Project: Probabilistic Bayesian soft sensor - a tool for on-line estimation of the key process variable in cold rolling mills
One of the key objectives of any rolling mill control system is to keep the thickness of the processed material within the prescribed tolerance band, which can be as low as +-10 microns for thin strips. Unfortunately, no practical direct measurement of the gauge within the rolling gap is possible. Failure to comply with the tolerances results in losses which, according to experts estimate, might go up to 10 % of the profit for poorly equipped rolling mills. _x000D__x000D_The continuous urge to alleviate the problem in the process of cold rolling has resulted in several remedies during the last decades. The gaugemeter principle, for example, combines knowledge of the mill stretch characteristic and the available measurement of the rolling force. Another option is use of direct feedback from the output thickness measurement. The great problem here is unavoidable transport delay, preventing elimination of fast thickness variability. Various model-based schemes for the AGC (Automatic Gauge Control) utilizing measurements of input thickness and other variables (tensions, speeds, etc.) have been suggested. Unfortunately the underlying sensors are unavailable in many situations. The same holds true for the mass-flow principle relying on precise measurement of both thicknesses and strip speeds, which implies utilization of expensive measurement equipment. The enduring problem with existing solutions is threefold. First, they are applicable to adequately equipped rolling mills and under specific conditions. Moreover, in some cases the sensor configuration frequently changes during the rolling process (for example the thickness meters can be withdrawn in some phases of rolling). Secondly, their performance is significantly conditioned by the reliability and accuracy of the collected data. Third, overlooked gradual drifts in signals, loss of precision and other incipient sensor malfunctions could significantly deteriorate the control system performance. _x000D__x000D_The underlying project focuses on small and medium European metal processing plants which have to face tough competition with limited funding potentials. Therefore the CO motivation arises from bold economic requirements for significant cost reduction while preserving high performance. In that respect the aims of the project are to propose new solutions to the AGC that will (1) avoid the need for purchasing expensive measurement systems and (2) to maximize the performance of the thickness control by providing reliable estimation of the strip thickness (gauge) within the rolling gap on-line and in real time._x000D_For example, it is expected that reliable gauge estimation would avoid the need for laser velocimeters on both sides of the mill which are normally required for working AGC of the mass-flow type. Commonly used measurement of speeds by incremental encoders connected to deflector rolls ought to be sufficient instead. Thus significant savings of order of magnitude 80,000 EUR could be achieved per mill._x000D_The existing solutions rely on tailoring of the control system according to the signals availability and in explicit switching of the control modes according to momentary rolling conditions. Apart from that, the key novelty of the project arises from the innovative design of high performance "soft sensor", which estimates the gauge in the rolling gap utilizing all available information and the above-mentioned methods at once. The Bayesian approach allows systematic treatment of uncertainty minimizing the need for heuristic solutions of partial problems such as controller's switching, etc. The estimated value should be directly usable for the thickness control and for mill operators. The estimator output will be in the form of probability distribution thus providing clear information about reliability of the estimation. Integral part of the developed tool will be the monitor of quality of all signals providing additional information for the mixing estimator. As a result, the estimator will provide value of the gauge in all times, differing in the shape of the probability density function according to quality and momentary availability of incoming data. _x000D_The project will be realized by a consortium with internationally renowned expertise in particular fields. COMPUREG will be responsible for data collection, design and programming of specific modules and their integration into the control system of the rolling mill, experiments and dissemination of practical results. INEA will be responsible for implementation of the advanced signal monitor within the PLC environment and its testing. JSI will be responsible for research and development of algorithms of signal filtering, monitoring and fault detection. UTIA will be responsible for research in the field of Bayesian mixing and decision making. _x000D__x000D_ProBaSensor will be prepared for the market within two years after completion of the project as an optional part of the rolling mill control system.
| Acronym | ProBaSensor (Reference Number: 4632) |
| Duration | 01/07/2009 - 30/06/2012 |
| Project Topic | The project aims to develop a novel on-line estimator of the key process variable in rolling mills by mixing multiple models with different sensitivities to inaccuracy in process data. The approach relies on the systematic treatment of uncertainty and merging of all available information. |
|
Project Results (after finalisation) |
1) The functional sample of the ProBaSensor system which was successively tested in two metal processing plants._x000D_2) The utility model (a national patent) concerning improved speed measurement._x000D_3) 9 papers in proceedings of international conferences. |
| Network | Eurostars |
| Call | Eurostars Cut-Off 2 |
Project partner
| Number | Name | Role | Country |
|---|---|---|---|
| 4 | COMPUREG Plzen, s.r.o. | Coordinator | Czech Republic |
| 4 | INEA - informatizacija, energetika, avtomatizacija d.o.o. | Partner | Slovenia |
| 4 | Institute of Information Theory and Automation, Academy of Sciences of the Czech Republic | Partner | Czech Republic |
| 4 | Jozef Stefan Institute | Partner | Slovenia |