Array eddy current sensing device for monorail inspection robot and signal processing method thereof

Abstract

This invention discloses an array eddy current sensing device and its signal processing method for a monorail inspection robot. The array eddy current sensing device includes: a sensor fixture with a contoured structure at its bottom adapted to the profile of the rail top and gauge angle, and multiple mounting holes on its surface; an eddy current sensor unit installed in each mounting hole, including a sensor housing containing a sensor main board, an excitation coil, and multiple receiving coils; when the eddy current sensor unit is installed in the mounting hole, the excitation coil faces the bottom surface of the mounting hole; the multiple receiving coils are adjacent and positioned below the excitation coil; the sensor main board is communicatively connected to the excitation coil and the receiving coils respectively; a host module is detachably connected to the side of the sensor fixture and communicates with the sensor main board of each eddy current sensor unit via a serial bus. This invention effectively suppresses dynamic lift-off interference and improves detection accuracy and stability through a non-contact sensor design and a corresponding signal processing method.

Description

Technical Field

[0001] This invention relates to the field of nondestructive testing technology, specifically to an array eddy current sensing device for a monorail inspection robot and its signal processing method. Background Technology

[0002] As a core load-bearing component of the railway system, the surface damage of rails, such as cracks, peeling, and wear, is a key factor affecting train operation safety. To ensure railway operational safety, it is necessary to periodically inspect rails using testing equipment. For example, an automated walking inspection robot equipped with an array of eddy current sensors can be used to detect surface damage. As the robot moves along the top of the rail, bumps, swaying, or changes in the curvature of the rail surface cause real-time variations in the lift-off distance between the coil in the array of eddy current sensors and the tested surface. These variations significantly alter the electromagnetic coupling strength between the coil and the tested rail, causing drastic fluctuations in the detection signal that are unrelated to actual surface damage, thus interfering with the eddy current detection signal.

[0003] In existing technologies, a mechanical constraint method involving an elastic clamping structure and auxiliary wheels is used to control the lift-off distance between the sensor and the rail surface to remain relatively constant, thereby reducing the interference of lift-off variations on the eddy current detection signal. However, with the increasing complexity of inspection scenarios and higher requirements for robot maneuverability, especially in some special areas (such as turnout areas, tunnels, and bridges), it is necessary to use a rail-top walking monorail inspection robot as an inspection device to inspect the rails.

[0004] Because monorail robots that only move along the top of the track have complex locomotion and support mechanisms, they require high levels of weight distribution and balance. When such monorail inspection robots are equipped with traditional array eddy current sensors, the added elastic clamping structures and auxiliary wheels significantly increase the robot's structural complexity and introduce additional, uncontrollable mechanical contact forces. This, in turn, interferes with the robot's stable movement on the smooth track, exacerbating its dynamic imbalance and instability, leading to detection deviations or operational malfunctions, and reducing the stability of the detection requirements. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides an array eddy current sensing device for a monorail detection robot and its signal processing method. The aim is to effectively suppress dynamic lift-off interference and improve detection accuracy and stability through non-contact sensor design and matching signal processing methods.

[0006] This invention discloses an array eddy current sensing device for a single-track detection robot, comprising: The sensor fixture has a contoured structure at its bottom that matches the profile of the rail top and gauge angle. The surface of the sensor fixture has multiple mounting holes. The sensor fixture can be detachably connected to the monorail inspection robot. An eddy current sensor unit, installed in each mounting hole, includes a sensor housing, a sensor main board, an excitation coil, and multiple receiving coils inside the sensor housing. The excitation coil is located at the bottom of the sensor housing, facing the bottom surface of the mounting hole when the eddy current sensor unit is installed in the mounting hole. The multiple receiving coils are adjacent to each other along their long or wide sides and are located below the excitation coil. The sensor main board provides communication connections between the excitation coil and the receiving coils. The host module is detachably connected to the side of the sensor fixture and communicates with the sensor motherboard of each eddy current sensor unit via a serial bus.

[0007] Preferably, there are four receiving coils, each with an area one-quarter the size of the excitation coil. The receiving coils are of the same size and are distributed sequentially in the first, second, third, and fourth quadrants of the excitation coil from the bottom view direction.

