Abstract: Addressing the limitations of online equipment fault diagnosis systems, which are unsuitable for fault detection and diagnosis in small and medium-sized equipment due to their high cost, inconvenient installation, and maintenance, this paper designs a real-time intelligent fault detection system for rolling bearings based on the TMS320C6713, meeting the requirements of real-time diagnosis, intelligent diagnosis, and equipment portability. The principles and methods of the system's hardware and software design are detailed, and the application of hardware resonance demodulation technology is demonstrated. Research results show that this system can be easily applied in engineering projects. Keywords: rolling bearing; fault diagnosis; DSP; resonance demodulation technology[b][align=center]Design of the Rolling Bearing Real-time Fault Diagnosis System Based on DSP YIN Jian-jun[/align][/b] Abstract: The on-line equipment fault diagnosis system is not suitable for fault detection and diagnosis of small and medium-sized equipment, because of its high cost and its inconvenience of installation and maintenance. This article designs a kind of intelligent rolling bearing real-time fault diagnosis diagnosis system, which is based on TMS320C6713 to meet real-time diagnosis, intelligent diagnosis, portable equipment and other aspects of requirements. This article describes the design principles and methods of the system hardware and software in detail, realizes the application of the demodulated resonance technique based on hardware. The results show that the system can be conveniently applied to the project. Key words: Rolling Bearing;Fault Diagnosis;DSP;Demodulated Resonance Technique 1. Introduction Rolling bearings are the most widely used general mechanical components in various rotating machinery. Whether their operating status is normal often directly affects the performance of the whole machine. According to statistics, in rotating machinery using rolling bearings, failures caused by rolling bearing damage account for about 30%[1]. It can be seen that the fault diagnosis of rolling bearings is still of great significance in engineering. On the one hand, important large-scale equipment of domestic enterprises often use online systems to monitor the working conditions of the equipment. However, online systems are expensive, have poor versatility, and are inconvenient to install and maintain, and are not suitable for a large number of small and medium-sized equipment. On the other hand, small vibration detectors only play the role of data acquisition and storage, and data analysis requires experienced people to do it, making it difficult to achieve the requirements of real-time diagnosis. The rolling bearing real-time fault diagnosis system designed in this paper, which uses the TMS320C6713 produced by TI as a digital signal processor, can meet the needs of real-time intelligent fault diagnosis of small and medium-sized equipment[2]. 2. Principle of Real-time Fault Diagnosis System for Rolling Bearings The real-time fault diagnosis system for rolling bearings mainly includes two aspects: hardware and software. (1) Hardware: mainly uses analog circuits to implement resonance demodulation technology [3-4] and digital signal processing circuit design. The impact signal generated by the fault causes the hardware resonator to resonate. The weak impact signal is modulated into the high-frequency resonance signal through the resonance response of the resonator. Then, the high-frequency signal is processed by demodulation to obtain a resonance demodulated wave that eliminates low-frequency vibration interference, thereby achieving the purpose of accurate diagnosis. Finally, the signal is converted by A/D and then sent to the DSP for FFT transformation. (2) Software: mainly based on the DSP to identify and diagnose the fault characteristics of the vibration signal after resonance demodulation. The DSP performs time-frequency conversion on the digital signal transmitted by the AD to obtain the spectrum of the vibration signal. The system can automatically analyze the spectrum and draw conclusions such as fault location, fault type, and severity, and display them on the LCD. [align=center] Figure 1. Block diagram of real-time fault diagnosis system for rolling bearing[/align] 3. Hardware structure of the system This system uses hardware resonance demodulation technology to realize the fault diagnosis of bearing. Its advantage over software resonance demodulation technology is that it can effectively extract the micro-impact signal of early mechanical faults and realize predictive maintenance in the early stage of faults[5]. In addition, the hardware implementation is faster than the software implementation, which better reflects the requirements of real-time diagnosis. The main principles of the system hardware design are: (1) Vibration signal processing circuit, which mainly includes vibration signal preprocessing circuit and resonance demodulation processing circuit, as shown in Figure 2: [align=center] Figure 2. Block diagram of vibration signal processing circuit[/align] ① The front-end vibration sensor detects the vibration signal of the bearing (which includes the resonance signal of the resonator). ② The vibration signal is conditioned and amplified to obtain a low-frequency signal that is easy to process. The charge amplifier and the programmable amplifier play the role of impedance matching and signal amplification of the sensor output and subsequent processing circuit. Since the vibration signal of bearing fault is very weak and the sensor output impedance is very large, a dedicated preamplifier integrated circuit with high signal-to-noise ratio and high gain is required. ③ The bandpass filter preprocesses the signal. After processing by the bandpass filter, low-frequency vibration and high-frequency interference can be effectively filtered out, and only the frequency range near the resonance frequency is retained. ④ The envelope detector detects the outer envelope of the resonance wave and converts the high-frequency resonance signal into a low-frequency envelope signal, thus realizing the resonance demodulation function. After processing by the envelope analyzer and low-pass filter, the interference of time-domain spike signals can be further eliminated, providing a more stable identification signal for the subsequent fault identification system. ⑤ Since the fault frequency of rolling bearings varies from tens to hundreds of hertz, it is necessary to filter out the high-frequency components before further processing. At the same time, selecting a suitable filter before A/D conversion can also play a role in anti-frequency