Student or educator? Sign up with your school email and get 6 months of Pro free.
CoChat

The DOI matches its Crossref record; 63 of the 67 references it cites were found in a registry or an index; 7 of the 18 sources its claims cite were checked for support.

Edge-Cloud Synergy for AI-Enhanced Sensor Network Data: A Real-Time Predictive Maintenance Framework

Open access
Sathupadi, Kaushik, Achar, Sandesh, Bhaskaran, Shinoy Vengaramkode, Faruqui, Nuruzzaman, Abdullah-Al-Wadud, M., Uddin, Jia · 2024 · Sensors

Abstract

Sensor networks generate vast amounts of data in real-time, which challenges existing predictive maintenance frameworks due to high latency, energy consumption, and bandwidth requirements. This research addresses these limitations by proposing an edge-cloud hybrid framework, leveraging edge devices for immediate anomaly detection and cloud servers for in-depth failure prediction. A K-Nearest Neighbors (KNNs) model is deployed on edge devices to detect anomalies in real-time, reducing the need for continuous data transfer to the cloud. Meanwhile, a Long Short-Term Memory (LSTM) model in the cloud analyzes time-series data for predictive failure analysis, enhancing maintenance scheduling and operational efficiency. The framework's dynamic workload management algorithm optimizes task distribution between edge and cloud resources, balancing latency, bandwidth usage, and energy consumption. Experimental results show that the hybrid approach achieves a 35% reduction in latency, a 28% decrease in energy consumption, and a 60% reduction in bandwidth usage compared to cloud-only solutions. This framework offers a scalable, efficient solution for real-time predictive maintenance, making it highly applicable to resource-constrained, data-intensive environments.

Overview

CoChat Summary

## TL;DR

This paper proposes a hybrid edge-cloud predictive maintenance framework for sensor networks, using KNN on a Raspberry Pi edge device for real-time anomaly detection and LSTM in the cloud for failure prediction. The reported system reduces latency by 35%, energy consumption by 28%, and bandwidth usage by 60% versus a cloud-only baseline.

## Key findings

  • The hybrid approach achieved a 35% reduction in latency and an average end-to-end latency of 500 ms, compared with a cloud-only baseline of 800 ms.
  • It produced a 28% decrease in energy consumption compared with the cloud-only solution, and the paper also reports 60% energy savings relative to edge-only processing under low data flow conditions.
  • Bandwidth usage was reduced by 60% compared with the cloud-only model; the paper also reports 70% less total data transmission than sending all raw data directly to the cloud.
  • For edge anomaly detection, the model was evaluated on 18,452 test instances; for the latency study, the paper says the summary was prepared from 20 discrete experiments.

## Methods

  • The study uses a hybrid edge-cloud design: KNN runs on the edge for anomaly detection, and LSTM runs in the cloud for time-series failure prediction.
  • Data come from an industrial sensor network with 123,013 instances and 20 sensors; the paper says only 20 sensors were accessible in the packaging-industry deployment, and the data were split 70:15:15 into train/validation/test using a temporal split.
  • Analysis included grid search and cross-validation for KNN parameters (selected k = 5 and a 30-time-step window), 10-fold cross-validation for both edge and cloud models, and experimental comparison against cloud-only, fog-related, and other hybrid approaches.

## Limitations

  • The paper states the current framework has no exception handling if the cloud server fails to respond; it assumes the cloud system provides uninterrupted service.
  • The edge device is a Raspberry Pi Zero 2 W, and the paper says this creates an extendability issue and limits workload distribution because of constrained computational capability.
  • The text also notes limits evident from the setup: evaluation is based on one industrial packaging environment with 20 accessible sensors, so broader generalization is not directly demonstrated.

Model-written summary. Check it against the paper before relying on it.

Integrity signals

Registry

Crossref · 10.3390/s24247918

Cites

63 of 63 references verified0 citation errors · 0 not checkable

Statistics

Nothing to recompute.