[0008] Preferably, the mounting holes are staggered and equidistantly distributed on the surface of the sensor fixture, with multiple mounting holes covering the top of the rail and the gauge angle area; The bottom surface of the mounting hole is tangent to the corresponding arc surface of the rail, so that each eddy current sensor unit is perpendicular to the corresponding rail surface after installation.

[0009] Preferably, the multiple mounting holes are arranged in an array of 4 rows and 10 columns, wherein the rows are along the length of the rail and the columns are along the tangent of the rail. The row-direction offset distance between two adjacent mounting holes in the same row is within the first interleaving distance range, and the column-direction offset distance is within the second interleaving distance range; The row-direction offset distance between two adjacent mounting holes in the same column is within the third interleaving distance range, and the column-direction offset distance is within the fourth interleaving distance range.

[0010] Preferably, the sensor housing is provided with a guide structure and a clamping structure on both sides, the guide structure is located below the clamping structure, the guide structure is a convex strip, and the clamping structure is an elastic buckle.

[0011] Preferably, the sensor motherboard integrates a microcontroller, a signal generation and acquisition circuit, a multi-channel analog front-end module, a phase-sensitive detection module, and a serial communication interface; The microcontroller is communicatively connected to the signal generation and acquisition circuit, the multi-channel analog front-end module, and the phase-sensitive detection module; the microcontroller is also communicatively connected to the host module via a serial communication interface. The signal generation and acquisition circuit is communicatively connected to the excitation coil; The multi-channel analog front-end module includes multiple analog front-end channels, and these channels are connected to multiple receiving coils in a corresponding communication manner. The phase-sensitive detection module is communicatively connected to the signal generation and acquisition circuit and the multi-channel analog front-end module.

[0012] Preferably, the analog front-end channel includes a preamplifier, a bandpass filter, and an ADC driver circuit; The input of the preamplifier is connected to its corresponding receiving coil, the input of the bandpass filter is connected to the output of the preamplifier, the input of the ADC driver circuit is connected to the output of the bandpass filter, and the output of the ADC driver circuit is connected to the microcontroller.

[0013] Preferably, the top of the sensor fixture is provided with a robot connection hole, the eddy current sensing unit is encapsulated in the mounting hole, and the serial communication interface is exposed.

[0014] This invention discloses a signal processing method for an array eddy current sensor device of a monorail inspection robot, applied to the aforementioned array eddy current sensor device of the monorail inspection robot. The signal processing method includes: During the process of the monorail inspection robot walking along the rail, the eddy current sensor unit continuously collects the sensing signals to form a raw signal matrix that is associated with the time series and the sensing signals. The sensing signals include the self-inductance signal generated by the excitation coil and the induced voltage signal generated by multiple receiving coils. The original signal matrix is ​​preprocessed using an eddy current sensor unit to obtain a digital signal matrix. The preprocessing includes signal amplification, signal bandpass filtering, and signal analog-to-digital conversion. The digital signal matrix is ​​demodulated using an eddy current sensor unit to obtain the demodulated signal of the excitation coil and the demodulated signal of the receiving coil. The inductive reactance of the excitation coil is calculated based on the demodulated signal of the excitation coil, and the absolute lift-off value is calculated based on the inductive reactance value of the excitation coil by calling a preset calibration function. The demodulated signal of the excitation coil includes the excitation amplitude and the excitation phase, and the demodulated signal of the receiving coil includes the receiving amplitude and the receiving phase. The host module polls each eddy current sensor unit and performs differential combination calculations on the received amplitude based on the spatial arrangement of multiple receiving coils in the eddy current sensor unit to obtain preliminary damage characteristic signals. Obtain the geometric model of the rail cross section and, in conjunction with the absolute lift-off value, calculate the tilt angle of the eddy current sensor unit. The tilt angle is caused by the swaying of the monorail detection robot as it walks along the rail. The influence coefficient of the differential signal is calculated based on the geometric model of the rail section and the tilt angle, and the ratio of the preliminary damage characteristic signal to the influence coefficient of the differential signal is calculated to obtain the corrected damage characteristic signal. Based on the preset damage characteristic signals and the magnitude of the corrected damage signals, the robot determines whether the rail surface is damaged and uploads the rail damage results to the robot's main controller.

