ECNN: A Low-complex, Adjustable CNN for Industrial Pump Monitoring Using Vibration Data

2025-03-10

AI Summary

To address the dual challenges of limited computational resources and poor cross-pump generalization in deploying industrial pump fault预警 on edge devices, this paper proposes a lightweight, configurable convolutional neural network (ECNN). ECNN employs a low-complexity CNN architecture, integrates pump-specific learnable parameters, and incorporates a statistical threshold fusion mechanism to enable rapid adaptation under small-sample conditions. The design jointly optimizes temporal modeling capability for vibration signals and inference efficiency, supporting real-time, on-device fault detection. Experimental results demonstrate that ECNN significantly outperforms conventional statistical methods and standard CNNs in detection accuracy, while reducing model size by over 60% and achieving inference latency below 15 ms. Thus, ECNN achieves a practical balance among high robustness, strong cross-pump generalization, and feasibility for edge deployment.

Technology Category

• Machine Learning: Learning on the Edge & Model Compression
• Intelligent Robots: State Estimation
• Knowledge Representation and Reasoning: Diagnosis and Abductive Reasoning

Application Category

• Systems and Infrastructure for Web, Mobile and WoT: Experiences and lessons learnt from Web-based algorithms and system deployments
• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs
• Web Mining and Content Analysis: Robustness and generalizability of Web mining methods

Abstract

Industrial pumps are essential components in various sectors, such as manufacturing, energy production, and water treatment, where their failures can cause significant financial and safety risks. Anomaly detection can be used to reduce those risks and increase reliability. In this work, we propose a novel enhanced convolutional neural network (ECNN) to predict the failure of an industrial pump based on the vibration data captured by an acceleration sensor. The convolutional neural network (CNN) is designed with a focus on low complexity to enable its implementation on edge devices with limited computational resources. Therefore, a detailed design space exploration is performed to find a topology satisfying the trade-off between complexity and accuracy. Moreover, to allow for adaptation to unknown pumps, our algorithm features a pump-specific parameter that can be determined by a small set of normal data samples. Finally, we combine the ECNN with a threshold approach to further increase the performance and satisfy the application requirements. As a result, our combined approach significantly outperforms a traditional statistical approach and a classical CNN in terms of accuracy. To summarize, this work provides a novel, low-complex, CNN-based algorithm that is enhanced by classical methods to offer high accuracy for anomaly detection of industrial pumps.

Problem

• Develops a low-complexity CNN for industrial pump failure prediction.
• Enables implementation on edge devices with limited computational resources.
• Enhances accuracy by combining ECNN with a threshold approach.

Innovation

• Low-complex CNN for edge device implementation
• Pump-specific parameter adaptation using minimal data
• Combined ECNN and threshold for enhanced accuracy