Analysis of Common Prediction Models for a Fuzzy Connected Source Target Production Based on Time Dependent Significance. Soller, S., Hoelzl, G., Greiler, T., & Kranz, M. In Proceedings of the 2022 International Conference on Embedded Wireless Systems and Networks, of EWSN '22, pages 226–231, USA, 2022. Junction Publishing. ![Analysis of Common Prediction Models for a Fuzzy Connected Source Target Production Based on Time Dependent Significance [pdf]](https://bibbase.org/img/filetypes/pdf.svg "pdf")
Paper abstract bibtex In an industrial setup quality measurements are taken in multiple steps of each production chain. Often a single product is evaluated for several steps in the production, but even more often pooled tests are done, to evaluate the quality of the used material or a charge. Instead of the traditional classification of sensor measurement to quality on a single process step, the question arises, how these steps interact with each other. Is it possible to foresee the faults of later production steps, by analyzing data gathered earlier? Especially in mass production this leads to unclear and fuzzy relationships. The material quality might not be known for every work piece produced but controlled in a fixed time interval. The challenge of these processes is, to correctly connect ground truth to feature vector by their temporal connection. In this work we show multiple steps to reach a better classification and insight into the production process. We gather data from a real-world environment and as a first shot, use common machine learning methods, which are available through public libraries. Therefore, we created a simple connection between the material trace elements and quality inspection. Further data analysis suggested the influence of the exact time of the quality inspection related to the measurement of trace elements. We performed a significance test, to proof the difference of time groups to each other. The identical machine learning methods were applied to these time groups and an improvement of classification accuracy of 2% could be detected. For feature approaches we propose an automatic split system, to find time dependent groups inside the data and split the data accordingly.
@inproceedings{ewsn2022_1,
abstract = {In an industrial setup quality measurements are taken in multiple steps of each production chain. Often a single product is evaluated for several steps in the production, but even more often pooled tests are done, to evaluate the quality of the used material or a charge. Instead of the traditional classification of sensor measurement to quality on a single process step, the question arises, how these steps interact with each other. Is it possible to foresee the faults of later production steps, by analyzing data gathered earlier? Especially in mass production this leads to unclear and fuzzy relationships. The material quality might not be known for every work piece produced but controlled in a fixed time interval. The challenge of these processes is, to correctly connect ground truth to feature vector by their temporal connection. In this work we show multiple steps to reach a better classification and insight into the production process. We gather data from a real-world environment and as a first shot, use common machine learning methods, which are available through public libraries. Therefore, we created a simple connection between the material trace elements and quality inspection. Further data analysis suggested the influence of the exact time of the quality inspection related to the measurement of trace elements. We performed a significance test, to proof the difference of time groups to each other. The identical machine learning methods were applied to these time groups and an improvement of classification accuracy of 2\% could be detected. For feature approaches we propose an automatic split system, to find time dependent groups inside the data and split the data accordingly.},
address = {USA},
author = {Soller, Sebastian and Hoelzl, Gerold and Greiler, Tobias and Kranz, Matthias},
booktitle = {Proceedings of the 2022 International Conference on Embedded Wireless Systems and Networks},
date-added = {2023-02-03 09:18:58 +0100},
date-modified = {2023-02-03 10:01:39 +0100},
keywords = {Reliability Keywords Industry 40, Machine Learning, Performance, Maintenance Prediction},
location = {Linz, Austria},
numpages = {6},
pages = {226--231},
publisher = {Junction Publishing},
series = {EWSN '22},
title = {Analysis of Common Prediction Models for a Fuzzy Connected Source Target Production Based on Time Dependent Significance},
url = {/publications/2022_ewsn_Fuzzy.pdf},
year = {2022},
bdsk-url-1 = {/publications/2022_ewsn_Fuzzy.pdf}}
Downloads: 0
{"_id":"5BDoJTX4oyEwC6WgD","bibbaseid":"soller-hoelzl-greiler-kranz-analysisofcommonpredictionmodelsforafuzzyconnectedsourcetargetproductionbasedontimedependentsignificance-2022","author_short":["Soller, S.","Hoelzl, G.","Greiler, T.","Kranz, M."],"bibdata":{"bibtype":"inproceedings","type":"inproceedings","abstract":"In an industrial setup quality measurements are taken in multiple steps of each production chain. Often a single product is evaluated for several steps in the production, but even more often pooled tests are done, to evaluate the quality of the used material or a charge. Instead of the traditional classification of sensor measurement to quality on a single process step, the question arises, how these steps interact with each other. Is it possible to foresee the faults of later production steps, by analyzing data gathered earlier? Especially in mass production this leads to unclear and fuzzy relationships. The