Optimisation des horaires des agents et du routage des appels dans les centres d’appels

Nous étudions la gestion de centres d'appels multi-compétences, ayant plusieurs types d'appels et groupes d'agents. Un centre d'appels est un système de files d'attente très complexe, où il faut généralement utiliser un simulateur pour évaluer ses performances. Tout d'abord, nous développons un simulateur de centres d'appels basé sur la simulation d'une chaîne de Markov en temps continu (CMTC), qui est plus rapide que la simulation conventionnelle par événements discrets. À l'aide d'une méthode d'uniformisation de la CMTC, le simulateur simule la chaîne de Markov en temps discret imbriquée de la CMTC. Nous proposons des stratégies pour utiliser efficacement ce simulateur dans l'optimisation de l'affectation des agents. En particulier, nous étudions l'utilisation des variables aléatoires communes. Deuxièmement, nous optimisons les horaires des agents sur plusieurs périodes en proposant un algorithme basé sur des coupes de sous-gradients et la simulation. Ce problème est généralement trop grand pour être optimisé par la programmation en nombres entiers. Alors, nous relaxons l'intégralité des variables et nous proposons des méthodes pour arrondir les solutions. Nous présentons une recherche locale pour améliorer la solution finale. Ensuite, nous étudions l'optimisation du routage des appels aux agents. Nous proposons une nouvelle politique de routage basé sur des poids, les temps d'attente des appels, et les temps d'inoccupation des agents ou le nombre d'agents libres. Nous développons un algorithme génétique modifié pour optimiser les paramètres de routage. Au lieu d'effectuer des mutations ou des croisements, cet algorithme optimise les paramètres des lois de probabilité qui génèrent la population de solutions. Par la suite, nous développons un algorithme d'affectation des agents basé sur l'agrégation, la théorie des files d'attente et la probabilité de délai. Cet algorithme heuristique est rapide, car il n'emploie pas la simulation. La contrainte sur le niveau de service est convertie en une contrainte sur la probabilité de délai. Par après, nous proposons une variante d'un modèle de CMTC basé sur le temps d'attente du client à la tête de la file. Et finalement, nous présentons une extension d'un algorithme de coupe pour l'optimisation stochastique avec recours de l'affectation des agents dans un centre d'appels multi-compétences.

Modeling and Analysis of Two-Part Type Manufacturing Systems

This paper presents a model and analysis of a synchronous tandem flow line that produces different part types on unreliable machines. The machines operate according to a static priority rule, operating on the highest priority part whenever possible, and operating on lower priority parts only when unable to produce those with higher priorities. We develop a new decomposition method to analyze the behavior of the manufacturing system by decomposing the long production line into small analytically tractable components. As a first step in modeling a production line with more than one part type, we restrict ourselves to the case where there are two part types. Detailed modeling and derivations are presented with a small two-part-type production line that consists of two processing machines and two demand machines. Then, a generalized longer flow line is analyzed. Furthermore, estimates for performance measures, such as average buffer levels and production rates, are presented and compared to extensive discrete event simulation. The quantitative behavior of the two-part type processing line under different demand scenarios is also provided.

The application of a reinforcement learning agent to a multi-product manufacturing facility

An intelligent agent-based scheduling system, consisting of a reinforcement learning agent and a simulation model has been developed and tested on a classic scheduling problem. The production facility studied is a multiproduct serial line subject to stochastic failure. The agent goal is to minimise total production costs, through selection of job sequence and batch size. To explore state space the agent used reinforcement learning. By applying an independent inventory control policy for each product, the agent successfully identified optimal operating policies for a real production facility.

On work in progress optimisation

In this paper, we investigate how to best optimise the level of work in progress (WIP) in a real world factory. Using a simulation model of the factory, we show that an optimum level of WIP can be attained. By systematically varying the maximum allowable level of WIP within different model runs, results show that the throughput reaches a high level very quickly and then tapers off. The production lead times, in contrast, begin at relatively low levels and increase after the optimum WIP level has been reached.

