Guido Buchholz

Self-optimising Production Systems

One of the central success factors for production in high-wage countries is the solution of the conflict that can be described with the term “planning efficiency”. Planning efficiency describes the relationship between the expenditure of planning and the profit generated by these expenditures. From the viewpoint of a successful business management, the challenge is to dynamically find the optimum between detailed planning and the immediate arrangement of the value stream. Planning-oriented approaches try to model the production system with as many of its characteristics and parameters as possible in order to avoid uncertainties and to allow rational decisions based on these models. The success of a planning-oriented approach depends on the transparency of business and production processes and on the quality of the applied models. Even though planning-oriented approaches are supported by a multitude of systems in industrial practice, an effective realisation is very intricate, so these models with their inherent structures tend to be matched to a current stationary condition of an enterprise. Every change within this enterprise, whether inherently structural or driven by altered input parameters, thus requires continuous updating and adjustment. This process is very cost-intensive and time-consuming; a direct transfer onto other enterprises or even other processes within the same enterprise is often impossible. This is also a result of the fact that planning usually occurs a priori and not in real-time. Therefore it is hard for completely planning-oriented systems to react to spontaneous deviations because the knowledge about those naturally only comes a posteriori.

Meta-modeling for manufacturing processes

Meta-modeling for manufacturing processes describes a procedure to create reduced numeric surrogates that describe cause-effect relationships between setting parameters as input and product quality variables as output for manufacturing processes. Within this method, expert knowledge, empiric data and physical process models are transformed such that machine readable, reduced models describe the behavior of the process with sufficient precision. Three phases comprising definition, generation of data and creation of the model are suggested and used iteratively to improve the model until a required model quality is reached. In manufacturing systems, such models allow the generation of starting values for setting parameters based on the manufacturing task and the requested product quality. In-process, such reduced models can be used to determine the operating point and to search for alternative setting parameters in order to optimize the objectives of the manufacturing process, the product quality. This opens up the path to self-optimization of manufacturing processes. The method is explained exemplarily at the gas metal arc welding process.

Machine vision system for online weld pool observation of gas metal arc welding processes

At the Welding and Joining Institute of the RWTH Aachen University, an optical sensor system which allows the online observation of weld pools when utilising gas metal arc welding (GMAW) processes has been developed. Moreover, an image processing software has been created which provides information about the actual weld pool width as well as the actual position of the gap in relation to the position of the wire electrode. This supplemental process information makes it possible to deduce the width of the resulting weld and the extent of the potential misalignment of the welding torch. One key requirement of the developed system and its algorithms was the assurance of a defined time constraint. If this is given, the GMAW process can be adapted to potential disturbances so that the desired weld quality can be achieved. The system has been successfully tested on typical groove preparations, like fillet welds, butt joints and V-joints. This paper focuses on the setup and the functionality of the machine vision system and the challenges of gaining usable information with simple, fast and reliable algorithms.

Possibilities of a control of the droplet detachment in pulsed gas metal arc welding

With the use of pulsed gas metal arc welding processes, a regular material transfer, resulting in a good weld quality, should be reached. Ideally, in every pulse, exactly one droplet should be detached. Such ideal processes can be achieved by now, but require process knowledge and stable boundary conditions. In most cases, the process is set up in such way, that the droplet detachment is forced through a higher energy input than necessary. Thereby, the energy input is measured per pulse, independent of the moment of the droplet detachment. The aim of the research behind this paper is it to automatically decide, if a droplet has been detached. With this information, a new time interval can be used to statistically analyse the real energy input regarding the material transfer; but also a detachment control can be (and has been) realised, which forces a droplet detachment even under process disturbances or unconfident parameter choices. As a technological vision, the droplet detachment control can be used to optimise the process regarding a lower energy input. So in the last part of this paper, different possibilities of adapting pulse parameters and their potential to reduce the energy input are discussed.

Progress Towards Model Based Optimisation Of Gas Metal Arc Welding Processes

Integrated welding production units are nowadays facing increasing demands for higher flexibility and autonomy in order to guarantee a high product quality and to reduce non-productive down time, which is caused by disturbances from changing process boundary conditions. On the one hand, this calls for concepts which support the set-up procedure in case of a product change and allow the efficient production of a wide product variety in order to reduce technical and economical expenditure. On the other hand, a more efficient exploitation of the technical potentials of the welding production systems is necessary. This can be achieved by improving the information processing capabilities of a production system and by enabling the system to autonomously control the process even at its technological limits. In order to manage these tasks, applicable and innovative control strategies are required. Within the scope of this article, a model based self-optimisation method for a gas metal arc welding (GMAW) process is explained. In automated welding production, self optimisation aims for the selection and provision of suitable welding parameters concurrent with changing tasks and requirements. A self-optimisation procedure can be defined as set-up procedure as well as for the use within an online process control strategy. Since both approaches are model based, the inverse usage of quality models and the applied optimisation algorithms represent a further priority of the paper.