aliasing. (2) Digital signal processing circuit part, as shown in Figure 3: [align=center] Figure 3. Block diagram of digital signal processing circuit[/align] ① The low-frequency demodulated signal is converted into a digital signal suitable for DSP processing by the A/D converter. This design selects the AD9221 A/D converter chip from TI, which features high sampling frequency, low power consumption, and single power supply. Its maximum sampling frequency reaches 1.5MHz, signal-to-noise ratio is 70dB, and sampling accuracy is high, sufficient to meet the system design requirements. ② The digital signal processor (DSP) is crucial to this design, as its processing speed directly affects the system's real-time performance. The DSP chip used in this system is the TI TMS320C6713. This is a floating-point DSP designed for high-precision, high-performance applications. Based on the C67x, it adds many peripheral devices and interfaces. The chip's core clock frequency can reach up to 300MHz, with a processing power of up to 2400MPIS. It has a configurable L2 cache, abundant peripheral resources, and a 32-bit External Memory Interface (EMIF) that allows seamless connection to SDRAM, FLASH, and other memory devices. It supports HPI, PCI, and I2C buses. ③ The TMS320C6713 has a maximum of 256KB of internal L2 SRAM, which is insufficient for the system's large-capacity data storage needs. External storage needs to be expanded. The system uses Micron's 32-bit SDRAM chip MT48LC4M32B2, expanding the dynamic storage space by 128MB. [align=center]Figure 4. Wiring diagram of DSP and SDRAM[/align] ④ The Flash memory of this system uses SST's SST39VF160. The SST39VF160 uses a single 2.7V power supply, with an access time of only 90ns, fast erasure (full chip erase in 15ms), and fast programming (full chip programming in 7s). ⑤ The TMS320C6713 requires a high-precision, stable dual power supply to ensure normal system operation. This system uses TI's high-precision power supply chip TPS54310 (not shown in the figure), which features external compensation circuitry and overcurrent protection circuitry. ⑥ The button control circuit and LCD display circuit are relatively simple and will not be described in detail here. 4. System Software Design The system software is powerful, and its software system covers the implementation of various algorithms; display of characters, Chinese characters and graphics; response of timers, serial ports, USB and external interrupts; setting and control of programmable amplifiers, LCD screens and system time; implementation of communication protocols; storage of file systems; memory management, etc. The main functions implemented by the system software are as follows: (1) The digital signal transmitted from AD is transformed into a frequency domain signal through FFT transformation, and then the spectrum of the vibration signal is obtained. (2) The system software uses the diagnostic method of BP neural network to perform intelligent diagnosis of the signal [6]. The information that reflects the characteristics of the vibration signal is used as the input of the neural network, and the diagnostic conclusions such as fault location, fault type and severity are used as the output of the network. The network is trained by BP algorithm [7], and then this neural network is used to automatically diagnose the actual bearing demodulated signal and report the fault. (3) The components of the demodulated wave and the data such as fault location, fault type, severity and bearing number are displayed on the large screen LCD, which is convenient for manual auxiliary judgment. (4) The keyboard has function shortcut keys, English and number keys to support manual data entry and interaction. During operation, system parameters and functions can be set through the keys, and the sampling length can be controlled, etc. (5) All measurement data are effectively saved, which is convenient for data management. The system can upload the measurement data to the host computer for storage and further fault analysis, and can also download the specific bearing model parameters and characteristic frequencies and other related information from the bearing library in the host computer. [align=center] Figure 5. System software function diagram[/align] 5. Conclusion Traditional vibration diagnostic instruments generally diagnose faults by judging the effective value, maximum amplitude, kurtosis and other time-domain characteristic information of vibration. The diagnostic method is simple, the signal processing is very rough and the reliability is low. This system uses BP neural network to diagnose faults, realizes intelligent diagnosis, and improves the diagnostic speed and diagnostic accuracy. Furthermore, this system employs hardware resonance demodulation technology for vibration signal analysis and fault diagnosis. Compared to software-based resonance demodulation, this approach allows for successful fault diagnosis in the early stages of fault formation, when the impact fault signal is weak. This enables focused monitoring of equipment with early-stage faults and provides ample time to procure replacement parts. Therefore, this system has broad application potential in engineering. The author's innovation lies in utilizing the high-speed signal processing capabilities of a DSP system to achieve real-time intelligent diagnosis of rolling bearings. The use of hardware resonance demodulation technology avoids the difficulty in detecting early-stage faults inherent in software-based resonance demodulation, making it widely applicable to rolling bearing fault diagnosis in small and medium-sized equipment. References [1] Chen Jin. Vibration monitoring and fault diagnosis of mechanical equipment [M]. Shanghai: Shanghai Jiaotong University Press, 1999.1-2 [2] Zhang Xiaoguang, Zhou Ning, Ding Yuquan. Design of CMOS image acquisition system based on DSP [J]. Microcomputer Information, 2007, 9-2: 193-196 [3] Mei Hongbin. Vibration monitoring and diagnosis of rolling bearings - theory, method and system [M]. Beijing: Machinery Industry Press, 1996.29-31 [4] Mei Hongbin, Yan Mingyin, Yang Shuzi. High frequency resonance method for fault diagnosis of rolling bearings [J]. Mechanical Design and Manufacturing, 1992, (2), 12-16. [5] Gao Lixin, Wang Dapeng, Liu Baohua et al. Application research of resonance demodulation technology in bearing fault diagnosis [J]. Journal of Beijing University of Technology, 2007, (1).1-5 [6] Wang Ping. Online intelligent diagnostic instrument for rolling bearing faults [J]. Bearing, 2003, (9). 35-38 [4] [7] HY Zhang, CW Chan. Fuzzy art map neural network and its application to fault diagnosis of navigation systems. Automatica, 2001, 37:1065-1070