Simulating the change from a push to pull production system

This paper will document the initial discrete-event simulation performed to study a proposed change from a push to a pull system in an existing manufacturing company. The system is characterised by five machine lines with intermediate buffers, and five major part groupings. A simulation model has been developed to mimic the flow of kanban cards in the physical system, by using a series of requests that propagate back through the facility, which the machines must respond to. The customer
demand therefore controls the level of activity in the plant. The results of the initial modelling steps will be presented in this paper, especially the impact of kanban lot size and demand variability on the output and stability of the production system, from which a set of future work is proposed.

Modeling a just-in-time plant using sequencing algorithms

An Australian automotive component company plans to assemble and deliver seats to customer on just-in-time basis. The company management has decided to model operations of the seat plant to help them make decisions on capital investment and labour requirements. There are four different areas in seat assembly and delivery areas. Each area is modeled independently to optimise its operations. All four areas are then combined into one model called the plant model to model operations of seat plant from assembly to delivery. Discrete event simulation software is used to model the assembly operations of seat plant.

A flow control methodology for the rapid generation of vehicle traffic models

This paper presents a conveyor-based methodology to model complex vehicle flows common to factory and distribution warehouse facilities. The AGV and human path modelling techniques available in many commercial discrete event simulation packages require extensive knowledge and time to implement even the simplest flow control rules for multiple vehicle interaction. Although discrete event simulation is accepted as an effective tool to model vehicle delivery movements, human paths and delivery schedules for modern assembly lines, the time to generate accurate models is a significant limitation of existing simulation-based optimisation methodologies. The flow control method has been successfully implemented using two commercial simulation packages. It provides a realistic visual representation, as well as accurate statistical results, and reduces the model development process cost.

Towards development of a quality cost model for automotive stamping

The current work used discrete event simulation techniques to model the economics of quality within an actual automotive stamping plant. Automotive stamping is a complex, capital intensive process requiring part-specific tooling and specialised machinery. Quality control and quality improvement is difficult in the stamping environment due to the general lack of process understanding and the large number to interacting variables. These factors have prevented the widespread use of statistical process control. In this work, a model of the quality control techniques used at the Ford Geelong Stamping plant is developed and indirectly validated against results from production. To date, most discrete event models are of systems where the quality control process is clearly defined by the rules of statistical process control. However, the quality control technique used within the stamping plant is for the operator to perform a 100% visual inspection while unloading the finished panels. In the developed model, control is enacted after a cumulative count of defective items is observed, thereby approximating the operator who allows a number of defective panels to accumulate before resetting the line. Analysis of this model found that the cost sensitivity to inspection error is dependent upon the level of control and that the level of control determines line utilisation. Additional analysis of this model demonstrated that additional inspection processes would lead to more stable cost structures but these structures many not necessarily be lower cost. The model was subsequently applied to investigate the economics of quality improvement. The quality problem of panel blemishes, induced by slivers (small metal fragments), was chosen as a case stuffy. Errors of 20-30% were observed during direct validation of the cost model and it was concluded that the use of discrete event simulation models for applications requiring high accuracy would not be possible unless the production system was of low complexity. However, the model could be used to evaluate the sensitivity of input factors and investigating the effects of a number of potential improvement opportunities. Therefore, the research concluded that it is possible to use discrete event simulation to determine the quality economics of an actual stamping plant. However, limitations imposed by inability of the model to consider a number of external factors, such as continuous improvement, operator working conditions or wear and the lack of reliable quality data, result in low cost accuracy. Despite this, it still can be demonstrated that discrete event simulation has significant benefits over the alternate modelling methods.

Estimating performance indexes of a baggage handling system using metamodels

In this study, we develop some deterministic metamodels to quickly and precisely predict the future of a technically complex system. The underlying system is essentially a stochastic, discrete event simulation model of a big baggage handling system. The highly detailed simulation model of this is used for conducting some experiments and logging data which are then used for training artificial neural network metamodels. Demonstrated results show that the developed metamodels are well able to predict different performance measures related to the travel time of bags within this system. In contrast to the simulation models which are computationally expensive and expertise extensive to be developed, run, and maintained, the artificial neural network metamodels could serve as real time decision aiding tools which are considerably fast, precise, simple to use, and reliable.

