A hybrid data-fusion system using modal data and probabilistic neural network for damage detection

2011 ◽  
Vol 42 (6) ◽  
pp. 368-374 ◽  
Author(s):  
Shao-Fei Jiang ◽  
Chun Fu ◽  
Chunming Zhang
Keyword(s):  
Neural Network ◽  
Data Fusion ◽  
Damage Detection ◽  
Probabilistic Neural Network ◽  
Fusion System ◽  
Modal Data ◽  
Hybrid Data

scholarly journals A Hybrid Data-Fusion System by Integrating CFD and PNN for Structural Damage Identification

2021 ◽  
Vol 11 (17) ◽  
pp. 8272
Author(s):  
Chun Fu ◽  
Shao-Fei Jiang
Keyword(s):  
Neural Network ◽  
Data Fusion ◽  
Structural Damage ◽  
Damage Identification ◽  
Fusion System ◽  
Fusion Model ◽  
Decision Level ◽  
Hybrid Data ◽  
Level Model ◽  
Level Fusion

Recently, a variety of intelligent structural damage identification algorithms have been developed and have obtained considerable attention worldwide due to the advantages of reliable analysis and high efficiency. However, the performances of existing intelligent damage identification methods are heavily dependent on the extracted signatures from raw signals. This will lead to the intelligent damage identification method becoming the optimal solution for actual application. Furthermore, the feature extraction and neural network training are time-consuming tasks, which affect the real-time performance in identification results directly. To address these problems, this paper proposes a new intelligent data fusion system for damage detection, combining the probabilistic neural network (PNN), data fusion technology with correlation fractal dimension (CFD). The intelligent system consists of three modules (models): the eigen-level fusion model, the decision-level fusion model and a PNN classifier model. The highlight points of this system are these three intelligent models specialized in certain situations. The eigen-level model is specialized in the case of measured data with enormous samples and uncertainties, and for the case of confidence level of each sensor is determined ahead, the decision-level model is the best choice. The single PNN model is considered only when the data collected is somehow limited, or few sensors have been installed. Numerical simulations of a two-span concrete-filled steel tubular arch bridge in service and a seven-storey steel frame in laboratory were used to validate the hybrid system by identifying both single- and multi-damage patterns. The results show that the hybrid data-fusion system has excellent performance of damage identification, and also has superior capability of anti-noise and robustness.

Rock hardness identification based on optimized PNN and multi-source data fusion

2021 ◽  
pp. 095440622110420
Author(s):  
Ying He ◽  
Muqin Tian ◽  
Jiancheng Song ◽  
Junling Feng
Keyword(s):  
Neural Network ◽  
Data Fusion ◽  
Probabilistic Neural Network ◽  
Wavelet Packet ◽  
Differential Evolution Algorithm ◽  
Vibration Signal ◽  
Optimization Method ◽  
Rock Wall ◽  
Rock Hardness ◽  
Source Data

To solve the problem that it is difficult to identify the cutting rock wall hardness of the roadheader in coal mine, a recognition method of cutting rock wall hardness is proposed based on multi-source data fusion and optimized probabilistic neural network. In this method, all kinds of cutting signals (the vibration signal of cutting arm, the pressure signal of hydraulic cylinders and current signal of cutting motor) are analyzed by wavelet packet to extract the feature vector, and the multi feature signal sample database of rock cutting with different hardness is established. To solve the problems of uncertain spread and complex network structure of probabilistic neural network (PNN), a PNN optimization method based on differential evolution algorithm (DE) and QR decomposition was proposed, and the rock hardness was identified based on multi-source data fusion by optimizing PNN. Then, based on the ground test monitoring data of a heavy longitudinal roadheader, the method is applied to recognize the cutting rock hardness, and compared with other common pattern recognition methods. The experimental results show that the cutting rock hardness recognition based on multi-source data fusion and optimized PNN has higher recognition accuracy, and the overall recognition error is reduced to 6.8%. The recognition of random cutting rock hardness is highly close to the actual. The method provides theoretical basis and technical premise for realizing automatic and intelligent cutting of heading face.

