www.as-se.org/ccse Communications in Control Science and Engineering (CCSE) Volume 4, 2016 Development of Bolt Crack Detection Device Based on Ultrasonic Wave Chuangang Wang 1, Fuqiang Li 1, Liang Lv 2, Xiao Lin 2 1 CRRC Qingdao Sifang CO., LTD., Qingdao, China 2 School of Mechanical, Electronic and Control Engineering, Beijing Jiaotong University, Beijing, China 1 wangchuangang@cqsf.com; 2 lifuqiang@cqsf.com; 3 15121249@bjtu.edu.cn; 4 12222088@bjtu.edu.cn Abstract A portable bolt crack detection device using ultrasonic was developed based on ARM and FPGA. This paper mainly introduced the operating principle of ultrasonic system, the hardware structure of the device and wave denoising technologies to realize this device. Finally, a lot of experiments on the device are carried out to detect the bolts, verifying the practicability and accuracy of the device. The results show that 1.5 mm-depth bolt crack can be recognized and alarmed. This device greatly reduced the work load and work requirements and can easily operate. Keywords Bolt Crack; Ultrasonic Detection; Wavelet Analysis Introduction With the improvement of train speed, the kinetic energy of train increases sharply, so the load of bolts has vastly increased. The crack of the brake disc bolt will lead to serious accidents. To avoid this situation, we shall start with the detection method and find the bolt with crack in advance. There are three kinds of traditional methods for the detection of bolt crack: magnetic particle testing, radiographic testing and ultrasonic testing. The ultra-sonic testing has many advantages. It is simple and fast, not required to be disassembled and suitable for automatic operation. Meanwhile, it can be used in detection on-line. The generic ultrasonic flaw detector is too complex for the operator lacking of experience. Currently, with the combination of torque testing, manual ultrasonic testing is adopted by vehicle factories; it consumes a lot of human and financial resources. In order to find the bolts with crack in advance to ensure the safety of the train in operation, this paper developed a system of detecting the crack of the brake disc bolt on-line. Operating Principle If there are holes, cracks and other flaws or inclusion in the metal, ultrasonic wave will travel to the flaw and be reflected from the surface of the flaw to the transducer. The reflected wave will be received by the transducer, processed by the internal circuit of instrument and displayed by the LCD screen. The depth and position of the flaw can be estimated according to the change of the waveform of the reflected wave. FIG. 1 THE PRINCIPLE OF PLUSE-ECHO METHOD 6
Communications in Control Science and Engineering (CCSE) Volume 4, 2016 www.as-se.org/ccse Pulse-echo method is commonly used in the bolt crack detection because of its high sensitivity and good positioning accuracy. Fig. 1 provides basic principle of the pulse-echo method: When the test specimen is flaw-free, the ultrasonic wave can travel to the bottom surface, there are only the transmitted wave (T) and the bottom echo (B) in the detection waveform, as shown in Fig.1(c). When the test specimen is with a flaw, the transmitted wave will be reflected by the surface of the flaw, there is another flaw echo (F) between the transmitted wave (T) and the bottom echo (B), as shown in Fig. 1(d). Designs of Hardware Structure Fig.2 shows the hardware structure of ultrasonic bolt crack detection device. The device mainly includes ultrasonic transducer, ultrasonic transmitting circuit, ultrasonic receiving circuit, FPGA, ARM, touch screen, etc. The working process is described below. First, a pulse signal generated with FPGA drives ultrasonic emission circuit transmits a pulse with a high amplitude and low pulse width to stimulate the trans-ducer. Next, the transducer produces ultrasonic wave travelling through the bolt under test and receives echo. The ultrasonic receiving circuit processes the echo by limiting, amplification, filtering, and 12 bit high-speed A/D conversion. Then, the processed signal is received by FPGA, in which FIFO buffer storage the signal and convert data to 16 bits. Finally, ARM microcontroller determinates the location and size bolt crack with the received data processed by wavelet denoising, and displays the useful information on the touch screen. FIG.2 THE HARDWARE STRUCTURE OF THE DEVICE The portable intelligent device, which can store the detection data and have buzzer alarming will satisfy the detecting requirement of improving the detection efficiency greatly. The device adopts 12V rechargeable lithium battery power supply, and the capacity can meet the requirements of long working-time sustained. The Circuit of Transducer The Excitation of Ultrasonic Waves Two kinds of frequently-used ultrasonic excitation signal are sharp pulse signal and square signal. The sharp pulse signal can encourage broadband narrow pulse, which is conducive to the production of a more narrow ultrasonic pulse signal, and the cost is low. So the sharp pulse is chosen as the excitation signal, and the RC discharge circuit is adopted as the sharp pulse signal excitation circuit. 7
