Udava

Husom, Erik Johannes; Tverdal, Simeon; Göknil, Arda; Sen, Sagar

doi:10.1145/3522664.3528603

Cited by 7 publications

(3 citation statements)

References 31 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Some works [32]- [34] discuss unsupervised learning for predictive maintenance and anomaly detection without mentioning AI engineering aspects. Only recently, Husom et al [35] explicitly discuss and evaluate their AI-based approach (i.e., an unsupervised learning pipeline for sensor data validation) for industrial settings from the AI engineering perspective. Our work presents the field knowledge of how a continual learning pipeline is engineered and its learning and inference experiences are employed, from which researchers can benefit.…”

Section: Related Workmentioning

confidence: 99%

Replay-Driven Continual Learning for the Industrial Internet of Things

Sen

Nielsen

Husom

et al. 2023

2023 IEEE/ACM 2nd International Conference on AI Engineering – Software Engineering for AI (CAIN)

Self Cite

View full text Add to dashboard Cite

The Industrial Internet of Things (IIoT) leverages thousands of interconnected sensors and computing devices to monitor and control large and complex industrial processes. Machine learning (ML) applications in IIoT use data acquired from multiple sensors to perform tasks such as predictive maintenance. While remembering useful learning from the past, these applications need to adapt learning for evolving sensor data stemming from changes in industrial processes and environmental conditions. This paper presents a continual learning pipeline to learn from the evolving data while replaying selected parts of the old data. The pipeline is configured to produce ML experiences (e.g., training a baseline neural network model), improve the baseline model with the new data while replaying part of the old data, and infer/predict using a specific model version given a stream of IIoT sensor data. We have evaluated our approach from an AI Engineering perspective using three industrial case studies, i.e., predicting tool wear, remaining useful lifetime, and anomalies from sensor data acquired from CNC machining and broaching operations. Our results show that configuring experiences for replay-driven continual learning allows dynamic maintenance of ML performance on evolving data while minimizing the excessive accumulation of legacy sensor data.

show abstract

Section: Related Workmentioning

confidence: 99%

Replay-Driven Continual Learning for the Industrial Internet of Things

Sen

Nielsen

Husom

et al. 2023

2023 IEEE/ACM 2nd International Conference on AI Engineering – Software Engineering for AI (CAIN)

Self Cite

View full text Add to dashboard Cite

show abstract

“…This section exemplifies unsupervised learning systems and their configuration. Figure 1 illustrates an example of an unsupervised learning system [39], i.e., an ML pipeline that automatically discovers reference patterns for process behavior in sensor data for AI-enabled IIoT. The pipeline consists of three main steps: data preprocessing, unsupervised learning of clusters, and labeling and validating new data.…”

Section: Configuration Of Unsupervised Learningmentioning

confidence: 99%

AutoConf: Automated Configuration of Unsupervised Learning Systems Using Metamorphic Testing and Bayesian Optimization

Shar,

Goknil,

Husom

et al. 2023

2023 38th IEEE/ACM International Conference on Automated Software Engineering (ASE)

View full text Add to dashboard Cite

Unsupervised learning systems using clustering have gained significant attention for numerous applications due to their unique ability to discover patterns and structures in large unlabeled datasets. However, their effectiveness highly depends on their configuration, which requires domain-specific expertise and often involves numerous manual trials. Specifically, selecting appropriate algorithms and hyperparameters adds to the complexity of the configuration process. In this paper, we propose, apply, and assess an automated approach (AutoConf ) for configuring unsupervised learning systems using clustering, leveraging metamorphic testing and Bayesian optimization. Metamorphic testing is utilized to verify the configurations of unsupervised learning systems by applying a series of input transformations. We use Bayesian optimization guided by metamorphic-testing output to automatically identify the optimal configuration. The approach aims to streamline the configuration process and enhance the effectiveness of unsupervised learning systems. It has been evaluated through experiments on six datasets from three domains for anomaly detection. The evaluation results show that our approach can find configurations outperforming the baseline approaches as they achieved a recall of 0.89 and a precision of 0.84 (on average).

show abstract

“…Process shifts and drifts are unexplained or unexpected trends of a measured process parameter(s) away from its intended target value in time-ordered analysis. Our unsupervised data validation pipeline [10] automatically discovers reference patterns representing modes of process behavior in training data from a reference production cycle. Its event detection service tracks deviations (process shifts and drifts) in production data by checking the recurrence of these patterns (see Figure 2(b)).…”

Section: Unsupervised Data Validationmentioning

confidence: 99%

Taming Data Quality in AI-Enabled Industrial Internet of Things

et al. 2022

Self Cite

View full text Add to dashboard Cite

Artificial intelligence (AI)-enabled Industrial Internet of Things (IIoT) marks the rise of systems at the convergence of tremendous amounts of data from multiple IoT devices for complex machine learning/AI software that supports decision making and predictive maintenance in various industries. However, the omnipresent neglect of data quality leads to the accumulation of dark data and the impregnation of biases in AI systems. We address the problem of taming data quality in AI-enabled IIoT systems by devising machine learning pipelines as part of a decentralized edge-to-cloud architecture. These pipelines generate services for (i) erroneous data repair and (ii) unsupervised detection of events and deviations in sensor data. We present the design and deployment of our approach from an AI Engineering perspective using two industrial case studies. THE INDUSTRIAL INTERNET OF THINGS(IIOT) revolutionizes several industries, such as manufacturing, transportation, and energy. It is a major driving force behind Industry 4.0 and employs Artificial Intelligence (AI) techniques, e.g., Machine Learning (ML), to exploit the massive interconnection and large volumes of IIoT data. AI-enabled Industrial IoT systems (IIoTs) improve decision-making [1] and perform predictive maintenance [2] (e.g., tool wear and product defect prediction in the manufacturing domain) in industrial processes. The quality and continuity of IIoT data is a bottleneck and makes these systems rather conservative in what they can achieve. Furthermore, the growing neglect of data quality in AI-enabled IIoTs [3] leads to the accumulation of dark data (unstructured, untagged, and untapped data not analyzed) [4] and the impregnation of biases [5].IIoT data endures a long journey on the edgecloud continuum: (i) data obtained by sensors observing industrial processes is consumed by a rugged industrial computer to control actuators, such as a machine tool in manufacturing; (ii) it is transferred to an edge device over wired/wireless communication channels using industrial communication protocols (e.g., OPC-UA, OPC-DA, NMEA, Bluetooth); and (iii) it is aggregated on edge to be transferred to the cloud using API protocols (e.g., REST, RPC, SOAP, GraphQL). Taming data quality in AI-enabled IIoTs aims to detect and manage data quality issues (bias, freezing, precision degradation, data drift in sensors) on this journey and preserve data continuity on the edge-cloud continuum. Sensor bias is an offset shifting sensor output by a constant value. A sensor freezes when its output is constant in successive samples. Precision degradation occurs IT Professional

show abstract

Udava

Cited by 7 publications

References 31 publications

Replay-Driven Continual Learning for the Industrial Internet of Things

Replay-Driven Continual Learning for the Industrial Internet of Things

AutoConf: Automated Configuration of Unsupervised Learning Systems Using Metamorphic Testing and Bayesian Optimization

Taming Data Quality in AI-Enabled Industrial Internet of Things

Contact Info

Product

Resources

About