[0015] Preferably, there are four receiving coils, each with an area one-quarter the size of the excitation coil. The receiving coils are of the same size and are distributed sequentially in the first, second, third, and fourth quadrants of the excitation coil from the bottom view direction. Based on the spatial arrangement of multiple receiving coils in the eddy current sensor unit, differential combination calculations are performed on the received amplitudes to obtain preliminary damage characteristic signals, including: The difference between the first amplitude sum and the second amplitude sum is calculated to obtain the preliminary damage characteristic signal. The first amplitude sum is the sum of the received amplitude of the receiving coil corresponding to the first quadrant and the received amplitude of the receiving coil corresponding to the third quadrant. The second amplitude sum is the sum of the received amplitude of the receiving coil corresponding to the second quadrant and the received amplitude of the receiving coil corresponding to the fourth quadrant.

[0016] Compared with the prior art, the beneficial effects of the present invention are as follows: This invention abandons the existing method of relying on auxiliary wheels or elastic clamping structures for physical constraint lifting, and proposes a collaborative anti-interference architecture of hardware spatial filtering and software dynamic compensation. This architecture utilizes an excitation coil and multiple quadrant-arranged receiving coils within a single sensor unit, combined with real-time absolute lift-off values ​​calculated based on the excitation coil's inductive reactance, to construct a dynamic reference benchmark. This fundamentally eliminates the interference of dynamic lift-off changes on the detection signal without any mechanical contact. Simultaneously, the bottom of the fixture features a contoured structure that conforms to the rail top and gauge angle, ensuring that each sensor unit automatically aligns perpendicular to the rail surface after installation, reducing system errors introduced by installation angle deviations. Attached Figure Description

[0017] Figure 1 This is a schematic diagram of the array eddy current sensing device for the monorail detection robot provided by the present invention. Figure 2 This is a schematic diagram of the structure of the eddy current sensor unit provided by the present invention; Figure 3 This is a schematic diagram of the insertion structure of the eddy current sensor unit provided by the present invention; Figure 4 This is a schematic diagram of signal acquisition for the array eddy current sensor provided by the present invention; Figure 5 The flowchart of the signal processing method provided by the present invention.

[0018] Illustration: 1. Sensor clamp; 11. Contouring structure; 12. Mounting hole; 13. Robot connection hole; 14. Threaded connection hole; 2. Eddy current sensor unit; 21. Sensor housing; 211. Clamping structure; 212. Guide structure; 22. Sensor main board; 23. Excitation coil; 24. Receiver coil A; 25. Receiver coil B; 26. Receiver coil C; 27. Receiver coil D; 3. Main module; 4. Rail. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0020] The present invention will now be described in further detail with reference to the accompanying drawings.

[0021] Combination Figure 1 , Figure 2 and Figure 3 As shown, this embodiment of the invention provides an array eddy current sensing device for a monorail inspection robot, including a sensor fixture 1, an eddy current sensor unit 2, and a main unit module 3. The sensor fixture 1 has a contoured structure 11 at its bottom that adapts to the profile of the rail top and gauge angle. The contoured structure 11 tightly fits the rail top and gauge angles on both sides of the rail 4. Multiple mounting holes 12 are provided on the surface of the sensor fixture 1, allowing for detachable connection between the sensor fixture 1 and the monorail inspection robot.

[0022] Eddy current sensor unit 2 is installed in each mounting hole 12. It includes a sensor housing 21, within which a sensor main board 22, an excitation coil 23, and multiple receiving coils are housed. The excitation coil 23 is located at the bottom of the sensor housing 21. When the eddy current sensor unit 2 is installed in the mounting hole 12, the excitation coil 23 faces the bottom surface of the mounting hole 12 to emit an excitation magnetic field onto the surface of the rail 4. Multiple receiving coils are adjacent along their long or wide sides and positioned below the excitation coil 23 to receive the induced magnetic field signal generated on the surface of the rail 4 due to the eddy current effect. The sensor main board 22 is communicatively connected to both the excitation coil 23 and the receiving coils, and is responsible for generating the excitation signal and processing and converting the received signal.

