Computational approaches for pattern/correlation analysis based on historical data
Preventive maintenance scheduling is needed by high value manufacturing industry, and the attention is stressed on exploration of techniques that can facilitate accurate prediction of unscheduled downtimes. The objective of this project was to develop an algorithm to analyze event logs that impro...
Saved in:
| Main Author: | |
|---|---|
| Other Authors: | |
| Format: | Final Year Project |
| Language: | English |
| Published: |
2011
|
| Subjects: | |
| Online Access: | http://hdl.handle.net/10356/42732 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Institution: | Nanyang Technological University |
| Language: | English |
| Summary: | Preventive maintenance scheduling is needed by high value manufacturing industry, and the attention is stressed on exploration of techniques that can facilitate accurate prediction of unscheduled downtimes.
The objective of this project was to develop an algorithm to analyze event logs that improves preventive maintenance scheduling. We proposed a simplified model for failure prediction-Two Tier approach. In his approach eqiupment labelled to be in one of the two states.The state of the machine is idetified thorugh some threshold. The algorithm has four following steps: Input of restriction parameters,choice of the metric of measurement, threshold identification, time to failure estimation, censor time estimation, performance testing.
In this project algorithm is implemented using C# programming language and GUI is implemented using Windows Form utilities. Database is stored using Oracle DBMS.
There were two experiments with two different sets of data. The first experiment gave results both during training and testing. We did not avoid a large number of unscheduled downtimes, but we decreased the number of redundant maintenances, giving factories an opportunity to cut down the cost. During the second experiment we achieved the results only during training, but the preventive profile did not give results during testing. We tried to preprocess the data differently, but could not achieve results we hoped for.
The strength of this algorithm is that it can work different types of data and it is reconfigurable according to factory specifications. Data preprocessing this algorithm is versatile and can be applied to different equipments. This algorithm is a simplified approach, is not able to track sophisticated relationship between faults and failures. |
|---|