Laurent Geneste

Dispatching rules in scheduling Dispatching rules in scheduling: a fuzzy approach

Abstract Job-shop scheduling through simulation uses various kinds of dispatching rules such as SPT or the slack time rule. Each of these rules aims at satisfying a single criterion although workshop management is a multi-criteria problem. This paper proposes a way to use fuzzy logic in order to build aggregated rules allowing to obtain a compromise between the satisfaction of several criteria. When the criteria of performance change with the evolution of the production environment, these aggregated rules can be parametrized in order to modify the respective influence of the elementary rules they are composed of.

Knowledge formalization in experience feedback processes: An ontology-based approach

Because of the current trend of integration and interoperability of industrial systems, their size and complexity continue to grow making it more difficult to analyze, to understand and to solve the problems that happen in their organizations. Continuous improvement methodologies are powerful tools in order to understand and to solve problems, to control the effects of changes and finally to capitalize knowledge about changes and improvements. These tools involve suitably represent knowledge relating to the concerned system. Consequently, knowledge management (KM) is an increasingly important source of competitive advantage for organizations. Particularly, the capitalization and sharing of knowledge resulting from experience feedback are elements which play an essential role in the continuous improvement of industrial activities. In this paper, the contribution deals with semantic interoperability and relates to the structuring and the formalization of an experience feedback (EF) process aiming at transforming information or understanding gained by experience into explicit knowledge. The reuse of such knowledge has proved to have significant impact on achieving the missions of companies. However, the means of describing the knowledge objects of an experience generally remain informal. Based on an experience feedback process model and conceptual graphs, this paper takes domain ontology as a framework for the clarification of explicit knowledge and know-how, the aim of which is to get lessons learned descriptions that are significant, correct and applicable.

Continuous improvement through knowledge-guided analysis in experience feedback

Continuous improvement in industrial processes is increasingly a key element of competitiveness for industrial systems. The management of experience feedback in this framework is designed to build, analyze and facilitate the knowledge sharing among problem solving practitioners of an organization in order to improve processes and products achievement. During Problem Solving Processes, the intellectual investment of experts is often considerable and the opportunities for expert knowledge exploitation are numerous: decision making, problem solving under uncertainty, and expert configuration. In this paper, our contribution relates to the structuring of a cognitive experience feedback framework, which allows a flexible exploitation of expert knowledge during Problem Solving Processes and a reuse such collected experience. To that purpose, the proposed approach uses the general principles of root cause analysis for identifying the root causes of problems or events, the conceptual graphs formalism for the semantic conceptualization of the domain vocabulary and the Transferable Belief Model for the fusion of information from different sources. The underlying formal reasoning mechanisms (logic-based semantics) in conceptual graphs enable intelligent information retrieval for the effective exploitation of lessons learned from past projects. An example will illustrate the application of the proposed approach of experience feedback processes formalization in the transport industry sector.

Integration of uncertain and imprecise orders in the MRP method

Nowadays, one of the main difficulties of Production Management is to take into account the increasing uncertainty of the customer demand. In an MRP system, this uncertainty is mainly managed at middle term and through successive actualizations of the planning. We suggest in this paper a way to explicitly model the uncertainty and imprecision of the demand allowing to pass through all the MRPII steps (Material Requirement Planning, Load Planning, Scheduling). This method, named Fuzzy-MRP (F-MRP) allows to visualize at each step a much more rich information for the decision makers, taking into account not only the certain data but also a quantification of the various eventualities that may arise. Decisions requiring a long preparation (sub-contracting, order of components, increase of capacity, etc.) can so be considered earlier, on the base of quantified data.

Adding decision support to workflow systems by reusable standard software components

Industrial information systems like Enterprise Resource Planning (ERP) systems are increasingly comprehensive and integrated. Nevertheless, satisfying all the user requirements regarding information processing or decision support within a unique tool seems still to be unrealistic. As a consequence, being able to quickly provide the users with additional pieces of software for supporting specific decisions remains more than ever a topic of interest. Specific developments take time, are costly, have usually low reliability and are often poorly integrated with the main information system. In order to address these drawbacks, we suggest a structure and the first elements of a toolbox aimed at allowing an easier development of additional pieces of information/decision support system (DSS) by reuse of standard software components. This toolbox allows the implementation of workflow and groupware facilities and the communication between modules is achieved through a database which provides the integration with the main information system. The first decision support modules developed include an expert system generator, a neural network simulator, a simplex module and a Case-Based Reasoning (CBR) module. Examples of applications developed using this toolbox are described, and a development methodology is suggested.

