Industrial Internet of Things (IIoT) systems generate high-dimensional sensor telemetry from interconnected components, where faults can propagate across the system. To address these challenges, we propose SFlexRCA (Scalable and Flexible Root Cause Analysis), a topology-free RCA framework designed for resource-constrained IIoT environments. SFlexRCA transforms multivariate telemetry into compact orthogonal representations and applies shared lightweight linear modeling, avoiding explicit graph construction, message passing, and per-variable or lag-specific parameter growth. We evaluate SFlexRCA on three publicly available IIoT datasets, BATADAL, SWaT, and WADI, spanning different numbers of monitored variables, temporal characteristics, and training-data regimes. SFlexRCA is compared with 10 statistical, causal, and non-causal baselines in terms of RCA accuracy, training efficiency, inference latency, and memory consumption. In addition, inference efficiency and memory consumption are evaluated on Raspberry Pi 3 and Raspberry Pi 5, while energy consumption is additionally measured on Raspberry Pi 5. We further investigate temporalcontext sensitivity, architectural and loss components, and alternative representations. Notably, SFlexRCA maintains strong localization performance on BATADAL despite its limited normal-operation training data, while its compact shared architecture avoids the parameter growth associated with causal and graph-based approaches. Its lightweight shared architecture further enables efficient deployment on resource-constrained IIoT edge devices. The SFlexRCA code is available at https://github.com/theamrzaki/RootCause- Analysis-Correlation-Attentive-Modeling.
Figures & tables
Method
Telemetry
Architecture
Edge Evaluation
Target
Statistical RCA
ϵ -Diagnosis [ 31 ]
M
Two-sample testing
–
Microservices
RCD [ 32 ]
M
Conditional independence
–
Microservices
BARO [ 33 ]
M
Robust statistical scoring
–
Microservices
Multimodal RCA
ART [ 10 ]
L+M+T
Transformer + GRU
–
Distributed systems
TABLE I : Comparison of Representative Root Cause Analysis and Efficient Time-Series Modeling Approaches.
Fig. 1 : SFlexRCA framework for unsupervised root-cause analysis in IIoT systems.
Method
Complexity
Description
cMLP
O(P2KH)
Separate MLP predictor for each target variable.
cLSTM
O(P2KH+PH2)
Separate recurrent predictor for each target variable.
GVAR
O(P2K)
Explicitly models pairwise variable relationships across lags.
AERCA
O(3KHP2)
Three GVAR-based networks generate a full P×P matrix at each lag.
CUTS+
O(LPH2+PH+P2)
Uses graph-based recurrent message passing with pairwise node interactions.
TABLE II : Complexity comparison of RCA approaches. P is the number of monitored variables, H the hidden dimension, K the temporal order, and L the number of hidden layers.
Metric
BATADAL
SWaT
WADI
System Type
Simulated water distribution (C-Town)
Physical water treatment testbed
Physical water distribution testbed
Sampling Frequency
1 hour
1 second
1 second
Number of Channels
43 sensor/actuator features
51 sensor/actuator features
127 sensor/actuator features
Training Data (Normal)
8,760 samples (365 days)
410,400–480,800 samples (7 days)
∼ 1,200,000 samples (14 days)
Amount of Attacks
14 attacks total (7 train, 7 test)
36–41 attack scenarios (4 days)
15 attack scenarios (2 days)
TABLE III : Comparison of the BATADAL, SWaT, and WADI datasets.
