints and formulate the according mathematical model in MIP approach. Due to the complexity of large-size problems in practice, the thesis proposes MIP-based heuristic approaches in a second step, such as aggregation and disaggregation techniques and decomposition. The model for implementation is developed that simplifies the original problem, while covering key aspects of location capacities, precedence constraints and minimization of tardiness. Numerical study is conducted in GAMS in a deterministic approach.
Besides the model formulation for the considered problem, this thesis is expected to explore potentials and limitations of MIP-based solution approaches by comparing key performance measures and analyze in terms of their robustness.
Line balancing with stochastic task completion times
The line balancing problem assigns tasks to stations in order to optimize the production process of a flow line. It solves the underlying problem of which task should be assigned to which station.
Recent studies have focused on incorporating uncertainty with regard to the task completion times. The goal of this thesis is to implement a sampling approach to account for stochastic task completion times. Each work piece can be represented as a vector containing its specific completion times for each task. The use of sampling necessitates a modified objective function. A service level will be set that determines the percentage of work pieces with binding task completion times. The total number of stations is then minimized for all binding work pieces and a given cycle time. A numerical study will be performed in order to analyze the impact of variability on the system performance.
Since the problem is NP-hard, a solution algorithm is developed to solve the problem within a reasonable time frame. Possible approaches are a Benders decomposition or a problem-specific branch and bound approach. The performance of the solution algorithm is evaluated with regard to its computation time for problem instances of different sizes.
Performance evaluation of time-dependent queueing systems with the Pointwise Stationary Fluid Flow Approximation: A numerical comparison
Several real-world queueing systems feature time-dependent parameters. Examples are time-dependent traffic volume at call centers and IT systems and time-dependent truck arrival rates at sea ports.
Explicit analytical solutions for the performance evaluation exist only for special cases. Hence, several approximation approaches have been developed. One of them is the Pointwise stationary fluid flow approximation (PSFFA). The basic idea of the approach is the combination of a deterministic fluid approximation with steady-state queueing formulas to integrate stochasticity. Based on this idea different versions of the PSFFA have been developed. They relax assumptions on the characteristics of the queuing system and improve the approximation quality.
The goal of the thesis is to provide a comprehensive numerical comparison of approximation results by the different versions of the PSFFA, other analytical approximations and discrete-event simulation. The implementation of the PSFFA is a key part of the thesis whereas existing implementations of other approximation methods and a discrete-event simulation can be used. As an initial step, meaningful test instance have to be created which capture the variety of operating environments of time-dependent queues. Based on the numerical study the strength and weakness of the PSFFA approach have to be identified.
Shadow prices in degenerated problems
In many applications in production and economics, one is not only interested in optimizing profit, revenue, lead times and so on, but also in the shadow prices. They give information about how the objective will change, if some parameters change. For example, this may be useful, if having to decide about increasing capacity.
Another example, where shadow prices play a crucial role is revenue management. Revenue management is mainly used in airline industry, where the customers are divided in different classes according to their willingness to pay. If low value customers arrive before high value customers, it may be not optimal to accept all requests. One common way to answer the question which customers to accept is to use shadow prices. They represent the opportunity costs of selling one seat now. A request is accepted, if the shadow price (in this context also called bid prices) is smaller than the price, the customer is willing to pay.
In order to compute shadow prices, one often solves a linear program and uses the dual variables as shadow prices approximating the marginal values of the corresponding resources. This is a justified approach, if the underlying problem is non-degenerate. In real life problems this is often times not the case. In practice and even in the scientific literature, this fact is widely ignored.
Simulation of dynamic end-to-end semiconductor production systems
(in cooperation with Infineon Technologies AG)
End-to-end semiconductor supply chains comprise players from different industries, e.g. automotive original equipment manufacturers (OEMs), tier one component suppliers, and tier two semiconductor manufacturers. These players differ with regard to production technologies. Hence, diverse production and supply chain management methodologies are used. The flow production of OEMs has short cycle times and is usually managed with methodologies like Just In Time, Just In Sequence, and Kanban. These operations management methods are not directly applicable to semiconductor manufacturing companies. Operation Curve Management has proven to be an efficient approach for their production management, which hast to cope with short product life cycles, cycle times of up to 6 months, and the curial factor of quickly gaining high yields in a stochastic production environment.
To coordinate production and material flows in the whole supply chain, the above described production management methodologies of automotive and semiconductor companies have to be aligned, rather than locally optimized per stage. This task becomes in particular challenging as in most of the mentioned supply chains semiconductor and automotive companies are not linked directly, but via the tier one supplier. This third player has to understand and to deal with both production management methodologies.
