Bolted Joint Monitoring Using Impedance Signatures and Autoencoders
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for monitoring bolted joints in constructions require complex calibration and classification steps, making them expensive and inefficient.
Innovation Solution
A method using an unsupervised machine learning algorithm, specifically an autoencoder, to analyze the electromechanical impedance signatures from sensor elements embedded in washers, nuts, or bolts, for anomaly detection in bolted joints, allowing for cost- and work-efficient monitoring.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If complex calibration and classification steps are used to monitor bolted joints, then measurement precision is improved, but device complexity and cost increase
Solution Approach 1:
The patent extracts and eliminates the complex calibration and classification steps from the monitoring system. By using unsupervised machine learning algorithms that automatically adapt to environmental conditions and joint characteristics, the system removes the need for manual calibration procedures while maintaining high measurement precision in detecting joint anomalies.
Solution Approach 2:
The monitoring system performs self-calibration through unsupervised learning mechanisms. The algorithm automatically adapts to the specific characteristics of each joint and environmental conditions without requiring external calibration inputs, enabling the system to serve itself and eliminate manual calibration requirements.
2Reliability
If extensive calibration across various torque and environmental conditions is performed, then reliability is improved, but loss of time and productivity decrease
Solution Approach 1:
The unsupervised machine learning algorithm performs preliminary adaptation automatically when deployed, learning the baseline characteristics of the joint under current environmental conditions without requiring pre-calibration across multiple conditions. This preliminary self-adjustment ensures reliability while eliminating time-consuming calibration procedures.
Solution Approach 2:
The system dynamically adapts to changing environmental conditions and joint characteristics in real-time through continuous unsupervised learning. Rather than requiring static pre-calibration for various conditions, the algorithm dynamically adjusts its parameters to maintain reliable monitoring across different operating conditions.
3Measurement precision
If multiple sensors are deployed in washers, nuts, or bolts, then measurement precision is improved, but device complexity and cost increase
Solution Approach 1:
The patent implements a universal sensor system where a single sensor type can be deployed in washers, nuts, or bolts interchangeably. The unsupervised machine learning algorithm processes signals from any sensor location uniformly, eliminating the need for location-specific calibration or multiple sensor types, thereby reducing system complexity while maintaining detection accuracy.
Solution Approach 2:
The patent merges the calibration and classification functions into the unsupervised machine learning algorithm itself. Rather than requiring separate calibration hardware or software modules for each sensor location, the algorithm integrates these functions, processing signals from multiple sensor locations through a unified approach that reduces overall system complexity.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
Enables real-time, sensitive, and accurate monitoring of bolted joints with reduced maintenance downtime, eliminating the need for extensive calibration and signal acquisition across various torque and environmental conditions.
Implementation Method 1
KR 10-2018-0023618A discloses a method for detecting loosening of a bolt and a nut by a pressure sensor transforming pressure into electrical capacitance
Implementation Method 2
Linsheng Huo et al (Smart Materials and Structures, 26:057004, 2014) discloses an electro-mechanical impedance method to investigate the pre-load condition on a bolted connection using a 'smart washer'
Data Source
Figure 1
Figure 2
Figure 3
AI summary
The present invention relates to a method for monitoring an installation comprising joints wherein one or more of the joints comprise a washer and/or a nut and/or a bolt containing a sensor element, the method comprising the steps of: (I) sending an electrical signal defined on a selected frequency range to the sensor element, (II) receiving the electrical response signal from the sensor element and processing the signal to obtain the impedance signature of the coupled system, (III) feeding the impedance signature signal and/or some or all of its features to an unsupervised machine learning algorithm and processing of the fed data by the unsupervised machine learning algorithm, (IV) detecting and reporting anomalous conditions of the monitored installation, and to a system implementing said method.