www.as-se.org/ccse Communications in Control Science and Engineering (CCSE) Volume 4, 2016 We present some improvement measures to optimize the sharp pulse signal according to actual application. 1) Pulse Amplitude as Large as Possible Transmit power of ultrasonic emission signal can be expressed as: P = V2 C t where V is the amplitude of sharp pulse signal, C is the capacitance of the emission circuit, t is discharge time. It can be seen that transmit power of ultrasonic emission signal is in proportion to V squared. The greater the power of the ultrasonic signal is, the greater the amplitude of the crack echo which can be detected easier will be. The amplitude of sharp pulse is determined by the voltage of the high voltage DC power supply, resist-ance and capacitance. The resistance and capacitance also affect the width of the sharp pulse. Therefore, the best method to improve the amplitude of the sharp pulse is to improve the voltage of the high voltage DC power supply directly. A 0~1000V voltage adjustable high voltage power supply is used in the system, the peak value of the pulse can be up to 800V in its 1000V voltage output, and meets the design requirements. 2) Rise Time Is as Short as Possible If the emission signal may include the inherent frequency of the transducer which is certain, the vibration effect of the piezoelectricity chip will achieve the best. The time constant of discharging can be expressed as: τ = RC (2) it is knowable that the rise time of the pulse can adjust by changing the circuit resistance and capacitance. While this method reduces the pulse amplitude at the same time, it is not effective in practical application. We improve the structure of the transmission circuit, as shown in Fig.3. The Reception of Ultrasonic Waves 1) Signal Modulating Because the device sends and receives ultrasonic wave (1) HV R C A D1 +/-5v D2 CRY MOSFET R1 R0 FIG. 3 EMISSION CIRCUIT OF SHARP PULSE OPTIMIZED by the same transducer, and the amplitude of the ultrasonic excitation pulse is very high. In order to protect the reception circuit, a limiting circuit is needed before other circuit. After limiting, the peaks of signal are flattened, and because its amplitude is not large, on which limiting circuit has limited impact, noise needs to be filtered the band pass filter in order to improve SNR. The amplitude of the ultrasonic signal has a very large range; it is likely to exceed the range of A/D conversion, so it is necessary to adjust the gain of the amplifier according to the change of the signal. 2) Signal Sampling The signal can be sent to the A/D converting module after the signal modulating circuit. The natural frequency of transducer is 5MHz. The sampling frequency should be twice greater than the highest frequency of signal 8
Communications in Control Science and Engineering (CCSE) Volume 4, 2016 www.as-se.org/ccse according to sampling theorem, 5~10 times in general are selected in order to ensure the effect of sampling. AD9433 was chosen as A/D converting chip which supports sampling frequency of 50MHz. Wavelet Denoising In order to analyse whether there is a crack signal in the ultrasonic signal, it is necessary to reduce the noise of the signal to prevent output noise jamming system. Wavelet has good analysis ability in time and frequency domain, and can be used to analyse the localization characteristics of signal. Wavelet analysis technique applied to ultrasonic detection system can improve not only the accuracy of the analysis on defect information, also the reliability of the system; it has a great significance in improving the detection performance. The aim of noise reduction is to eliminate the noise in the signal and reconstruct the original signal. Wavelet threshold denoising method used in this paper includes three basic steps: Wavelet decomposition, Threshold determination and Signal reconstruction. Wavelet Decomposition The wavelet is decomposed by two wavelet decomposition filters, both of which are essentially two sets of coefficients. After filtering, the detail coefficients and scale coefficients of the layer can be obtained, and then the scale coefficients will be resolved after quadratic extraction sampling, until a predetermined number of decomposition levels are reached. The wavelet decomposition level of the system is deter-mined as the 5 levels. The first important thing in wavelet decomposition is the choice of wavelet function. The Daubechies wavelet function is generally