Print Email Facebook Twitter Optimizing the maintenance interval by modeling the state of a conveyor belt Title Optimizing the maintenance interval by modeling the state of a conveyor belt Author Berenbak, C.J. Contributor Pang, Y. (mentor) Faculty Mechanical, Maritime and Materials Engineering Department Maritime and Transport Technology Programme Transport Engineering and Logistics Date 2015-01-27 Abstract Traditionally, maintenance of conveyor belts is performed using corrective maintenance. This means that maintenance to conveyor belts is only performed if the belt fails or severe damage is detected to the belt. Visual inspections to the belt are carried out to detect defects to the belt before the belt fails. Once a belt fails or a defect is detected, maintenance is performed. The amount of maintenance to a system of belts heavily depends on the failures and defects detected. So during a period with little detected defects, the workload is low while the opposite also is true. This leads to a very uneven work load for the department or company performing the maintenance to the belts. Because maintenance is only performed once severe damage is detected or the belt has already failed, the system reliability is lowered. The reliability of the system and the spread of the work load can be improved by using preventive maintenance instead of corrective. Preventive maintenance is performed before severe damage to the belt is present. But how do you know when to perform the preventive maintenance. Performing the maintenance too early will lead to an increase of the number of maintenance actions over time. Performing the maintenance too late and the corrective maintenance has already taken place. Developing a method to determine when the maintenance has to take place, the so called maintenance interval, is the main focus of this research. To determine the optimum maintenance interval a model has been developed. The model has been designed with conveyor belts into mind but can easily be used on every component that requires preventive maintenance. The model is developed using the Bayesian Belief Network (BBN). Belief networks are graphical representations of models that capture the relationships between the model's variables. The variables that interact directly are identified and are limited to the variables to which they are directly connected. Belief networks may use directed or undirected graphs to represent a dependency model. The directed acyclic graph (DAG) provides a better representation than the undirected graphs. The DAG is also more flexible and is able to represent a wider range of probabilistic independencies. An undirected graph is one where the edges have no direction meaning (A, B) is equal to (B, A). The BBN is a specific type of causal belief network. As for any causal belief network, the nodes represent stochastic variables and the arcs identify direct causal influences between the linked variables. The fundamentals of the Bayesian methodology is too enable prior knowledge of a certain event to calculate the posterior probability of a hypothesis based on the probability of the event. One of the challenges of the BBN method is incorporating information with a large number of possible values. The thickness of a conveyor belt for example changes of its lifetime because of wear. To take this type of information into account in the model, fuzzy logic is introduced. Fuzzy logic is used to assign a degree of membership to an event. By assigning the thickness of the belt a number of ranges instead of thickness in millimetre, the amount of variations for this node is limited to the number or ranges. The BBN model is created by using both historical data as the knowledge of experts concerning the part where the model is used for into account. The historical data provide the basic information necessary for the model. The expert opinion can be used to check the information supplied by the historical data as fill in missing data. By introducing reliability of the data and information to the model, the usefulness in practice can be increased. Data for the model determined by a large number of sources and checked by an expert can be considered as reliable. The opposite is also valid and by looking at the reliability of the outcome of the model, the influence of the model on the decision making process can be described. Another factor that is taken into account in the model is the spread in the output. The output of the model will always have an uncertainty that is translated in a spread. This spread can be influenced by the reliability of the model and for example a safety factor for the maintained part in question. The developed BBN model can provide a boost in both the reliability of the system the part is present in as reducing the fluctuations in the workload for maintenance operations. The workings of the model have been proven with the usage of a test case at the company Tata Steel although the output was not accurate enough to use in practice. Further research is necessary to increase the accuracy of the model to enable the industry to use the method during normal operations. Subject Conveyor beltsPreventive maintenanceBayesian Belief NetworkFuzzy logic To reference this document use: http://resolver.tudelft.nl/uuid:59ff732f-9292-455c-b40a-191d183d50cb Embargo date 2020-01-01 Part of collection Student theses Document type master thesis Rights (c) 2015 Berenbak, C.J. Files PDF Master_Thesis_Coen_Berenb ... L.7898.pdf 3.82 MB Close viewer /islandora/object/uuid:59ff732f-9292-455c-b40a-191d183d50cb/datastream/OBJ/view