Percussion-based vibration and audio signal analysis for structural health monitoring of bolted joints

Current switch bolt inspection on rail systems is a labor intensive and sometimes unreliable approach to maintaining the switch integrity. Recent rail accidents in the United Kingdom (Potters Bar in 2002 and Grayrigg in 2007) underscore the need for routine inspections of the switch mechanisms. From the Grayrigg report of 23 February 2007 the main causes of the accident were found to be the loosening and, as a result, the initiation and growth of cracks, and, eventually, rupture of the bolts of the switch bars, especially the one maintaining the switch rails at a correct distance apart. Such findings also resulted from the 2002 crash report but unfortunately frequent visual inspections were not forthcoming. In this paper, an effective method for monitoring the loosening of the switch bolts is described. As the loosening of the bolts further causes the crack formation in the bolted joints, it seems valid to say that the early detection of loosening of bolted joints in railroad switches will be of great importance in eliminating the need for frequent visual inspection by totally automating inspection of the switches’ mechanical condition. The first part of the present paper focuses on the use of smart materials and structures for the health monitoring of bolted joints in railroad switches. It is shown that using the piezoelectric transducers and the impedance-based structural health monitoring technique, the loosening of the bolted joints are detectable. The accuracy in loosening detection is as high as 25 ft-lbs which corresponds to merely 1/10th of a bolt turn. Being able to detect the loosening of the bolted joints in railroad switches, the concept of self-healing bolted joints is applied in the next part in order to automatically retighten the loosened bolts to their prescribed functional conditions.

This paper presents the use of audio and vibration signals in fault diagnosis of a circulating pump. The novelty of this paper is the use of audio signals acquired by microphones. The objective of this paper is to determine if audio signals are capable to distinguish between normal and different abnormal conditions in a circulating pump. In order to compare results, vibration signals are also acquired and analysed. Wavelet package is used to obtain the energies in different frequency bands from the audio and vibration signals. Neural networks are used to evaluate the discrimination ability of the extracted features between normal and fault conditions. The results show that information from sound signals can distinguish between normal and different faulty conditions with a success rate of 83.33%, 98% and 91.33% for each microphone respectively. These success rates are similar and even higher that those obtained from accelerometers (68%, 90.67% and 71.33% for each accelerometer respectively). Success rates also show that the position of microphones and accelerometers affects on the final results.

Classification and regression models of audio and vibration signals for machine state monitoring in precision machining systems

Condition monitoring of rolling bearing using bearing's audio signal and vibration signal via cost-effective accelerometer is experimented with and analyzed for both localized faults and distributed/ generalized roughness faults. Even though the Fast Fourier Transform (FFT) of the bearing's audio signal is not appropriate to diagnose bearing faults under varying background noises it was possible to observe characteristic frequency hikes related to misalignment from the FFT of the vibration signal. Localized faults are processed using Kurstogram, and Hilbert transform. Misalignment experiments for different types of bearings at different speeds and related fault frequencies are identified through both bearing audio signal and vibration signal. Artificial Neural Network (ANN) and Convolutional Neural Network (CNN) are trained using the Mel Frequency Cepstral Coefficient (MFCC) feature of bearing audio signals with various background noises and/or the FFT of the vibration signal to distinguish between healthy bearing and bearing having distributed /generalized roughness faults. The ANN and CNN models trained using fused MFCC of bearing audio signal and FFT of the vibration signal yield greater accuracy than that of models trained using MFCC of the audio signal or trained using FFT of the vibration signal

The traditional methods of structural health monitoring (SHM) have obvious disadvantages such as being time-consuming, laborious and non-synchronizing, and so on. This paper presents a novel and efficient approach to detect structural damages from real-time vibration signals via a convolutional neural network (CNN). As vibration signals (acceleration) reflect the structural response to the changes of the structural state, hence, a CNN, as a classifier, can map vibration signals to the structural state and detect structural damages. As it is difficult to obtain enough damage samples in practical engineering, finite element analysis (FEA) provides an alternative solution to this problem. In this paper, training samples for the CNN are obtained using FEA of a steel frame, and the effectiveness of the proposed detection method is evaluated by inputting the experimental data into the CNN. The results indicate that, the detection accuracy of the CNN trained using FEA data reaches 94% for damages introduced in the numerical model and 90% for damages in the real steel frame. It is demonstrated that the CNN has an ideal detection effect for both single damage and multiple damages. The combination of FEA and experimental data provides enough training and testing samples for the CNN, which improves the practicability of the CNN-based detection method in engineering practice.