used in the processing of ultrasonic signals. However, after many simulations and experiments have been done, the Symlet wavelet function is adopted in the paper, as seen in Fig.4. It is found that waveform denoised by the Sym8 wavelet basis is more close to original waveform, and SNR is higher. Threshold Determination FIG. 4 COMPARISON BETWEEN DB AND SYM WAVELET BASIS Due to the detail coefficients taking up less percentage in the original ultrasonic signal and as SNR is not very high, fixed threshold method widely used is adopted because of its convenient calculation and significant effect. The fixed threshold thr can be estimated by: tthrr = σσ 2llllll (nn) (3) where n is the length of signal, σ is the standard deviation of noise, it can be computed as: σ = median Det1 /0.6745 (4) where median is to the median function. Det1 is the detail coefficients of first level in wavelet decompo-sition. Namely, mmmmmmmmmmmm DDDDDD1 is intermediate value of the wavelet detail coefficients of the corresponding level obtained after decomposition. 0.6745 is the adjustment coefficient of white Gaussian noise. 9
www.as-se.org/ccse Communications in Control Science and Engineering (CCSE) Volume 4, 2016 Signal Reconstruction The last step of wavelet denoising is to reconstruct the signal, which is actually the inverse process of wavelet decomposition. Two wavelet reconstruction filters are used in the reconstruction process, which are also essentially two sets of coefficients and can be computed according to wavelet decomposition filters. The detail coefficients of the upper layer can be calculated by the detail coefficients and scale coefficients of each level convolution with wavelet reconstruction filter after quadratic interpolation. Finally, the reconstructed signal can be obtained. Identification of Crack Positions The identification of the crack positions by traditional gate judgment is dominated by a selected threshold, if there is a peak value in the detected area of ultrasonic wave exceeding the threshold value, we can identify the presence of a crack echo, as shown in Fig. 5. FIG.5 THE HARDWARE STRUCTURE OF THE DEVICE Because of the different testing conditions, the thres-hold can be adjusted after calibration. However, there are certain defects existing in this method: for cracks less than 2mm, the impact on the ultrasonic signal is small, and the change is relatively small, it is likely to filter the useful signal in denoising processing, while the useful signal will be drowned in noise signal without denoising. It is difficult to solve this problem. Since the pulse-echo method is adopted to crack detection, so the location of the crack can be obtained by the formula: s = v t/2 (5) where s is the distance between surface of crack and transducer, v is the ultrasonic velocity (5,900 m/sec in steel) and t is the time of wave travelling through the bolt and back to transducer. specimens Crack free TABLE 1 THE EXPERIMENT RESULTS 2mm crack at 7th thread 3mm crack at 7th thread 3mm crack at 6th thread 3mm crack at 5th thread No.1 100 0 0 0 0 100 No.2 0 98 2 0 0 100 No.3 0 0 96 0 4 100 No.4 0 0 0 99 1 100 No.5 0 0 0 2 98 100 total Experiment and Error Analysis In the experiment, 5 MHz round type transducer with the diameter of 14mm have been chosen. The bolt with no notch (No.1), the bolt with 2mm notch at the seventh thread root (No.2), the bolt with 3mm notch at the seventh thread root (No.3), the bolt with 3mm notch at the sixth thread root (No.4) and the bolt with 3mm notch at the fifth 10
Communications in Control Science and Engineering (CCSE) Volume 4, 2016 www.as-se.org/ccse thread root (No.5) were selected to be tested. The test at the same condition for each bolt has been repeated 100 times. Statistic results of the test can be seen in Table 1. It can be concluded from table I that the correct rate of measuring device to bolt detection is very high, only 1.8% of the rate of false positives. These false positives are caused by the error in manual operation and the noise jamming of the circuit. The false positive rate can be acceptable. And it has no effect on detecting the existence of bolt crack, so this system is reliable and meets the basic needs. Conclusions This paper has developed a portable ultrasonic bolt crack detection device based on ARM and FPGA. It is proved that the device can detect and size bolt crack accurately and effectively, and automatically alarm. It has a good human-computer interface and makes the bolt crack detection intelligent. REFERENCES [1] Fushao Tang. Study and Design of Ultrasonic Flaw Dectector Based on ARM11 Master diss., Nanchang University, 2014. [2] Bin Jin, Minzhu Lou, liping Zhang. Technique and System for Auto Eddy Current Detection of the Cracks in Bolts Nondestructive Testing, 2003, 25(11): 565-567, 588. 11