[0023] The host module 3 is detachably connected to the side of the sensor fixture 1, for example, by means of a threaded connection hole 14 on the side of the sensor fixture 1. It is also connected to the sensor motherboard 22 of each eddy current sensor unit 2 via a serial bus to control each eddy current sensor unit 2 and to summarize and analyze the detection data.

[0024] During use, the sensor fixture 1 is connected to the bottom of the monorail inspection robot via the robot connection hole 13 provided with bolts or other fasteners, enabling quick assembly and disassembly of the device and facilitating subsequent maintenance and replacement. During the inspection of the rail, the monorail inspection robot moves at a constant speed along the track. The arrayed eddy current sensor unit 2 excites an alternating magnetic field in real time and simultaneously collects electromagnetic response signals from the surface and near-surface of the rail. The host module 3 polls each individual eddy current sensor unit 2 via an RS485 serial bus, receiving the reported detection signals and lift-off values. It then processes and packages the summarized data, and interacts with the robot or other communication devices via a four-pin main interface connector.

[0025] In this embodiment of the invention, there are four receiving coils, each with an area one-quarter the area of ​​the excitation coil. The receiving coils are identical in size and are distributed sequentially in the first, second, third, and fourth quadrants of the excitation coil from the bottom view direction. Figure 1 As shown, the four receiving coils are receiving coil A (24), receiving coil B (25), receiving coil C (26), and receiving coil D (27), where receiving coil A (24) is located in the first quadrant, receiving coil B (25) is located in the second quadrant, receiving coil C (26) is located in the third quadrant, and receiving coil D (27) is located in the fourth quadrant. This symmetrical four-quadrant distribution design allows a single eddy current sensor unit to collect induction signals from different spatial orientations, providing a hardware foundation for subsequent differential combination operations. It also helps to suppress the effects of environmental interference and sensor tilt, thereby extracting rail damage characteristics more accurately.

[0026] In this embodiment of the invention, the mounting holes 12 are staggered and equidistantly distributed on the surface of the sensor fixture 1, with multiple mounting holes 12 covering the rail top and gauge angle areas. This distribution ensures a comprehensive and thorough scan of the critical detection areas of the rail, avoiding blind spots in detection.

[0027] The bottom surface of mounting hole 12 is tangent to the corresponding arc surface of the rail, ensuring that each eddy current sensor unit 2 is perpendicular to its corresponding rail surface after installation. The bottom plane of each mounting hole 12 is tangent to the arc surface of the clamp at that location, guaranteeing that the installed sensor unit is perpendicular to the rail surface. Since the rail top and gauge angle are arc surfaces with different curvatures, the normal direction of the bottom plane of each mounting hole points towards the center of the rail section, ensuring that all sensor units are perpendicular to their respective local tangent planes on the rail surface after installation. This provides a unified geometric reference for subsequent lift-off value calculation and differential processing. Through this design, when the eddy current sensor unit 2 is installed in mounting hole 12, its excitation coil 23 and receiving coil can face the rail surface at the optimal perpendicular angle, thus ensuring the efficiency of electromagnetic coupling and the accuracy of the detection signal, and reducing detection errors caused by sensor tilt.

[0028] Furthermore, the mounting holes are arranged in a 4x10 array, with the rows along the length of the rail and the columns along the tangent to the rail. This 4x10 array layout forms multiple rows of detection points along the length of the rail, enabling simultaneous detection of different locations on the same cross-section of the rail, thus improving detection resolution and reliability. Specifically, the row-direction offset distance between two adjacent mounting holes in the same row is within a first staggered distance range, and the column-direction offset distance is within a second staggered distance range; the row-direction offset distance between two adjacent mounting holes in the same column is within a third staggered distance range, and the column-direction offset distance is within a fourth staggered distance range. This staggered arrangement allows the detection areas of the eddy current sensor units within adjacent mounting holes to complement and cover each other, further optimizing the spatial sampling density of the entire detection device and ensuring accurate detection even when the rail surface has complex contours or minor damage.