Workforce scheduling in queueing systems with batch arrivals
Airports and authorities continuously look for opportunities to reduce costs while keeping security and service at a high level. While many costs are fixed, one major lever is the deployment ofworkforce. The staffing and rostering of customs and immigration officers is complex as available information changes over time with limited possibilities to react. Flight plans and the approximate number of passengers of a flight are known in advance and can be taken into account when schedules and rosters are created. Contrary, flights may be delayed or arrive early, which is known only a few minutes or hours before the actual landing. The stochasticity of the actual arrival times has an impact on the actual, time-dependent demand for officers. The fact that arrivals are stochastic and the possibility to shift officers from one booth to another should be taken into account when planning the deployment of officers. In this thesis, the student will give an introduction to the planning problem and will formulate a model that takes stochasticity into account. The student will further implement an optimal solution approach and analyze its performance.
Appointment systems and truck arrival management: A literature review and research agenda
Innovative demand management mechanisms have emerged in the past decade to improve the performance of truck handling operations in different areas of application along the supply chain, such as production plants, distribution centers, seaport container terminals, and air cargo terminals. In particular, the implementation of terminal appointment systems has received increasing attention in practice and research. Such a system's overall objective is to smooth demand by shifting truck arrivals from peak to off-peak periods. Appointment systems are prevalent in a variety of other areas, in particular applied for health care and service operations. The goal of the thesis is to provide a comprehensive literature review, combining the streams of truck arrival management and appointment systems. The student analyzes and classifies the design characteristics of the various fields of application and their corresponding quantitative analysis and planning approaches in the academic literature in order to identify research gaps for potential future research efforts.
Production ramp up - A literature survey
The production ramp up is the phase from manufacturing the first unit until the desired production level is met. An adequate planning and management of this crucial phase is of high economic importance. Delays and resulting lost sales as well as quality issues can become serious threads to a successful product lunch. The literature provides performance measures and models for different aspects of the production ramp up. The goal of the thesis is to provide a comprehensive survey of the existing literature. Therefore, a classification scheme has to be developed in order to structure the literature by the focus of the papers and the used methodologies. Moreover, the gained overview is the starting point for the identification of uncovered fields regarding production ramp ups.
Simultaneous requirement planning and scheduling for queuing systems with time-dependent demand
The typical objective of workforce scheduling is to minimize costs while maintaining a minimum service level. This is commonly achieved by a two-step approach: first, the period-dependent workforce requirements are determined. Secondly, the allocation of the workforce to shifts is optimized. However, the so-called hierarchical approach has serious shortcomings: among the most important ones are the assumptions of a steady state and period-independence. In dynamic environments, such as call centers, these assumptions do not hold and lead to understaffing. This Master's thesis presents alternative approaches that integrate demand planning and workforce scheduling. Therefore the models are implemented in Java and evaluated regarding their preciseness, flexibility, underlying assumptions and computation times.
Energy cost minimal production planning using resource task network models
(in cooperation with ABB AG)
As electricity prices are time-dependent, electricity-intensive industries can reduce their costs significantly by shifting their production timing to low-cost periods. This thesis deals with the optimization of the melt shop schedule of a steel plant regarding energy costs. The melt shop process is modeled as a Resource Task Network, a general modeling technique consisting of resources and tasks. Based on an existing discrete-time Resource-Task-Network model, the thesis aims to develop a heuristic approach that is able to handle larger problem sizes. Furthermore, a more accurate continuous-time model shall be developed for the same problem.
Fairness in operations management
Operational decision making in organizations has an impact on other actors (employees, business partners, public, etc.) these organizations interact with. Nevertheless, these actors may have different perceptions on the fairness of the decision making process, and may (re-)act in consequence. The scope of this thesis is many-fold. On the one hand, it aims to provide a framework of the concept of fairness and its several dimensions. On the other hand, it deals with evaluating how can fairness be measured/
A comparative study of capacitated lot‐sizing and scheduling models with sequence‐dependent setup times and costs
Lot sizing is an essential optimization problem in operational production planning. Changing the product type produced on a resource usually requires a setup operation and incurs setup times and costs. On the other hand, the inventory to meet the demand has to be kept low because it incurs holding costs. Lot sizing optimization models strive to find the best trade-off between setup and holding costs. The complexity of this problem increases, if the setup times and costs are sequence-dependent. In this thesis, an extensive overview of optimization models for the lot sizing problem with sequence-dependent setup costs is given. Furthermore, a selection of these models is implemented in a mathematical programming language and compared in a numerical study.
