Project: Intelligent Process Control in Foundry Manufacturing
Increasing individualization and the joint proliferation of options with simultaneously growing product complexity, results inevitably in a huge effort in manufacturing processes. Successful manufacturing of enterprises is COly dependant how quick the reaction on new demands of customers and market can be realized. It requires accordingly highest standards of process reliability and stability, which require extensive process knowledge. One of the most complex manufacturing processes is the foundry process, which is defined by the particularly high number of sub-processes of the process chain, by the individual dependencies between the process parameters and variables, and by numerous elements of uncertainty, which affect the stability of any sub-process applied and, consequently, product quality. To this end, it is necessary to know and to control the causes, factors and their interaction which contribute to this quality. These quality factors will extend with increasing production flexibility and product complexity. Providing sufficient process chain stability under these conditions, processes and practises used in manufacturing of castings should be consistent. The basic problem in foundries is that the numerous attributive variables may be non-linear. Delays in feedback make the data meaningless for real time control, key variables often are not identified, therefore not measured or not even controlled, measured data are not reliable and a lot more. Even when foundries are often confronted with this situation, the causes and the solutions for quality improvement are hardly known. Sufficient data collection systems and intelligent information process-ing trace casting defects back to processes, operations and the factors accountable for them: by consequent application of intelligent information processing, sufficient process controlling is assured to manufacture by stable processes. The particular demand is to keep the process in a narrow range. This necessitates a combination of collected continuously monitored process data along the complete process chain, suitable analysing tools, foundry expert knowledge, control of every single procedure as well as of the process chain, materials and subprocesses, process control tools to join input variables and output results to a closed self controlling loop._x000D__x000D_During the last 20 years a lot of different tools have been developed for data analysing in manufacturing processes. These are COly neuronal networks (NN), Bayesian networks (BN), fuzzy logic techniques, generic algorithms, statistical data analyses and expert systems. Most of them are applied individual, some have been combined like NN with fuzzy technologies or NN with BN. Additionally most of these tools have been designed for a specific application and do not have the ability to get applied on foundry problems. Some tools for process and quality control in metallurgical processes, like steelmaking and rolling, have been developed; specific solutions for foundry are not available. _x000D_ _x000D_Against this background the goal of this project is to develop a platform for intelligent process control in manufacturing. Specific interest will be laid on foundry production. The technical demands on this tool have been described above. Preliminary tests of the consortium showed that the application of single tools do fulfil these demands COly because of insufficient data preparation, specific constraints related to the individual structure and behaviour of variables and processes, and lack of integration of expert knowledge. The innovation of the new tool will be the automatic data and process controlled method combination, based on the integration of intelligent data preprocessing, combination of different analysis methods, and foundry expert knowledge. The tool will be evaluated, modified and optimized on different foundry systems._x000D__x000D_The project consortium consists of foundries which manufacture high quality cast parts in cast iron and Aluminium for automotive and engineering industry. They represent the actual state of the art in the most important areas of foundry production regarding technology and market._x000D__x000D_The Ps are Eidologic GmbH (Recklinghausen, Germany, dataprocessing, data analysis, knowledge based systems), Kemptener Eisengießerei AG (Germany, iron castings for mechanical engineering), ClaasGuss GmbH (Bielefeld, Germany, parts in nodular cast iron for traffic and mechanical engineering), Fundiciones Garbi, S.A. (Abadiño,Spain, safety and precision components for automotive), TS Fundiciones S.A. (Arroa-Behea, Spain, cast parts in grey and nodular iron for the windmill industry), Cofundi S.L. (Mungia, Spain, Aluminium foundry, heat treatment, finishing), University of Kempten (Germany, foundry process technology, knowledge base)._x000D_Subcontractors are: Azterlan (Durango, Spain, engineering and development of foundry processes) and University of Deusto (Bilbao, Spain, intelligent data analyses systems).
| Acronym | IPRO (Reference Number: 5092) |
| Duration | 01/07/2010 - 31/10/2013 |
| Project Topic | Process reliability is an extraordinary demandig factor in nearly all manufacturing processes. The necessity of stable process variables requires to develop an innovative intelligent process control software based on a self controlling combination of different data analysis methods. |
|
Project Results (after finalisation) |
The aim of the project was to develop a software that is able to predict process results from independant input parameters. The prediction should be based on methods of machine learning. Such a software is essential for further improvement of foundry processes as these processes are complicate and include complex interaction of input parameters. Therefore it is often difficult to derive laws of cause and effect by conventional statistical methods. This target was achieved by the developed software EIDOminer. To support the software development CLAAS GUSS (CG) descibed all relevant sub processes of the foundry process in detail and defined the respective input and output parameters. Availability of process data was checked and a data base was established where all relevant process parameteres were gathered. This data base formed the basis to generate data sets to test the software at the different stages of development. Input was given for further software development. With the data the function box was trained for selected foundry questions and the prognisis of the software was compared to measured process results - with good results. Furthermore CG gave input to the development of the knowledge data base. |
| Network | Eurostars |
| Call | Eurostars Cut-Off 3 |
Project partner
| Number | Name | Role | Country |
|---|---|---|---|
| 7 | Claas Guss GmbH | Partner | Germany |
| 7 | Cofundi productos no férricos de Mungia, S. L. | Partner | Spain |
| 7 | Eidologic GmbH | Coordinator | Germany |
| 7 | Fundiciones Garbi, S.A. | Partner | Spain |
| 7 | Kemptener Eisengießerei Adam-Hönig AG | Partner | Germany |
| 7 | T. S. Fundiciones, S. A. | Partner | Spain |
| 7 | University of Applied Science, Kempten | Partner | Germany |