Additive manufacturing (AM) was originally developed to manufacture polymer prototypes. Today, it has been used for the manufacturing of many critical machine components. Most of the structural health monitoring (SHM) methods were developed for monitoring the condition of large and thin plates on airplane fuselages. Additively manufactured parts are generally small, thick, and have complex geometries. SHM methods have been improved to sense load, detect defects, and identify loose bolts with the help of a permanently installed sensor. In this study, the adaptability of SHM methods was researched with additively manufactured metal parts with complex geometry. Magnets were used to apply pressure to 9 different locations on the surface of a stainless steel additively manufactured thick plate with deep groves. SHM was used to estimate the magnets’ location. Many SHM (Lamb wave) methods cannot work on smaller parts since their dimensions are shorter or very close to the wavelength of the created oscillations. Surface response to excitation (SuRE) method which has similar characteristics to electromechanical impedance methods was used for data collection. To obtain descriptive features of the time domain data, fast Fourier transformation (FFT), short-time Fourier transformation (STFT), continuous wavelet transformation (CWT), and synchrosqueezing transform (SST) were applied to CWT. 1D and 2D convolutional neural networks (CNN) were used to classify the cases. When CNN was optimized for the analysis of our data, 100% location estimation accuracy was obtained by using 50% of 320 scalograms for training. The scalograms were obtained by enhancing the CWT results with SST. STFT-CNN combination was the second best. It obtained 95% accuracy with the same number of spectrograms and training allocation.

PurposeThe objective of this article is to evaluate and compare the performance of two machine learning (ML) algorithms, i.e. support vector machines (SVMs) and random forests (RFs), when classifying seven states of operation of an electric motor using the Mel-frequency cepstral coefficients (MFCCs) as extracted representative features.Design/methodology/approachThe extracted MFCCs are calculated using the motor’s vibration and audio signals separately.FindingsAfter the training, the SVM model obtained a mean accuracy of 100% for the MFCCs obtained from database vibration signals and 69.6% for the database of audio signals.Research limitations/implicationsThe ML strategies and results reported are limited to the well-known data for industrial electric motors used in the evaluations, although it was performed tests and cross-validations with unseen data and the information from the confusion matrix.Practical implicationsThe success of these methodologies in defect classification, where the RF presented a mean accuracy of 99.15% for the vibration signals and 63.82% for the audio signal, enables the use of this ML and extracted features as a predictive tool for failure and anomaly detection, lifetime predictions and online real-time monitoring.Originality/valueIt is the first time that the MFCCs are being used for anomaly detection in vibration and audio signals for electrical motors, as this extracted feature is usually used for human speech identification in the literature.

Health monitoring of bolted joints via electrical conductivity measurements

Fault degradation assessment of water hydraulic motor by impulse vibration signal with Wavelet Packet Analysis and Kolmogorov–Smirnov Test

This study focuses on developing and implementing Mamdani hybrid fuzzy logic inference system (FIS) for transverse crack detection and fault diagnosis in a woven fiber laminated glass/epoxy composite beam using different vibration modes of natural frequencies. The shifting of vibration is attributed to the implication of cracks. These vibration signatures are fuzzified through hybrid fuzzy sets (triangular, trapezoidal, Gaussian) and scaled to crack location and depth using the fuzzy rules and defuzzification process. The vibration signatures are recorded using ABAQUS finite element (FE) simulation software for a fixed beam and are fed as input parameters to the developed FIS for computing the desired outputs. The realization for crack depth and position is experimentally verified through a Fast Fourier Transform (FFT) analyzer. The experimental results with simulated data show that fuzzy logic application detects crack positions and depth accurately at different levels. It is concluded that the hybrid FIS bears a close resemblance to the experimental analysis and also stands out as an effective method for crack detection in LCB over other standalone methods. The current method can be used as a cost-effective non-destructive technique for health monitoring and fault diagnosis of composite beam structures in any practical field.

Floating wind turbines (FWTs) operate in offshore environments under harsh and varying operating conditions, making frequent in situ monitoring dangerous for maintenance teams and costly for operators. Remote and automated diagnosis, including the stages of detection, identification, and severity characterization of early stage damages in FWTs through advanced vibration-based structural health monitoring (SHM) methods of the machine learning (ML) type, is evidently critical for timely repairs, extending their operational lifecycle, reducing maintenance costs, and enhancing safety. This study investigates, for the first time, the complete (all stages) damage diagnosis problem by employing well-established ML SHM methods and conducting hundreds of experiments on a lab-scale FWT model operating under different wind speeds and directions, both in healthy and damaged states. The latter include two distinct blade cracks of limited length, two added masses attached to the blade edge simulating possible accumulation of ice, and connection degradation at the mounting of the main tower with the floater. The results indicate that the proper training of advanced ML methods using damage-sensitive feature vectors that represent the structural dynamics within the entire frequency bandwidth of measurements may achieve flawless damage diagnosis, reaching 100% success at all diagnosis stages, even when only a minimal number of vibration signals from a limited number of sensors (a single sensor in this study) are used.

