Design of vibration monitoring system for track beam of suspended permanent magnetic levitation train

The suspended magnetic levitation train, as an innovative mode of transportation, has made significant progress in China in recent years. This train achieves levitation through the repulsive force between permanent magnets and permanent magnetic tracks, allowing it to travel stably in the air without the need for electrical power. It offers advantages such as high speed, low energy consumption and environmental friendliness (Zigang et al., 2022; Yingzhuo and Chunzhi, 2017). The domestically developed “Xingguo” permanent magnetic levitation aerial train stands out as a notable representative in this field, with a designed speed of up to 120 kilometers per hour and advanced features such as autonomous operation and intelligent control (Yang, 2022; Jie et al., 2020). Despite the numerous advantages of suspended magnetic levitation trains, they still encounter several challenges in practical applications, including the design, manufacturing and maintenance of track beams.

As the infrastructure for the operation of suspended magnetic levitation trains, track beams support the weight of the train and guide it along the designated path. Any deformation or damage to the track beams can directly impact the stability of the train’s levitation and its operational safety (Chen et al., 2021). Therefore, designing a health monitoring system for track beams is of significant importance in preventing train operational failures. This paper presents an integrated compact device for collecting vibration acceleration signals and wireless transmission, establishing a long-term vibration monitoring system for track beams. The device is compact and can be flexibly installed on the track beams.

The vibration monitoring system comprises a three-axis accelerometer module, a wireless Microcontroller Unit (MCU) main control module, a Security Digital (SD) storage module, a wireless access point and an upper computer. The system structure is illustrated in Figure 1. The accelerometer is compact, highly accurate, and adaptable to various complex environments. It primarily handles the data acquisition of vibration information and communicates with the wireless main control MCU via Serial Peripheral Interface (SPI). The SD storage module is primarily used to store vibration information during server disconnections caused by network instability. The wireless MCU module primarily implements functions such as sensor data packet processing, connection to the wireless AP, TCP client creation and communication with the upper computer server. The wireless AP serves as a bridge for data transmission between the lower-level wireless MCU and the upper-level host computer, providing a stable data link for efficient transmission. The upper computer primarily receives data packets transmitted from lower-level devices, parses and extracts vibration data, and visualizes this information. Additionally, the upper computer is capable of issuing commands to control the sensors and retrieve power status. The entire system is powered by a battery pack and employs a sleep-wake mode, enhancing energy efficiency for long-term vibration monitoring.

The sensing and acquisition module employs the digital three-axis accelerometer ADXL357. The ADXL357 is a compact, ultra-low-power three-axis accelerometer that features a built-in A/D converter, eliminating the need for complex signal conditioning circuits and allowing for direct output of digital signals (Yuliang et al., 2018). This sensor offers a resolution of 14 bits and a measurement range of ±40 g, enabling precise detection of minor vibration signals. The internal First In First Out (FIFO) memory allows for Direct Memory Access (DMA) operations, which reduces CPU load and enhances data transfer speed (Da et al., 2019). The sensor is controlled via the SPI interface of the MCU main control unit, enabling rapid data reading and writing. The sensor circuit is illustrated in Figure 2.

The wireless MCU module employs the low-cost, low-power and high-performance CC3200 chip (Jian and Lihua, 2020) produced by Texas Instruments (TI). It integrates an application MCU subsystem, a wireless Wi-Fi network processor subsystem and a power management subsystem. The application MCU subsystem features an industry-standard ARM Cortex-M4 core along with various peripheral interfaces. The Wi-Fi network processor subsystem supports 802.11 b/g/n RF, baseband processing, and a robust encryption engine in its MAC, enabling secure and fast connections to both local area networks and the Internet. Additionally, it includes a dedicated ARM MCU that alleviates the processing burden on the application MCU, thereby enhancing the overall performance and response speed of the system (Shihao et al., 2018). The power management subsystem supports low-power modes, including Sleep, Deep Sleep, Low Power Deep Sleep (LPDS) and Hibernate (HIB). This paper’s system software design adopts the LPDS low-power mode to facilitate long-term vibration monitoring.

This module is responsible for collecting sensor data and power voltage information. It can directly control the on-chip Wi-Fi to connect to a wireless network using the Cortex-M4 core, allowing data to be transmitted to the upper computer. This design eliminates the need for additional hardware support, thereby simplifying the requirements for peripheral circuit design. The MCU circuit is illustrated in Figure 3.

The data storage module utilizes SD NAND for storing sensor data. Compared to traditional SD cards, it offers higher storage density and lower costs (He et al., 2023). Additionally, it exhibits excellent performance in writing speed and durability, with a write cycle endurance of 50,000–100,000 times. Furthermore, its low-power design makes it suitable for the long-term vibration monitoring system discussed in this paper. The circuit diagram of the storage module is shown in Figure 4.

To facilitate the storage and management of collected sensor data, this system implements the FatFs file system (Shiqi et al., 2011) on the SD NAND based on the SD 2.0 communication protocol, allowing for fast and accurate retrieval of sensor information stored during network disconnections.

The wireless AP primarily functions as a bridge between wired and wireless networks, enabling wireless devices to connect to local area networks (LAN) or the Internet via Wi-Fi. In this paper, a router is selected as the wireless AP device. The wireless MCU connects to the router via Wi-Fi, placing it within the same LAN to ensure reliable data transmission. A computer is chosen as the upper computer, responsible for data collection, monitoring, processing, analysis and displaying information to users, as well as controlling remote devices.

