Improvement of quality management in the cold rolling and finishing area by combining failure mode and effect analysis with data-based approaches

The aim of this project was the combination of traditional methods for the determination of failure causes, like the FMEA, and data-based methods. The usability of the approach was shown by means of another method, the AFD. For the data-based methods the univariate and multivariate statistics, artificial neural networks and decision trees were used. During this project the following types of combinations were investigated: 1) FMEA parameter calculation with data based methods At first the data-based methods are used to calculate the FMEA parameters occurrence and severity. The detectability is in the investigated context constant due to the usage of measured values. The parameters are calculated by means of a Self Organising Map (SOM). With the developed and implemented software tools it is possible to make a summarization and a simulation of the FMEA parameters due to changing process variables. 2) On-line plant diagnosis Based on a FMEA made for the plant a data-base was designed and implemented, in which the FMEA rules were stored. By means of a Fuzzy Inference Engine a new virtual sensor was created, which is able to determine the plant status using the FMEA database and measured process data. The result is shown on-line on the monitoring system of the plant. The occurrence of a plant anomaly is shown in form of optical signals, the highlighting of the involved plant section as well as an anomaly description and the corresponding actions. 3) Investigation of cause-and-effect relationships of quality defects In this part several real quality problems of flat products were analysed. The causes of surface defects on flat strip were investigated by means of a data-based analysis of process and quality data of the whole production process. The results of the data-based analysis were evaluated with FMEA rules as well as the FMEA rule-base was enhanced with results gathered during the data-based analysis. The usability of the AFD was proven by the investigation of two different quality problems at two different plants. By means of a commercial tool the methodology was used in the failure analysis mode. The AFD was detected to be a very helpful method for the in-depth analysis of failure situations. It helps to avoid the usual analytical way to look on failures following the intended design and functionality of the plant or the production unit. The involved technicians are not locked in a kind of 'routine blindness'. For the daily usage of the investigated tool a on-site demonstration has shown that the commercial product is too general, it lacks of specificity to solve typical problems of the production of flat steel products. The methodology for the combined analysis was implemented on different sites producing different types of flat products. The concepts developed for the systems installed are in the main components very similar: interfaces to data acquisition systems of central databases, development of FMEA rule-bases and user interfaces. The systems are embedded into the environment of the different plants using different technologies (web-based application or stand alone applications). The production lines involved are producing flat products of different steel types: stainless steel and tinplate and one pilot line for cold treatment. Furthermore the several types of plants necessary for the production of flat products are taken into consideration: from melt shop over hot and cold rolling, annealing and slitting lines. During this project the methodology of the combined analysis has shown its usability. The experience gathered during the test phase of the developed and implemented systems at the different plants has shown the proof of the concept and the implementations. By means of the combined analysis some reasons for quality deficiencies of flat products could be identified and, by means of the knowledge gathered with the FMEA, suggestions for corrective actions were given. After that, the quality of the product could be improved. The developed methodology is well capable for other plants.