Context:
This data set originates from a practice-relevant degradation process, which is representative for Prognostics and Health Management (PHM) applications. The observed degradation process is the clogging of filters when separating of solid particles from gas. A test bench is used for this purpose, which performs automated life testing of filter media by loading them. For testing, dust complying with ISO standard 12103-1 and with a known particle size distribution is employed. The employed filter media is made of randomly oriented non-woven fibre material.
Further data sets are generated for various practice-relevant data situations which do not correspond to the ideal conditions of full data coverage. These data sets are uploaded to Kaggle by the user "Prognostics @ HSE" in a continuous process. In order to avoid the carryover between two data sets, a different configuration of the filter tests is used for each uploaded practice-relevant data situation, for example by selecting a different filter media.
Detailed specification:
For more information about the general operation and the components used, see the provided description file Preventive to Predicitve Maintenance dataset.pdf
Given data situation:
The data set Preventive to Predicitve Maintenance is about the transition of a preventive maintenance strategy to a predictive maintenance strategy of a replaceable part, in this case a filter. To aid the realisation of predictive maintenance, life cycles have already been recorded from the application studied. However, the preventive maintenance in place so far causes them to be replaced after a fixed period of time, regardless of the condition of the degrading part. As a result, the end of life is not known for most records and thus they are right-censored. The so given training data are recorded runs of the filter up to a periodic replacement interval.
When specifying the interval length for preventive maintenance, a trade-off has to be made between wasted life and the frequency of unplanned downtimes that occur, when having a particularly short life. The interval here is chosen so that, on average, failure is observed at the shortest 10% of the filter lives in the training data. The other lives are censored. The filter failure occurs when the differential pressure across the filter exceeds 600 Pa. The maintenance interval length depends on the amount of dust fed in per time, which is constant within a test run. For example, at twice the dust feed, the maintenance interval is half as long. The same relationship therefore applies to the respective censoring time, which scales inversely proportional with the particle feed. The variations between lifetimes are therefore primarily based on the type of dust, the flow rate and manufacturing tolerances. The filter medium CC 600 G was used exclusively for these measurement samples, which are included in this data set.
Task:
The objective of the data set is to precisely predict the remaining useful life (RUL) of the filter for the given test data, so a transition to predictive maintenance is made possible. For this purpose, the dataset contains training and test data, consisting both of 50 life tests respectively. The test data contains randomly right-censored run-to-failure measurements and the respective RUL as a ground truth to the prediction task. The main challenge is how to make the most use of the right-censored life data within the training data.
Due to the detailed description of the setup and the various physical filter models described in literature, it is possible to support the actual data-driven models by integrating physical knowledge respectively models in the sense of theory-guided data science or informed machine learning (various names are common).
**Acknowledgement: **
Thanks go to Marc Hönig (Scientific Employee), Marcel Braig (Scientific Employee) and Christopher Rein (Research Assistant) for contributing to the recording of these life tests.
Data set Creator:
Hochschule Esslingen - University of Applied Sciences
Research Department Reliability Engineering and Prognostics and Health Management
Robert-Bosch-Straße 1
73037 Göppingen
Germany
Dataset Citation:
Hagmeyer, S., Mauthe, F., & Zeiler, P. (2021). Creation of Publicly Available Data Sets for Prognostics and Diagnostics Addressing Data Scenarios Relevant to Industrial Applications. International Journal of Prognostics and Health Management, Volume 12, Issue 2, DOI: 10.36001/ijphm.2021.v12i2.3087