Task assignment for check-in counters at airports
Ground handling agencies provide airlines with the personnel required for passengers check-in process at airports. It is, thus, important for the ground handling agencies to efficiently manage their workforce to fulfill the service level required by the different airlines. This thesis deals with the assignment of the agencies' employees to specific flights within a single day planning. These assignments have to be done considering the availability and qualifications of the employees, as well as the flight schedules, location of the check-in counter for each flight, etc. The goal is to obtain an optimization model to bring about the assignments that maximize personnel utilization.
Robust planning in the German emergency medical services: A sampling approach
The emergency medical service's (EMS) aim is to provide a timely first aid to emergency patients. Hence, it is important to plan resources accordingly to save human life. However, the resource planning is influenced by different aspects such as legal and personal restrictions and demand volume. The demand represents a cyclical pattern for different daytimes, weekdays and weeks. The difficulty lies within the demand realization. If there are not sufficient resources available, the EMS cannot respond timely to an emergency call and this may cost a human life. Therefore, robust planning is important. In the Master's thesis different possible demand scenarios are generated by a sampling approach and then used as input for a staffing model which determines individual work schedules for each employee. The aim of the thesis is to determine whether this is an appropriate approach.
Simultaneous ramp-up and allocation planning in the automotive industry
(in cooperation with Daimler AG)
During a production ramp-up the produced goods are usually scarce. In a global industry, like the automotive industry, several markets compete for the produced goods. Demand size and market entry date further characterize the demand. This thesis aims to provide a linear program (LP) to support the assignment of products to the different markets. Thereby, strategic prioritizations have to be considered before the distribution decision can be made. However, usually it is desirable to maintain a certain level of fairness regarding demand fulfillment between the different markets.
Transient analysis of finite capacity tandem queues with arbitrary distributions: An extension of the SBC-approximation
An approximation to analyze the time-dependent behavior of single stage queueing systems is the stationary backlog-carryover (SBC) approach by Stolletz 2008 and Stolletz/
Cumulative diagrams for the performance analysis of check-in counters and security checks in airport terminals
Waiting lines in front of check-in counters at airports are typical examples for dynamic queueing systems. In this example the arrival process of passengers and the service process are both characterized by time-dependent behavior. An approximation by fluid models is an often used method to analyze the behavior of such queueing system and shall be presented in this diploma thesis.
Simulation-based slot scheduling of truck handling operations at the Lufthansa Cargo Hub in Frankfurt
Airfreight deliveries for outbound flights and pick-ups from inbound flights are shipped by trucks to and from cargo terminals. At Lufthansa Cargo in Frankfurt, truck arrivals and departures typically occur dynamically and stochastically over time, making cost-efficient capacity provision a challenging task. In order to improve performance by shifting demand from peak to off-peak times, the airline plans to implement a slot booking system for scheduling loading and unloading activities at the truck handling facilities. By conducting a simulation analysis, the student evaluates and quantifies the potential performance improvement of such a slot-based appointment system and compares different slot scheduling scenarios on the basis of multiple performance measures. Furthermore, implications of expected future demand growth and variations in capacity provision are analyzed as well.
Operations scheduling with sequence-dependent processing times
The student gave an overview on the topic of scheduling (i.e. assigning pending jobs to available resources and putting them into a favorable order) under the assumption that the setup time or processing time depends on the order. The student performed a thorough literature research, outlining and evaluating about 200 scientific articles concerning this topic and classifying them by their assumptions and solution approaches.
Vehicle routing with time windows: Models and applications
The Vehicle Routing Problem (VRP) aims at finding optimal routes for delivering goods to customers. The student is to conduct a literature research, creating an overview on the current state of research on the topic, covering different objectives, assumptions and restrictions (like limited time windows) as well as solution approaches. The paper „The Vehicle Routing Problem: An overview of exact and approximate algorithms“ by Laporte (1992) gives a first (but possibly outdated) overview.
The application of queueing theory for health care in hospitals
Different examples for the application of queueing theory can be found in the healthcare and the hospital sector. Examples are the patient care in a hospital emergency room, waiting lists for organ transplantations or even the management of the inventory of hospital beds. This diploma thesis shall give an overview of the use of queueing theory in the healthcare sector especially the hospital sector. Furthermore the different applications shall be described and analyzed.