This study investigates the effectiveness of geodetic methods in Structural Health Monitoring (SHM), focusing on the utilization of the High-Rate Global Navigation Satellite System (HR-GNSS) and Robotic Total Station (RTS) for monitoring structural movements. Experiments were conducted on a horizontal single-axis shake table to simulate various frequencies and amplitudes. Data were analyzed using time series and Fast Fourier Transform (FFT) techniques to evaluate the performance of geodetic measurement methods in SHM studies Two applications were conducted using a single-axis shake table. In the first, the table oscillated at 0.25 Hz frequency and 20 mm amplitude, while data from a GNSS receiver on the upper table underwent processing with the TRACK module of GAMIT/GLOBK software using the kinematic post-process (KPP) GNSS technique. In the second, the reflector on the shake table moved through eight oscillations at various amplitudes and frequencies, monitored automatically with a LEICA TPS1200 RTS. Time series and FFT analyses were performed on all application data to determine oscillation frequencies and amplitudes. Method accuracy was assessed by comparing these values with data from the shake table’s high-precision position sensor (Linear Variable Differential Transformer-LVDT). Results showed good agreement between HR-GNSS measurements and LVDT data, with a -1.6mm amplitude difference for KPP GNSS. Additionally, RTS measurements accurately determined frequency values, with amplitude differences ranging from 0.2 mm to 6.5 mm. Root Mean Square Error (RMSE) values for eight RTS tests, covering frequencies between 0.25-0.50 Hz and amplitudes between 4.5-73.4 mm, varied from 2.1mm to 6.3mm, reflecting performance variability across different conditions.

This study presents a novel fast Fourier transform (FFT)‐based non‐contact vibrational harmonics measurement system using a position sensitive detector (PSD) along with calibration using a piezoelectric accelerometer. Frequency‐domain vibrational analysis is required as the changes in machine dynamics are directly related to its failures and could provide more insight into the vibration signal. In this regard, FFT is used for spectral analysis to detect the harmonics in the vibration signal. The novelty of the applied technique for detecting vibrational harmonics lies in its innate contactless nature where the vibration detection sensor i.e. PSD is placed at a particular distance from the vibrating target. Additionally, the parasitic and external vibrations, which might pose unforeseen errors in the detected vibration data, have been nullified by employing a self‐vibration technique using an ADXL‐345 three‐axis accelerometer. The results obtained through PSD have been calibrated via a standard Brüel & Kjaer (B & K) vibration measurement system which uses a piezoelectric accelerometer (B & K accelerometer). The proposed measurement technique is equipped with NI compact RIO‐9074 that features a real‐time processor and an FPGA. The system was observed to effectively measure the frequency range 5–600 Hz with a maximum relative error of 2% in FFT amplitudes.

A non-destructive evaluation technique using piezoceramic (PZT) as an actuator-sensor has an ability to efficiently detect structural damage. In this technique, a PZT actuator-sensor patch is bonded on a structure. Through the measurement of its electrical impedance, which is related to mechanical impedance of the structure being bonded, the change in structure properties due to damage can be detected. This paper presents the use of PZT in structural health monitoring to quantitatively detect damage of bolted joints. The structure used in this study consists of two aluminum beams connected by a bolted joint. The damage is simulated by loosening of the bolts. To quantitatively monitor the damage, a numerical model of the structure is formulated. Spectral element method (SEM) based on wave propagation approach is used to model the structure. A bonded-PZT beam and a bolted joint element are developed by using SEM. The equations of motion are derived by using Hamilton's principle subsequently, the spectral element matrices are formulated. Experimental results show the ability of this method to detect the damage. By using the proposed model, the loosening of bolts can be quantitatively identified as the change in stiffness and damping at the bolted joint. Therefore, this method has high potential to quantitatively monitor damage of bolted joints.

This paper introduces an efficient and reliable unsupervised method for detecting faults in a brushless direct current (BLDC) motor based on abnormality identification in sensor-acquired vibration and sound signals through multi resolution decompostion and analysis. The research utilizes the double-density dual-tree complex wavelet transform (DD-DT-CWT) to extract important features from vibration signals, and incorporates audio feature extraction for the sound signals. The captured signals are divided into overlapping segments to improve fault localization, and the features of each segment are organized in a coefficient matrix. Subsequently, singular value decomposition (SVD) is applied to the resulting coefficient matrix from the vibration and audio signals. To effectively monitor the motor’s condition, the singular values from both sets of sensor data are combined. Analysing the decay patterns of the singular values enables the identification of faults in the BLDC motor under test. By establishing a suitable threshold for the decay slope of the singular values, the proposed method can accurately and precisely identify and categorize various faults in BLDC motors. This early fault detection can prompt predictive maintenance to ensure the optimal performance, reduced downtime and longevity of BLDC motors.