Development of methods and procedures to automate the operative production control and scheduling considering the aspects of sustainability

The production order doesn't only influence logistic and operational ratios, but also shows effects on the environment as well as ergonomics and the staff's psychology. Aim of the project was to develop a method to optimize the job order planning in consideration of an integrated evaluation in terms of sustainable economic activity.

Short Description




Initial Situation and Motivation

Today's production processes are more or less controlled by classical economical criteria. Typical key performance indicators are resource utilisation, stock volumes, throughput times and due dates. Criteria like energy and operating resources consumption, waste and pollutant emissions, stress charges, work monotony are rarely and especially not yet systematically considered in sequence planning and scheduling. Both order sequences as well as lot sizes show an essential impact on sustainability indicators. Inappropriate sequences require additional set-up, cleaning and adjusting procedures, that yield in additional energy consumption (e.g. for tool preheating) as well as an increased pollutant emission caused by detergents or residual material (e.g. caused by material or colour change). Further, the order sequence also influences work ergonomics. Giving the same mean value, manual tasks may be planned in a manner that short charges are followed by breaks of the same lengths, or in such a way that longer charges are followed by longer breaks. A sensitised production control system may consider this fact by analysing not only the mean time of breaks but also their distribution and individual durations. Considering these aspects in a holistic evaluation of schedules can prevent mental exhaustion, stress reactions, failure rates and unconfident behaviour. Further, a balanced coordination of following production orders will make manual tasks more diversified, which in turn work against fatigue and increases production quality.

Contents and Objectives

Aim of the project was to develop methods and procedures to automate the operative production control (rolling detailed production planning) considering sustainable economics. The envisaged solution considers criteria of sustainability like work ergonomics, energy and resources consumption, amount of waste etc. and holds a production automatically at an – from a holistic point of view – optimal point of operation. The aimed automation of the planning tasks ensures a stable planning quality at a high optimisation level as well as an increased transparency and a full documentation of planning decisions. The results were evaluated in a casting production which is characterised by a high level of manual tasks as well as many chronologically overlapping operations.

Methodological Approach

A simulation-based method was selected as the basis of the intended decision support system: Different scenarios are created within the discrete event process model by the use of the optimization methods, which are being evaluated by the rating model. The rating model gives feedback to the optimization method (scheduler, dispatcher) about the quality of the generated solution, so that a continuous process of improvement can take place. The results of the planning/optimization runs are being visualized as Gantt charts or as strategy dialog and are being communicated to the employees in that way.

Due to the vast stochastic influences by everyday disruptions the idea arose to develop both schedules and priority rules at the same time, and to switch from scheduling to dispatching in case of a disruption.

For the creation of schedules the concept of Resource Constraint Project Scheduling (RCPS) was chosen and implemented as Adaptive Large Neighbourhood Search (a neighbourhood heuristic).

For the rule based dispatching the process model was enhanced with rule interpreters, which evaluate the priority rules at the decision points and therefore make decisions based on the actual situation. The rules were synthesized and optimized automatically with Genetic Programming.

For the holistic rating of the generated scenarios the indicators which were observed within the process model are being weighted and merged into a single fitness value within the subsequent rating model. The most important elements of the rating models are the transfer functions, which transform the different values which are represented in different measurement units into one and the same unit.


In the course of the project, the methodological approach has been implemented prototypically in .NET as a testing and evaluation environment using the example of a plastics casting process. SiRO, a discrete event simulation library implemented by Profactor, has served as a foundation of the implementation of the simulation and rating models. HeuristicLab, an optimization framework developed by HEAL, was used for the synthesis and the optimization of dispatching rules. It was also combined with the process and rating models to form an optimization system. The Adaptive Large Neighbourhood Search algorithm was implemented in C++. The reference data necessary for the initialization of the process models as well as the order data were drawn from the database of the existing ERP system APplus. The process model involves all the relevant worksteps from the core preparation to the finishing process of the moulded parts. The individual rating model takes into consideration economical criteria (throughput, timeliness, pass-through time), ecological criteria (material consumig setup tasks, reworking tasks) and work ergonomic aspects (regular pauses, regeneration before extra hours, monotony, heat and cold). Additionally to the single indicators and the merged fitness value the solution quality of either schedules or dispatching rules is visualized by a Gantt-Chart.

The experiments which were conducted during the project confirmed the chosen approaches: Both implemented methods can be used to optimize order sequences, although this does not yet state anything about the quality of both the chosen algorithms and their parameterization.

Due to the high complexity of the real processes, especially of the manual tasks, more effort than initially planned had to be put into the creation of the process and rating models. In spite of the high level of detail, a few aspects of the real production environment (constraints) could not yet have been completely modelled. Also the rating model requires further research, which could not be implemented and tested completely during the project runtime.

For a deeper scientific study of the interdependencies between different configurations and parameter settings of optimization algorithms, rating model and simulation model a follow-up project was initiated. In this project, especially the algorithms and models should be explored in terms of robustness and performance. The goal is to clarify how the algorithms and the synthesized complex priority rules react to data flaws, change of process parameters, different order data or changes in the production strategy (changes in the rating model). Another point to be examined scientifically is the interpretation of the dispatching rules synthesized by the genetic programming (complex priority rules).

Project Partners

Project management

Dr. Markus Vorderwinkler

Project collaborator

Helga Heiß

Projekt partners

  • arbeitsleben KG
  • FH OÖ Forschungs & Entwicklungs GmbH
  • Universität Wien, Lehrstuhl für Produktion und Logistik
  • ABF Industrielle Automation GmbH
  • ASMA GmbH
  • Schneegans-Silicon GmbH

Contact Address

Dr. Markus Vorderwinkler
Im Stadtgut A2, A-4407 Steyr-Gleink
Tel.: +43 (7252) 885