[0029] For example, there are 40 mounting holes (12 in total), arranged in a staggered, equidistant pattern of 4 rows and 10 columns. The center-to-center distance between adjacent mounting holes in each row is 8 mm. The first row, first column mounting hole serves as the starting point, located at the gauge angle on one side of the rail, while the fourth row, tenth column mounting hole serves as the ending point, located at the gauge angle on the other side of the rail. The second row, first column mounting hole is offset by 18 mm in the row direction and 2 mm in the column direction relative to the starting point; the third row, first column mounting hole is offset by 36 mm in the row direction and 4 mm in the column direction relative to the starting point; the fourth row, first column mounting hole is offset by 54 mm in the row direction and 6 mm in the column direction relative to the starting point. The remaining columns of mounting holes maintain an equidistant distribution in the row direction and an equal column spacing in the column direction, forming a staggered array structure.

[0030] Through this staggered distribution, 40 mounting holes form a continuous detection array on the rail surface, covering the rail top and the gauge angle areas on both sides. The mounting holes in adjacent rows are staggered in the row direction, which effectively avoids longitudinal detection blind spots caused by alignment within the same column, ensuring the continuity and coverage of detection during robot movement.

[0031] In this embodiment of the invention, a guide structure 212 and a clamping structure 211 are provided on both sides of the sensor housing 21. The guide structure 212 is located below the clamping structure 211. The guide structure 212 is a protruding strip, and the clamping structure 211 is an elastic buckle.

[0032] The inner wall of the mounting hole 12 is provided with a guide groove that matches the guide structure 212, and a locking groove that cooperates with the locking structure 211. When the eddy current sensor unit 2 is installed, precise positioning is first achieved through the cooperation of the guide structure 212 and the guide groove, ensuring that the sensor unit is smoothly inserted into the mounting hole 12 in a preset direction. As the insertion depth increases, the elastic locking structure 211 will undergo elastic deformation and lock into the locking groove when passing through the locking groove position, thereby firmly fixing the eddy current sensor unit 2 in the mounting hole 12 and preventing it from shifting or falling off due to vibration or shaking during robot operation. This design makes the installation and removal of the eddy current sensor unit 2 convenient and efficient, without the need for complex tools, facilitating on-site maintenance and quick replacement of the sensor unit.

[0033] In this embodiment of the invention, the sensor motherboard 22 integrates a microcontroller, a signal generation and acquisition circuit, a multi-channel analog front-end module, a phase-sensitive detection module, and a serial communication interface. The microcontroller is communicatively connected to the signal generation and acquisition circuit, the multi-channel analog front-end module, and the phase-sensitive detection module. The microcontroller is also communicatively connected to the host module via the serial communication interface. The signal generation and acquisition circuit is communicatively connected to the excitation coil. The multi-channel analog front-end module includes multiple analog front-end channels, each of which is communicatively connected to a corresponding receiving coil. The phase-sensitive detection module is communicatively connected to both the signal generation and acquisition circuit and the multi-channel analog front-end module.

[0034] like Figure 4 As shown in this embodiment of the invention, the sensor motherboard 22 integrates a microcontroller (MCU), a signal generation and acquisition circuit, a multi-channel analog front-end module, a phase-sensitive detection module, and a serial communication interface (RS485 communication interface). The microcontroller (MCU) is an STM32G4 series microcontroller with embedded ADC and DAC, serving as the core control and data processing unit. The MCU controls the DAC to generate the excitation waveform, simultaneously acquires and receives coil signals through four ADCs, executes a digital demodulation algorithm, and finally communicates with the outside via the RS485 bus.

[0035] In the signal generation and acquisition circuit, the MCU controls the DAC (digital-to-analog converter) to generate a sinusoidal digital sequence, which is then low-pass filtered and sent to the power amplifier circuit. The amplified excitation current drives the excitation coil 23, inducing an eddy current field on the rail surface. The power amplifier circuit uses an operational amplifier (OPA547) to form a voltage-to-current conversion circuit to ensure stable excitation current.