This paper employs the FreeRTOS framework (Jialiang et al., 2019) to develop a real-time operating system, facilitating the scheduling of multiple tasks, including device initialization, serial command reception and parsing, sensor data acquisition and data packet uploading. The execution flow of the real-time operating system software is illustrated in Figure 5. The multitask scheduling is based on task priority and status, implemented using a time-slicing approach. By invoking delay functions to modify task states, the scheduler’s decision-making is influenced, allowing for task switching.

The network connection management task primarily involves establishing Wi-Fi connections, switching between low-power modes and storing data on the SD card. The serial reception task is responsible for parsing incoming AT commands to configure sensor-related parameters and updating the firmware through In Application Programming (IAP). The ADXL357 data acquisition task receives data from the sensor’s internal FIFO memory by setting up a circular queue buffer, adjusting corresponding flags based on the network connection status for appropriate data storage, as illustrated in Figure 6. The TCP client creation task communicates with the server using the TCP/IP protocol, transmitting data via sockets while also parsing and processing commands issued by the server.

The upper computer primarily implements the reception, processing, and display of wireless sensor data. Designed in the Qt development environment on a Windows system using C++, it creates a TCP server to listen for incoming data from wireless sensors. The received data packets are parsed to extract the x, y and z axis data from the sensors, which are then displayed in real-time as waveforms. Configuration of the sensors is achieved by sending AT commands via the serial interface. Additionally, specified data packets can be sent over TCP to set parameters, perform device self-checks, configure acquisition modes and facilitate data return. The upper computer also stores sensor data in local files for subsequent algorithmic classification and processing. Furthermore, the sensors periodically send heartbeat packets to the upper computer, allowing for timely detection and handling of anomalies, thereby enhancing system reliability. The functional structure of the upper computer software is illustrated in Figure 7.

In the upper computer software, the data processing primarily involves parsing the data packets transmitted by the sensors and extracting the corresponding three-axis data. Each data packet sent via TCP contains 16 sets of x, y and z vibration data along with a timestamp for the current moment. To accurately capture the vibration state at the current time, the average value of each dimension (x, y z) across these 16 sets of data is calculated. The calculation method is shown in the formula:

x_i, y_i and z_i represent the vibration values in the x, y and z directions, respectively, for the i-th data set. Avg_x、Avg_y、Avg_z denote the average values of all data in these three directions, collectively forming the vibration information at the current moment. This approach effectively reduces random fluctuations in the data, providing a more stable and reliable description of the vibration state.

Open the upper computer software and use the serial connection to send AT commands to configure the Wi-Fi and TCP server IP addresses and port numbers that the sensor needs to connect to. Once the configuration is complete, the sensor will return status information to the serial reception box. Click on the TCP settings in the connection method section and set it to the corresponding IP and port configured previously for the sensor. Then, click “Start Listening.” If the listening is successful, a dialog box will appear indicating that the client is connected. At this point, click the “Start Acquisition” button in the data processing section, and the sensor data will begin to be received by the upper computer, displaying in real-time on the software window. The monitoring interface is shown in Figure 8, where it can be observed that the sensor is in normal acquisition mode. Once the keep-alive time ends, it immediately enters Low Power Deep Sleep (LPDS) mode. After the sleep duration ends, the sensor is awakened and returns to normal acquisition mode, continuing to collect data. This method enables long-term vibration monitoring of the train track beams.

When you want to remotely query the sensor’s status, you can click the “Instruction delivery” button to open the command transmission and reception window. A specified format data packet is sent to the sensor via TCP, where the sensor receives and parses the command data packet, then returns the sensor status data in a specific format to the upper computer software. The command transmission interface is shown in Figure 9, where the upper computer receives feedback from the sensor after sending a query command. The command issuance data packet consists of several fields. The header of the packet, “AA 55 BB,” functions as a static prefix for identifying and differentiating various commands. The command field, denoted by “23,” specifies the particular operation or function to be performed. Subsequent to this, the two bytes “00 00” serve as a serial number, uniquely identifying the data packet. Finally, the final two bytes “00 00” are utilized to denote the length of any subsequent data, such as configuration information, that may follow.

To test the accuracy of the sensors in the system, a vibration table is used to provide known standard vibration signals. The vibration table can generate vibrations of various frequencies and amplitudes, simulating conditions in actual working environments. By mounting the vibration sensor on the table and recording its output signals, the sensor’s output can be compared with the actual vibration parameters of the table, allowing for an evaluation of the sensor’s performance and accuracy. The physical sensor is shown in Figure 10, and the laboratory test vibration table is depicted in Figure 11.

By setting the vibration table to an acceleration of 1 g and a vibration frequency ranging from 50 Hz to 500 Hz, with increments of 50 Hz, the table vibrates according to a sine wave pattern. The acceleration signals collected by the sensor are recorded, as shown in Table 1.

By conducting multiple measurements at an acceleration of 1 g across the frequency range of 50–500 Hz, the maximum acceleration error does not exceed 5%.The test results demonstrate that the vibration sensor possesses high measurement accuracy and stability, making it suitable for track beam vibration monitoring applications. These findings provide critical data support for the subsequent design and optimization of the monitoring system.

To address the issue of information collection for suspended magnetic levitation train track beams, this paper designs a wireless device for acquiring vibration signals and constructs a remote monitoring system to provide data support for train operational safety. Test results indicate that the device can accurately measure acceleration signal values, and the system can acquire and display vibration information in real-time on the upper computer, as well as send commands to obtain the operational status of the wireless acquisition device. In the future, the graphical display interface and command transmission interface will be optimized, and data processing algorithms will be introduced to assess the health status of the track beams, ensuring safer and more stable train operations.