Scheduling uncertain orders in the customer-subcontractor context

Within the customer–subcontractor negotiation process, the first problem of the subcontractor is to provide the customer with a reliable order lead-time although his workload is partially uncertain. Actually, a part of the subcontractor workload is composed of orders under negotiation which can be either confirmed or cancelled. Fuzzy logic and possibility theory have widely been used in scheduling in order to represent the uncertainty or imprecision of processing times, but the existence of the manufacturing orders is not usually set into question. We suggest a method allowing to take into account the uncertainty of subcontracted orders. This method is consistent with list scheduling: as a consequence, it can be used in many classical schedulers. Its implementation in a scheduler prototype called TAPAS is described. In this article, we focus on the performance of validation tests which show the interest of the method.

Distributed machining control and monitoring using smart sensors/actuators

The study of smart sensors and actuators led, during the past few years, to the development of facilities which improve traditional sensors and actuators in a necessary way to automate production systems. In another context, many studies have been carried out aimed at defining a decisional structure for production activity control and the increasing need of reactivity leads to the autonomization of decisional levels close to the operational system. We study in this paper the natural convergence between these two approaches and we propose an integration architecture, dealing with machine tool and machining control, that enables the exploitation of distributed smart sensors and actuators in the decisional system.

Multi-heuristic scheduling: three approaches to tune compromises

Most of the available industrial schedulers are based on a simulation approach using dispatching rules. These rules are often dedicated to the satisfaction of a single performance criterion, and are used whatever the characteristics of the workshop or of the set of jobs. An approach which allows one to bring in compromises between rules is set out in this paper. These compromises can be parametered in accordance with the objectives of the workshop and the characteristics of the jobs in order to introduce some reactivity in the decision system. Three ways to set up the parameters are compared: experimental design, fuzzy expert system and neural network. The method allowing one to define compromises can be implemented on each scheduler that uses a simulation approach. Tests have been made with an industrial scheduler called SIPAPLUS, the results of which are developed in this paper.

Management of imprecision and uncertainty for production activity control

The operational levels of production management, often called production activity control (PAC) or manufacturing process control, require increasing reaction capabilities in order to adapt the workshop management to the changes of its environment. It often implies giving more responsibilities to the low decision levels. However, the management of the corresponding degrees of freedom is generally unusual. In such a situation, decision support systems (DSSs) provide a way to reconcile the satisfaction of mid-level objectives and the reaction requirements. A conceptual model is described that provides a design framework for a PAC DSS. Since the available knowledge lies mainly in expertise, a DSS has been implemented using a knowledge-based system. The uncertainty and imprecision of the managed information led to the use of fuzzy logic as a modeling tool. Moreover, various inference semantics have been implemented in the expert rules because different kinds of reasoning have been identified. Two versions of the DSS are described and several examples of implemented reasoning processes are developed.

Application of optimization techniques to parameter set-up in scheduling

Scheduling requires to set-up a number of parameters that have a direct influence on the schedule quality. Since scheduling is a highly unstable process, it is usually a long and complex task to tune manually these parameters in order to optimize a set of objectives. Meta-heuristics have recently been successfully used for schedule optimization, but an important modeling effort is usually required in order to express the problem to solve within the specific framework of each method. Moreover, these techniques are often time-consuming and their application to problems of industrial size may be hazardous. It is suggested in this article a way to combine meta-heuristics in a black box approach in order to select, then set-up scheduling parameters on industrial-scale scheduling problems, i.e. problems where several tens of criteria can be combined in order to build an objective function, several tens of parameters can be used, with a schedule involving several hundreds of machines and several thousands of tasks. An implementation framework has been developed and tested on an industrial scheduler, named Ortems®. The first results of the use of this framework on real industrial databases are described and commented.