Enterprise modelling using hybrid techniques and process optimization: A decision-support system for mail processing centres

Performance and quality are the main resulting factors for the analysis and evaluation of the business and manufacturing processes. Different modelling architectures, simulation methods and evaluation techniques have been applied to these processes for performance improvement. In the postal company Royal Mail (UK), these processes involve complex logistics because of the size of its facilities (e.g. 72 sorting centres), the diversity and the volume of products (e.g. 122 millions items per day), and the constraints of resources (e.g. 220 000 employees). The company faces tremendous pressure from globalization, competition and the postal industry regulator to improve performance and service quality. Furthermore, the modelling and simulation tools that the company relies on have demonstrated some inefficiencies: - 1) the modelling is not formalised, 2) there is no model management capability, and 3) there is a lack of analytical and meta-heuristic methods used to solve planning problems. Thus the company is in an acute need of better modelling and evaluation approaches to improve its performance.

The challenges faced are substantial and this research has provided a decision support system approach to tackle the inefficiencies above. First the CIMOSA modelling concept is coupled with the high level coloured Petri Net simulator to implement a formalized modelling environment. Second, the Case-Based Reasoning concept is used with semantic and graph matching techniques to provide an intuitive model management facility. Thirdly, neural networks and heuristic methods are used to optimize the schedule for the mail sorting process.

The integrated formalized modelling approach enables the integrated view of the different perspectives of an enterprise which allows a more holistic study of the mail process. The model management approach provides a more accurate case retrieval with the use of information on the model’s structure, characters and problems. The approach also added an adaptation case reuse mechanism which allows modelling experiences to be accumulated. The neural network and heuristic optimization method allows the mail sorting schedules to be improved in terms of the latest completion times as compared to the priority rule used in the sorting centre (e.g. First Come First Serve).