Publications
2026
Schultheis, Alexander; Späth, Maximilian; Weich, Justin; Bergmann, Ralph
Towards Explainable and Reusable Temporal Case-Based Reasoning in Predictive Maintenance: A Research Environment Proceedings Article
In: Artificial Intelligence XLIII - 46th SGAI International Conference on Artificial Intelligence, AI 2026, Cambridge, UK, December 15-17, 2026, Proceedings, Springer, 2026.
Abstract | BibTeX | Tags: Industry 4.0 Data, Internet of Things, Predictive Maintenance, Temporal Case-Based Reasoning, Time Series Similarity
@inproceedings{SchultheisSWB2026,
title = {Towards Explainable and Reusable Temporal Case-Based Reasoning in Predictive Maintenance: A Research Environment},
author = {Alexander Schultheis and Maximilian Späth and Justin Weich and Ralph Bergmann},
year = {2026},
date = {2026-01-01},
booktitle = {Artificial Intelligence XLIII - 46th SGAI International Conference on Artificial Intelligence, AI 2026, Cambridge, UK, December 15-17, 2026, Proceedings},
publisher = {Springer},
series = {Lecture Notes in Computer Science},
abstract = {Predictive Maintenance (PredM) relies on multivariate sensor time series to detect emerging failures and support timely interventions. Temporal Case-Based Reasoning (TCBR) is promising in such data-scarce maintenance settings because similar past temporal cases can support retrieval, explanation, and later reuse. Existing work, however, mainly evaluates fault classification and lacks reusable environments for investigating these directions in PredM. This paper presents a case-based research environment for PredM as a foundation for explainable and reusable TCBR. Requirements for similarity assessment are derived, and a systematic literature study is conducted to select a suitable approach and dataset. Based on this, an expert-knowledge-enhanced Siamese Neural Network (SiamNN) as a similarity measure and an industry-near Fischertechnik dataset are selected. For this dataset, a suitable vocabulary is investigated, and a case base is provided based on it. This case base and the SiamNN integration are implemented within the Case-Based Reasoning Framework ProCAKE. The validation shows preserved retrieval behavior and substantially accelerated retrieval of new events. This contribution and the corresponding implementation form the research foundation for explanation and case reuse through adaptation.},
keywords = {Industry 4.0 Data, Internet of Things, Predictive Maintenance, Temporal Case-Based Reasoning, Time Series Similarity},
pubstate = {published},
tppubtype = {inproceedings}
}
Predictive Maintenance (PredM) relies on multivariate sensor time series to detect emerging failures and support timely interventions. Temporal Case-Based Reasoning (TCBR) is promising in such data-scarce maintenance settings because similar past temporal cases can support retrieval, explanation, and later reuse. Existing work, however, mainly evaluates fault classification and lacks reusable environments for investigating these directions in PredM. This paper presents a case-based research environment for PredM as a foundation for explainable and reusable TCBR. Requirements for similarity assessment are derived, and a systematic literature study is conducted to select a suitable approach and dataset. Based on this, an expert-knowledge-enhanced Siamese Neural Network (SiamNN) as a similarity measure and an industry-near Fischertechnik dataset are selected. For this dataset, a suitable vocabulary is investigated, and a case base is provided based on it. This case base and the SiamNN integration are implemented within the Case-Based Reasoning Framework ProCAKE. The validation shows preserved retrieval behavior and substantially accelerated retrieval of new events. This contribution and the corresponding implementation form the research foundation for explanation and case reuse through adaptation.