Damage Detection of Composite Materials Using Data Fusion With Deep Neural Networks

2020 ◽  
Author(s):  
Shweta Dabetwar ◽  
Stephen Ekwaro-Osire ◽  
João Paulo Dias
Keyword(s):  
Neural Network ◽  
Neural Networks ◽  
Composite Materials ◽  
Data Fusion ◽  
Damage Detection ◽  
Convolutional Neural Network ◽  
Deep Neural Networks ◽  
Mixed Data ◽  
Image Encoding ◽  
Using Data

Abstract Composite materials have enormous applications in various fields. Thus, it is important to have an efficient damage detection method to avoid catastrophic failures. Due to the existence of multiple damage modes and the availability of data in different formats, it is important to employ efficient techniques to consider all the types of damage. Deep neural networks were seen to exhibit the ability to address similar complex problems. The research question in this work is ‘Can data fusion improve damage classification using the convolutional neural network?’ The specific aims developed were to 1) assess the performance of image encoding algorithms, 2) classify the damage using data from separate experimental coupons, and 3) classify the damage using mixed data from multiple experimental coupons. Two different experimental measurements were taken from NASA Ames Prognostic Repository for Carbon Fiber Reinforced polymer. To use data fusion, the piezoelectric signals were converted into images using Gramian Angular Field (GAF) and Markov Transition Field. Using data fusion techniques, the input dataset was created for a convolutional neural network with three hidden layers to determine the damage states. The accuracies of all the image encoding algorithms were compared. The analysis showed that data fusion provided better results as it contained more information on the damages modes that occur in composite materials. Additionally, GAF was shown to perform the best. Thus, the combination of data fusion and deep neural network techniques provides an efficient method for damage detection of composite materials.

DATA FUSION-BASED STRUCTURAL DAMAGE DETECTION UNDER VARYING TEMPERATURE CONDITIONS

2012 ◽  
Vol 12 (06) ◽  
pp. 1250052 ◽  
Author(s):  
YUEQUAN BAO ◽  
YONG XIA ◽  
HUI LI ◽  
YOU-LIN XU ◽  
PENG ZHANG
Keyword(s):  
Data Fusion ◽  
Damage Detection ◽  
Temperature Effect ◽  
Test Data ◽  
Mode Shapes ◽  
Detection Accuracy ◽  
Modal Data ◽  
Temperature Conditions ◽  
Uncertainty Measurement ◽  
Structural Responses

A huge number of data can be obtained continuously from a number of sensors in long-term structural health monitoring (SHM). Different sets of data measured at different times may lead to inconsistent monitoring results. In addition, structural responses vary with the changing environmental conditions, particularly temperature. The variation in structural responses caused by temperature changes may mask the variation caused by structural damages. Integration and interpretation of various types of data are critical to the effective use of SHM systems for structural condition assessment and damage detection. A data fusion-based damage detection approach under varying temperature conditions is presented. The Bayesian-based damage detection technique, in which both temperature and structural parameters are the variables of the modal properties (frequencies and mode shapes), is developed. Accordingly, the probability density functions of the modal data are derived for damage detection. The damage detection results from each set of modal data and temperature data may be inconsistent because of uncertainties. The Dempster–Shafer (D–S) evidence theory is then employed to integrate the individual damage detection results from the different data sets at different times to obtain a consistent decision. An experiment on a two-story portal frame is conducted to demonstrate the effectiveness of the proposed method, with consideration on model uncertainty, measurement noise, and temperature effect. The damage detection results obtained by combining the damage basic probability assignments from each set of test data are more accurate than those obtained from each test data separately. Eliminating the temperature effect on the vibration properties can improve the damage detection accuracy. In particular, the proposed technique can detect even the slightest damage that is not detected by common damage detection methods in which the temperature effect is not eliminated.

Prediction of Unmeasured Mode Shape Using Artificial Neural Network for Damage Detection

2013 ◽  
Vol 61 (1) ◽  
Author(s):  
L. D. Goh ◽  
N. Bakhary ◽  
A. A. Rahman ◽  
B. H. Ahmad
Keyword(s):  
Neural Network ◽  
Damage Detection ◽  
Mode Shape ◽  
Health Monitoring ◽  
Spline Interpolation ◽  
Measured Data ◽  
Mode Shapes ◽  
Modal Data ◽  
Artificial Neural ◽  
Shape Data

Artificial neural networks (ANNs) have received much attention in the field of vibration–based damage detection since the 1990s, due to their capability to predict damage from modal data. However, the accuracy of this method is highly dependent on the number of measurement points, especially when the mode shape is used as an indicator for damage detection. With a high number of measurement points, more information can be fed to the ANN to detect damage; therefore, more reliable results can be obtained. Nevertheless, in practice, it is uneconomical to install sensors on every part of a structure; thus the capability of ANNs to detect damage is quite limited. In this study, an ANN is applied to predict the unmeasured mode shape data based on a limited number of measured data. To demonstrate the accuracy of the proposed method, the results are compared with the Cubic Spline interpolation (CS) method. A parametric study is also conducted to investigate the sensitivity of the number of measurement points to the proposed method. The results show that the ANN provides more reliable results compared to the CS method as it is able to predict the magnitude of mode shapes at the unmeasured points with a limited number of measurement points. The application of a two–stage ANN showed results with a high potential for overcoming the issue of using a limited number of sensors in structural health monitoring.

Sign in / Sign up

Export Citation Format

Share Document