Self-optimizing Production Technologies

Customer demands have become more individual and complex, requiring a highly flexible production. In high-wage countries, efficient and robust manufacturing processes are vital to ensure global competitiveness. One approach to solve the conflict between individualized products and high automation is Model-based Self-optimization (MBSO). It uses surrogate models to combine process measures and expert knowledge, enabling the technical system to determine its current operating point and thus optimize it accordingly. The objective is an autonomous and reliable process at its productivity limit. The MBSO concept is implemented in eight demonstrators of different production technologies such as metal cutting, plastics processing, textile processing and inspection. They all have a different focus according to their specific production process, but share in common the use of models for optimization. Different approaches to generate suitable models are developed. With respect to implementation of MBSO, the challenge is the broad range of technologies, materials, scales and optimization variables. The results encourage further examination regarding industry applications.

Determination of process variables in melt-based manufacturing processes

Industrial manufacturing requires continuous production at reliable quality to be competitive. Many manufacturing processes are run close to their technological limits to increase productivity what leads to a significant threat of malfunction in the case of limited control over setting parameters or deviating boundary conditions. This paper discusses the difficulties of determining process variables during manufacturing for two melt-based manufacturing processes, laser cutting and gas metal arc welding. Both manufacturing processes show a highly dynamic and complex behaviour which today prevents a physical description of all interactions of the process variables. On the practical side, even the dominant process variables cannot be measured as they are not directly accessible. The approach that is presented here suggests a combined solution with both modelling and measuring tools that connect through surrogate criteria. It involves a simplified modelling of the manufacturing process that describes the process behaviour well enough and that can be evaluated numerically within a short time frame. The measurement evaluates a property of the process which is well accessible. This ensures robust signal processing and stable information about the surrogate criterion. In combination with the simplified model, the operating point of the process can easily be determined. For laser cutting of metal sheets and gas metal arc welding, it is demonstrated how to acquire information about the process and how to model surrogates. The research is focused on providing tools for fast machine set-up and for components which can be used for self-optimisation.

Surrogate Modelling as an Enabler for Self Optimisation for Production Processes

To persist the ever-increasing market challenges, production processes must be capable to raise their flexibility without incremented expenses. Therefore, the topic of self optimisation for production processes is gaining more and more in importance. In this paper, a process independent method for self optimisation with focus on surrogate modelling is presented. Especially the creation of surrogate models and the possibilities of model evaluation to improve optimisation methods will be a central theme. All described methods will be illustrated by reference to examples from the gas metal arc welding process.

Model-based description of arc length as a synergetic system parameter in pulsed GMAW

Gas metal arc welding demonstrates a dynamic and inherently stochastic process behavior. Interdependencies of machine setup and process boundary conditions and the resulting process and weld product qualities are complex in nature, as the optimization of GMAW processes with regard to product quality demands expert knowledge. This work introduces a practical approach of mapping the arc length as a physical key process variable to the most influential pulsed GMAW setup parameters by means of generation and application of statistical process models. It further describes the necessary steps, including the determination of certain investigated variables, generating a statistical reasonable design of experiment and the actual model calculation and evaluation. Furthermore, field test evaluations of derived statistical models have been discussed, proving a successful statistical model generation process.

Process control of gas metal arc welding processes by optical weld pool observation with combined quality models

At the Institute of Welding and Joining Technology at RWTH Aachen University, an optical sensor system has been realized which allows the online observation of the weld pool while the process is executed. Therefore, an image processing software has been developed which provides information about the actual width of the weld pool as well as the actual position of the gap in relation to the position of the wire electrode. With this process information, the resulting width of the weld and the width of the misalignment of the welding torch can be deduced. These data shall be used to adapt setting parameters at the occurrence of deviations to guarantee a consistent weld geometry and position. On the one hand, it will be illustrated how information about the weld width can help to improve and simplify quality models. On the other hand, the weld width obtained by image evaluation will be correlated to real macro-sections and the adjusted values will be used as input for a calculation of further geometry parameters. Associated with setting parameters like welding speed and wire feed speed, these information provide the basis for quality models which are used to control and supervise the process.