Dynamic control of skilled and unskilled labour task assignments

Efficient allocation of skilled and non-skilled workers allow a company to improve productivity and usually requires an understanding of personnel capability, operating conditions and resource availability. This paper examines a labour control strategy that optimises labour skill level, utilisation, task execution time and processing error. The proposed controller manages different labour groups in a multiple work cell environment, providing real-time job assignment, as well as guiding and navigation features. These features can be used to enhance the performance of existing MRP-based or Just-In-Time production systems. A discrete event simulation-based manufacturing model has been developed to assess the performance of the labour controller. Experiments conducted for the selected production scenarios have demonstrated a productivity improvement when using the proposed control. A second experiment has shown that when a skilled labour uses the labour controller to guide them through the job, their utilisation also increases. The proposed controller also has potential application in other domains, such as minimising the shopping time at a supermarket

Future integrated factories: a system of systems engineering perspective

The manufacturing sector has gone through tremendous change in the last decade. We have witnessed the transformation from stand alone, manual processes to smart and integrated systems, from hand written reports to interactive computer-based dashboards. Future integrated factories will operate as a system of systems through intelligent machines, human factors integration, and integrated supply chains. To effectively operate and manage these emerging enterprises, a systems science approach is required. Modelling and simulation is recognised as a key enabling technology, with application from stakeholder engagement and knowledge elicitation to operational decision support through self-tuning and self-assembling simulations. Our research has led to the introduction of effective modelling and simulation methods and tools to enable real time planning, dynamic risk analysis and effective visualisation for production processes, resources and systems. This paper discusses industrial applicable concepts for real-time simulation and decision support, and the implications to future integrated factories, or factories of the future, are explored through relevant case studies from aerospace manufacturing to mining and materials processing enterprises.

A multi-level approach to planning and scheduling resources for aviation training

This paper describes a multi-level system dynamics (SD) / discrete event simulation (DES) approach for assessing planning and scheduling problems within an aviation training continuum. The aviation training continuum is a complex system, consisting of multiple aviation schools interacting through interschool student and instructor flows that are affected by external triggers such as resource availability and the weather.
SD was used to model the overall training continuum at a macro level to ascertain relationships between system entities. SD also assisted in developing a shared understanding of the training continuum, which involves constructing the definitions of the training requirements, resources and policy objectives. An end-to-end model of the continuum is easy to relate to, while dynamic visualisation of system behaviour provides a method for exploration of the model.
DES was used for micro level exploration of an individual school within the training continuum to capture the physical aspects of the system including resource capacity requirements, bottlenecks and student waiting times. It was also used to model stochastic events such as weather and student availability. DES has the advantage of being able to represent system variability and accurately reflect the limitations imposed on a system by resource constraints.
Through sharing results between the models, we demonstrate a multi-level approach to the analysis of the overall continuum. The SD model provides the school’s targeted demand to the DES model. The detailed DES model is able to assess schedules in the presence of resource constraints and variability and provide the expected capacity of a school to the high level SD model, subjected to constraints such as instructor availability or budgeted number of training systems. The SD model allows stakeholders to assess how policy and planning affect the continuum, both in the short and the long term.
The development of this approach permits moving the analysis of the continuum between SD and DES models as appropriate for given system entities, scales and tasks. The resultant model outcomes are propagated between the continuum and the detailed DES model, iteratively generating an assessment of the entire set of plans and schedule across the continuum. Combining data and information between SD and DES models and techniques assures relevance to the stakeholder needs and effective problem scoping and scaling that can also evolve with dynamic architecture and policy requirements.
An example case study shows the combined use of the two models and how they are used to evaluate a typical scenario where increased demand is placed on the training continuum. The multi-level approach provides a high level indication of training requirements to the model of the new training school, where the detailed model indicates the resources required to achieve those particular student levels.