Premium
Automatic Detection of Critical Points in Bottling Plants with a Model‐based Diagnosis Algorithm
Author(s) -
Flad Stefan,
Struss Peter,
Voigt Tobias
Publication year - 2010
Publication title -
journal of the institute of brewing
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.523
H-Index - 51
eISSN - 2050-0416
pISSN - 0046-9750
DOI - 10.1002/j.2050-0416.2010.tb00786.x
Subject(s) - bottling line , downtime , computer science , algorithm , component (thermodynamics) , data mining , engineering , mechanical engineering , physics , bottle , thermodynamics , operating system
The efficiency of bottling plants typically ranges between 40 to 70%. Automatic conditioning monitoring helps to find critical points in a plant and supports the operator in optimizing the plant. But bottling plants are complex lines of several linked machines and currently critical points can only be identified manually. This paper presents a model‐based efficiency analysis tool. It automatically localizes critical points that decrease the efficiency of the whole plant and the tool is adaptable to different plants solely through parameterization. The algorithm compares the behaviour of a plant with an OK‐model of the plant. If there are inconsistencies, the commercial tool RAZ'R finds a failure model that is consistent with the observed plant behaviour, thus localizing the component that causes downtime of the filler. The algorithm succeeds in identifying the cause of the downtime in 90% of the cases. A demonstrator application, which runs on different plants, has been implemented and requires only simple adaptation steps. The algorithm only needs information about the configuration of the plant and the production data, which normally exists in automated plants. In the future, it is expected that the project partners will integrate the algorithm with their PDA‐systems, such that the model‐based analysis will help to increase the efficiency of bottling plants.