BATADAL (K=16, P=43, training samples=8k)
SWAT (K=8, P=51, training samples=400k)
WADI (K=8, P=127, training samples=1.2M)
Method
Params
AC@1
AC@5
AC@10
auc@10
Params
AC@1
AC@5
AC@10
auc@10
Params
AC@1
AC@5
AC@10
auc@10
Statistical Methods
rcd
0
0.000 ± 0.00
0.086 ± 0.00
0.070 ± 0.00
0.091 ± 0.00
0
0.000 ± 0.00
0.000 ± 0.00
0.300 ± 0.00
0.100 ± 0.00
0
0.000 ± 0.00
0.000 ± 0.00
0.000 ± 0.00
0.000 ± 0.00
baro
0
0.286 ± 0.00
0.162 ± 0.00
0.221 ± 0.00
0.200 ± 0.00
0
0.000 ± 0.00
0.200 ± 0.00
0.250 ± 0.00
0.180 ± 0.00
0
0.267 ± 0.00
0.267 ± 0.00
0.356 ± 0.00
0.283 ± 0.00
torai
0
0.286 ± 0.00
0.178 ± 0.03
0.312 ± 0.04
0.241 ± 0.03
0
0.117 ± 0.03
0.200 ± 0.00
0.350 ± 0.00
0.227 ± 0.01
0
0.267 ± 0.00
0.267 ± 0.00
0.278 ± 0.00
0.269 ± 0.00
Non-Causal Methods
TABLE IV : Root Cause Analysis Accuracy and Model Parameters Across BATADAL and SWAT and WADI (Mean ± SD).
Fig. 2 : Training efficiency metrics (average training time per epoch and peak training memory) across BATADAL, SWaT, and WADI datasets.
Fig. 3 : Edge deployment efficiency metrics (inference latency on Pi 3/5 and energy consumption on Pi 5) across BATADAL, SWaT, and WADI datasets.
BATADAL
SWAT
WADI
Method
Params
AC@1
AC@5
AC@10
auc@10
Params
AC@1
AC@5
AC@10
auc@10
Params
AC@1
AC@5
AC@10
auc@10
Loss Ablations
w/o Sparse
279.4K
0.381 ± 0.08
0.181 ± 0.02
0.247 ± 0.04
0.238 ± 0.03
74.2K
0.217 ± 0.03
0.333 ± 0.03
0.400 ± 0.00
0.338 ± 0.02
84.5K
0.133 ± 0.00
0.278 ± 0.02
0.319 ± 0.01
0.265 ± 0.01
w/o Smooth
279.4K
0.381 ± 0.08
0.181 ± 0.02
0.247 ± 0.04
0.238 ± 0.03
74.2K
0.167 ± 0.03
0.383 ± 0.03
0.433 ± 0.06
0.353 ± 0.01
84.5K
0.133 ± 0.00
0.289 ± 0.02
0.404 ± 0.03
0.294 ± 0.02
w/o KL
279.4K
0.476 ± 0.08
0.162 ± 0.02
0.213 ± 0.01
0.234 ± 0.01
74.2K
0.217 ± 0.03
0.400 ± 0.05
0.433 ± 0.03
0.367 ± 0.04
84.5K
0.133 ± 0.00
0.271 ± 0.01
0.293 ± 0.04
0.266 ± 0.02
Architecture Ablations
TABLE V : Case-Based Ablation Study of Architectural Components (Mean ± SD). Boxed cells ( … ) indicate selected configurations.
Fig. 4 : Sensitivity analysis of RCA accuracy with respect to window size for BATADAL, SWaT and WADI datasets.
BATADAL
SWAT
WADI
Method
AC@1
AC@5
AC@10
auc@10
AC@1
AC@5
AC@10
auc@10
AC@1
AC@5
AC@10
auc@10
Baselines
none
0.429 ± 0.00
0.143 ± 0.03
0.209 ± 0.02
0.210 ± 0.02
0.217 ± 0.06
0.317 ± 0.03
0.400 ± 0.00
0.308 ± 0.01
0.333 ± 0.00
0.289 ± 0.02
0.367 ± 0.00
0.309 ± 0.00
learned
0.381 ± 0.08
0.149 ± 0.03
0.177 ± 0.05
0.193 ± 0.05
0.217 ± 0.03
0.367 ± 0.03
0.417 ± 0.03
0.360 ± 0.02
0.333 ± 0.00
0.289 ± 0.02
0.367 ± 0.00
0.308 ± 0.00
Static Projections
legendre
0.381 ± 0.08
0.152 ± 0.04
0.178 ± 0.03
0.204 ± 0.03
0.200 ± 0.00
0.333 ± 0.03
0.400 ± 0.05
0.318 ± 0.01
0.333 ± 0.00
0.289 ± 0.02
0.367 ± 0.00
0.307 ± 0.01
TABLE VI : Root Cause Analysis Accuracy Across Different Projections For BATADAL and SWAT and WADI (Mean ± SD).