Figures

Figure 1 Impact of k on KNN classification accuracy.
Figure 1 Impact of k on KNN classification accuracy.CC BY · PubMed Central
Figure 2 The sensor locations on four different types of machines.
Figure 2 The sensor locations on four different types of machines.CC BY · PubMed Central
Figure 3 The learning curve analysis of the KNN anomaly detector. The graph is divided into three distinct regions: (a) rapid improvement region, where the accuracy improves significantly with increasing training data; (b) stable region, where the rate of improvement slows down; and (c) convergence region, where accuracy plateaus, indicating diminishing returns with additional data. This division highlights the efficiency of the KNN model in achieving high performance with a moderately sized dataset.
Figure 3 The learning curve analysis of the KNN anomaly detector. The graph is divided into three distinct regions: (a) rapid improvement region, where the accuracy improves significantly with increasing training data; (b) stable region, where the rate of improvement slows down; and (c) convergence region, where accuracy plateaus, indicating diminishing returns with additional data. This division highlights the efficiency of the KNN model in achieving high performance with a moderately sized dataset.CC BY · PubMed Central
Figure 4 The learning curve analysis of the LSTM failure prediction model. The graph shows three key regions: (a) rapid improvement region, where both training and validation losses decrease substantially; (b) stable region, where the rate of decrease slows, showing consistency; and (c) convergence region, where losses reach minimal levels, indicating a well-trained model with low overfitting risk. These regions emphasize the LSTM’s effective learning and generalization capabilities.
Figure 4 The learning curve analysis of the LSTM failure prediction model. The graph shows three key regions: (a) rapid improvement region, where both training and validation losses decrease substantially; (b) stable region, where the rate of decrease slows, showing consistency; and (c) convergence region, where losses reach minimal levels, indicating a well-trained model with low overfitting risk. These regions emphasize the LSTM’s effective learning and generalization capabilities.CC BY · PubMed Central
Figure 5 The workflow of the implemented system.
Figure 5 The workflow of the implemented system.CC BY · PubMed Central
Figure 6 Confusion matrix analysis of the anomaly prediction by KNN.
Figure 6 Confusion matrix analysis of the anomaly prediction by KNN.CC BY · PubMed Central
Figure 7 The consistency of performance in all evaluation metrics in k-fold cross-validation.
Figure 7 The consistency of performance in all evaluation metrics in k-fold cross-validation.CC BY · PubMed Central
Figure 8 The consistency of MAE and RMSE in the LSTM network in k-fold cross-validation.
Figure 8 The consistency of MAE and RMSE in the LSTM network in k-fold cross-validation.CC BY · PubMed Central
Figure 9 Latency performance of the proposed edge-cloud framework across 20 experiments, showing reductions in end-to-end latency compared to a cloud-only baseline along with other parameters.
Figure 9 Latency performance of the proposed edge-cloud framework across 20 experiments, showing reductions in end-to-end latency compared to a cloud-only baseline along with other parameters.CC BY · PubMed Central
Figure 10 Energy consumption performance of the edge-cloud framework across 20 experiments, showing energy savings compared to a cloud-only baseline.
Figure 10 Energy consumption performance of the edge-cloud framework across 20 experiments, showing energy savings compared to a cloud-only baseline.CC BY · PubMed Central
Figure 11 Bandwidth usage performance of the edge-cloud framework across 20 experiments, showing substantial reductions in bandwidth consumption compared to a cloud-only baseline.
Figure 11 Bandwidth usage performance of the edge-cloud framework across 20 experiments, showing substantial reductions in bandwidth consumption compared to a cloud-only baseline.CC BY · PubMed Central

Verification

Claims the paper makes

4 Valid · 3 Not valid · 8 Not checked · 66 Own results and interpretations

38 more claims

Create a free account to see all 81 claims and to verify the 1 reference still open.

See all claims — free account
  • Probably not a substantial claim. 19
  • It may state two separate claims as one. 20
  • The quote only partly states this claim. 2

Create a free account to see these claims.

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

Self-check results appear here once an independent check has confirmed them.

Cited works that were retracted

Works this paper cites that were later retracted, from the retraction notices publishers deposit. Major when the notice came out before this paper was published.

1 of 1

Citations

Cites

1 retracted

Show all 67 references

Cited by

See all 89

Metadata

Authors
Sathupadi, K., Achar, S., Bhaskaran, S. V., Faruqui, N., Abdullah-Al-Wadud, M., & Uddin, J.
Affiliations
  • Google (United States) (US)
  • Department of Software Engineering, Walmart Global Tech, Sunnyvale, CA 94086, USA
  • Zoom Video Communications, Sanjose, CA 95113, USA
  • Daffodil International University (BD)
  • King Saud University (SA)
  • Woosong University (KR)
Corresponding author
  • Abdullah-Al-Wadud, M.
  • Uddin, Jia
Venue
Sensors
Publisher
MDPI AG
Published
2024-12-11
Volume
24
Issue
24
Pages
7918
ISSN
1424-8220 (ISSN-L)
OpenAlex
W4405284906
Cited by
89
References
67
Citations by year
2026: 52 · 2025: 37
FWCI
14.73
Type
journal article
Language
en
Open access
Locations
  • Sensors (publishedVersion, cc-by)
  • PubMed Central (publishedVersion, cc-by)
  • PubMed (publishedVersion)
  • DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion)
Funders
Researchers Supporting Project, King Saud University, Saudi Arabia (RSPD2024R951)
Topics
  • Physical Sciences › Computer Science › Computer Networks and Communications › IoT and Edge/Fog Computing
  • Physical Sciences › Engineering › Safety, Risk, Reliability and Quality › Reliability and Maintenance Optimization
  • Physical Sciences › Computer Science › Information Systems › Internet of Things and AI
Keywords
predictive maintenance, sensor networks, anomaly detection, failure prediction, k-nearest neighbors, long short-term memory, time series analysis, workload management, latency reduction, energy consumption, bandwidth reduction, maintenance scheduling

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

The claim verdicts this report counts are experimental: their accuracy is not yet measured.

A language model drew the claims from the paper on 2026-10-09, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).

A second model labelled each claim on 2026-10-09 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).

Cited sources are checked with a language model’s help (gpt-6-sol).

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review

How to cite

Cite CoChat for the checks on this page, and the paper for its own findings.

This page
CoChat. (2026). Citation check: Edge-Cloud Synergy for AI-Enhanced Sensor Network Data: A Real-Time Predictive Maintenance Framework. https://app.cochat.ai/papers/10.3390/s24247918
The paper
Sathupadi, K., Achar, S., Bhaskaran, S. V., Faruqui, N., Abdullah-Al-Wadud, M., & Uddin, J. (2024). Edge-Cloud Synergy for AI-Enhanced Sensor Network Data: A Real-Time Predictive Maintenance Framework. Sensors. https://doi.org/10.3390/s24247918

Data sources

Bibliographic data: Semantic Scholar, OpenAlex, Crossref.

Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).

Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.