material quality might not be known for every work piece produced but controlled in a fixed time interval. The challenge of these processes is, to correctly connect ground truth to feature vector by their temporal connection. In this work we show multiple steps to reach a better classification and insight into the production process. We gather data from a real-world environment and as a first shot, use common machine learning methods, which are available through public libraries. Therefore, we created a simple connection between the material trace elements and quality inspection. Further data analysis suggested the influence of the exact time of the quality inspection related to the measurement of trace elements. We performed a significance test, to proof the difference of time groups to each other. The identical machine learning methods were applied to these time groups and an improvement of classification accuracy of 2% could be detected. For feature approaches we propose an automatic split system, to find time dependent groups inside the data and split the data accordingly.","address":"USA","author":[{"propositions":[],"lastnames":["Soller"],"firstnames":["Sebastian"],"suffixes":[]},{"propositions":[],"lastnames":["Hoelzl"],"firstnames":["Gerold"],"suffixes":[]},{"propositions":[],"lastnames":["Greiler"],"firstnames":["Tobias"],"suffixes":[]},{"propositions":[],"lastnames":["Kranz"],"firstnames":["Matthias"],"suffixes":[]}],"booktitle":"Proceedings of the 2022 International Conference on Embedded Wireless Systems and Networks","date-added":"2023-02-03 09:18:58 +0100","date-modified":"2023-02-03 10:01:39 +0100","keywords":"Reliability Keywords Industry 40, Machine Learning, Performance, Maintenance Prediction","location":"Linz, Austria","numpages":"6","pages":"226–231","publisher":"Junction Publishing","series":"EWSN '22","title":"Analysis of Common Prediction Models for a Fuzzy Connected Source Target Production Based on Time Dependent Significance","url":"/publications/2022_ewsn_Fuzzy.pdf","year":"2022","bdsk-url-1":"/publications/2022_ewsn_Fuzzy.pdf","bibtex":"@inproceedings{ewsn2022_1,\n\tabstract = {In an industrial setup quality measurements are taken in multiple steps of each production chain. Often a single product is evaluated for several steps in the production, but even more often pooled tests are done, to evaluate the quality of the used material or a charge. Instead of the traditional classification of sensor measurement to quality on a single process step, the question arises, how these steps interact with each other. Is it possible to foresee the faults of later production steps, by analyzing data gathered earlier? Especially in mass production this leads to unclear and fuzzy relationships. The material quality might not be known for every work piece produced but controlled in a fixed time interval. The challenge of these processes is, to correctly connect ground truth to feature vector by their temporal connection. In this work we show multiple steps to reach a better classification and insight into the production process. We gather data from a real-world environment and as a first shot, use common machine learning methods, which are available through public libraries. Therefore, we created a simple connection between the material trace elements and quality inspection. Further data analysis suggested the influence of the exact time of the quality inspection related to the measurement of trace elements. We performed a significance test, to proof the difference of time groups to each other. The identical machine learning methods were applied to these time groups and an improvement of classification accuracy of 2\\% could be detected. For feature approaches we propose an automatic split system, to find time dependent groups inside the data and split the data accordingly.},\n\taddress = {USA},\n\tauthor = {Soller, Sebastian and Hoelzl, Gerold and Greiler, Tobias and Kranz, Matthias},\n\tbooktitle = {Proceedings of the 2022 International Conference on Embedded Wireless Systems and Networks},\n\tdate-added = {2023-02-03 09:18:58 +0100},\n\tdate-modified = {2023-02-03 10:01:39 +0100},\n\tkeywords = {Reliability Keywords Industry 40, Machine Learning, Performance, Maintenance Prediction},\n\tlocation = {Linz, Austria},\n\tnumpages = {6},\n\tpages = {226--231},\n\tpublisher = {Junction Publishing},\n\tseries = {EWSN '22},\n\ttitle = {Analysis of Common Prediction Models for a Fuzzy Connected Source Target Production Based on Time Dependent Significance},\n\turl = {/publications/2022_ewsn_Fuzzy.pdf},\n\tyear = {2022},\n\tbdsk-url-1 = {/publications/2022_ewsn_Fuzzy.pdf}}\n\n","author_short":["Soller, S.","Hoelzl, G.","Greiler, T.","Kranz, M."],"key":"ewsn2022_1","id":"ewsn2022_1","bibbaseid":"soller-hoelzl-greiler-kranz-analysisofcommonpredictionmodelsforafuzzyconnectedsourcetargetproductionbasedontimedependentsignificance-2022","role":"author","urls":{"Paper":"https://hasisaurus.at/publications/2022_ewsn_Fuzzy.pdf"},"keyword":["Reliability Keywords Industry 40","Machine Learning","Performance","Maintenance Prediction"],"metadata":{"authorlinks":{}}},"bibtype":"inproceedings","biburl":"https://hasisaurus.at/publications/All4Web.bib","dataSources":["5TQQC7wyLLts8eKF7"],"keywords":["reliability keywords industry 40","machine learning","performance","maintenance prediction"],"search_terms":["analysis","common","prediction","models","fuzzy","connected","source","target","production","based","time","dependent","significance","soller","hoelzl","greiler","kranz"],"title":"Analysis of Common Prediction Models for a Fuzzy Connected Source Target Production Based on Time Dependent Significance","year":2022}