Fig. 5 : t-SNE latent feature projections on BATADAL corresponding to baseline methods, static polynomial/frequency projections, and our proposed orthogonal projection (matching Table VI ).
Deploying large language models (LLMs) on Industrial Internet of Things (IIoT) edge devices demands extreme compression, yet existing structured pruning methods collapse at high compression ratios due to one-shot importance estimation, and their cross-architecture behavior remains unpredictable. This article presents a cascaded multi-granularity pruning framework that removes layers, attention heads, and feed-forward channels in coarse-to-fine order, with lightweight low-rank recovery between stages to re-estimate component importance. An information-theoretic analysis motivates this ordering, and the Structural Independence Assumption (SIA) is formalized as a checkable condition predicting whether per-component pruning criteria are reliable for a given architecture: Multi-Head Attention (MHA)+GELU designs satisfy the SIA, whereas Grouped Query Attention (GQA)+SwiGLU designs violate it. On bearing fault diagnosis spanning 88M to 6.25B-parameter models, the framework extends achievable compression to 13.8 times on MHA+GELU architectures with 83.82% accuracy (+3.70 percentage points (pp) over the strongest baseline), while exposing a ~74pp accuracy collapse on GQA+SwiGLU architectures that violate the SIA. Deployed on an industrial slewing bearing fault diagnosis platform with NVIDIA DGX Spark, compressed models reduce inference latency by up to 67.2% and peak memory by 62.5%, demonstrating viability for IIoT edge inference.
Jinghan Wang, Yanjun Chen, Wei Zhang +3
School of Mechanical and Electrical Engineering, Harbin Institute of Technology, Harbin 150001, China · School of Information Science and Technology, Eastern Institute of Technology, Ningbo, China
In the Industrial Internet of Things (IIoT) systems, edge devices often operate under strict constraints in memory, compute capability, and wireless bandwidth. These limitations challenge the deployment of advanced data analytics tasks, such as predictive and prescriptive maintenance. In this work, we explore hyperdimensional computing (HDC) as a lightweight learning paradigm for resource-constrained IIoT. Conventional centralized HDC leverages the properties of high-dimensional vector spaces to enable energy-efficient training and inference. We integrate this paradigm into a federated learning (FL) framework where devices exchange only prototype representations, which significantly reduces communication overhead. Our numerical results highlight the potential of federated HDC to support collaborative learning in IIoT with fast convergence speed and communication efficiency. These results indicate that HDC represents a lightweight and resilient framework for distributed intelligence in large-scale and resource-constrained IIoT environments.
Nikita Zeulin, Olga Galinina, Nageen Himayat +1
Tampere University, Tampere, Finland · Tampere Institute for Advanced Study, Tampere University, Finland · Intel Corporation, Santa Clara, CA, United States
Root cause analysis (RCA) in networked industrial systems, such as supply chains and power networks, is notoriously difficult due to unknown and dynamically evolving interdependencies among geographically distributed clients. These clients represent heterogeneous physical processes and industrial assets equipped with sensors that generate large volumes of nonlinear, high-dimensional, and heterogeneous IoT data. Classical RCA methods require partial or full knowledge of the system's dependency graph, which is rarely available in these complex networks. While federated learning (FL) offers a natural framework for decentralized settings, most existing FL methods assume homogeneous feature spaces and retrainable client models. These assumptions are not compatible with our problem setting. Different clients have different data features and often run fixed, proprietary models that cannot be modified. This paper presents a federated cross-client interdependency learning methodology for feature-partitioned, nonlinear time-series data, without requiring access to raw sensor streams or modifying proprietary client models. Each proprietary local client model is augmented with a Machine Learning (ML) model that encodes cross-client interdependencies. These ML models are coordinated via a global server that enforces representation consistency while preserving privacy through calibrated differential privacy noise. RCA is performed using model residuals and anomaly flags. We establish theoretical convergence guarantees and validate our approach on extensive simulations and a real-world industrial cybersecurity dataset.
Ayush Mohanty, Paritosh Ramanan, Nagi Gebraeel
School of Industrial and Systems Engineering, Georgia Institute of Technology, Atlanta, GA 30332 · School of Industrial Engineering and Management, Oklahoma State University, Stillwater, OK 74708