[0036] In the multi-channel analog front-end module, the weak voltage signals induced by four receiving coils are input to independent analog front-end channels. Each analog front-end channel includes a preamplifier, a bandpass filter, and an ADC driver circuit. The input of the preamplifier is communicatively connected to its corresponding receiving coil; the input of the bandpass filter is communicatively connected to the output of the preamplifier; the input of the ADC driver circuit is communicatively connected to the output of the bandpass filter; and the output of the ADC driver circuit is communicatively connected to the microcontroller.

[0037] Specifically, the preamplifier uses a low-noise instrumentation amplifier (such as INA128) to convert the differential signal into a single-ended signal and amplify it. The bandpass filter is a second-order bandpass filter composed of an operational amplifier (OPA227) and an RC network, with its center frequency matching the excitation frequency, filtering out out-of-band noise. The ADC driver circuit uses an operational amplifier to form a voltage follower, increasing the input impedance and ensuring that the signal enters the ADC without distortion.

[0038] The conditioned four analog signals are sent to the MCU's built-in multi-channel ADC for synchronous sampling. The ADC adopts a successive approximation (SAR) structure, with a sampling rate of no less than 10 times the excitation frequency, and is configured in continuous conversion mode, triggered synchronously by a timer.

[0039] In the phase-sensitive detection module, after the MCU reads the ADC conversion result, it implements the phase-sensitive detection (PSD) algorithm in software to calculate the amplitude and phase of the signal in each channel. The demodulated data is packaged in a fixed format, converted into differential signals via an RS485 transceiver (MAX3485), and transmitted to the array eddy current sensor host via the bus. The RS485 bus uses a master-slave communication protocol, with each sensor unit acting as a slave, using address identification to achieve data polling or active uploading.

[0040] In this way, the entire array eddy current sensor device is a standalone electronic package, with the RS485 interfaces of all individual eddy current sensor units 2 connected in parallel to the host module 3 via a bus. The host module 3 consists of an independent host MCU (STM32H7 series microcontroller) and its peripheral circuitry, housed in a resin-encapsulated shielded box on the side of the fixture. After completing excitation control, signal acquisition, and demodulation, the MCU within each eddy current sensor unit 2 uploads its lift-off value h and the amplitude data of the four receiving coils to the host module 3 via the RS485 bus. The array eddy current sensor host module 3 is responsible for polling all 40 sensor units, summarizing the data, and performing spatial differential calculations, dynamic compensation based on the rail model, and damage identification, ultimately generating the detection results.

[0041] The array eddy current sensor main module 3 exchanges data with the robot's main controller via a second RS485 connection. All power and communication lines converge to a four-pin main interface connector (located on the side of the fixture), which includes a power line (DC 5V), a ground line, and a pair of RS485 differential signal lines (for communication with the robot). The monorail inspection robot powers the entire array eddy current sensor through this connector and communicates with the array eddy current sensor main module 3 via its own main controller to receive the final inspection results.

[0042] This architecture encapsulates the entire sensor array into a fully functional electronic package, with independent data acquisition, processing, and decision-making capabilities internally, while only requiring a power supply and a simple communication interface externally, greatly facilitating the rapid mounting, replacement, and maintenance of robots.

[0043] In this embodiment of the invention, the top of the sensor fixture 1 is provided with a robot connection hole 13, the eddy current sensing unit 2 is encapsulated in the mounting hole 12, and the serial communication interface is exposed.

[0044] For example, the entire sensor unit is encapsulated in the array eddy current sensor fixture 1 using epoxy resin, with only the RS485 serial port interface exposed, to achieve electrical connection and physical protection, thereby improving waterproof and dustproof performance.

[0045] The following specific embodiment describes the complete usage process of the array eddy current sensing device for the monorail detection robot of the present invention: When the monorail inspection robot carries this device along the rail, the sensor fixture 1 fits tightly against the top and gauge angle of the rail 4 via its contoured structure 11 at the bottom, ensuring the stability of the entire sensing device during the inspection process. The eddy current sensor unit 2 within each mounting hole 12 generates an alternating electromagnetic field under the control of the sensor mainboard 22 via its excitation coil 23. This electromagnetic field acts on the surface of the rail 4, inducing eddy currents within the rail. Defects on or near the surface of the rail (such as cracks or wear) interfere with the distribution of eddy currents, causing changes in the induced magnetic field received by the receiving coil. The receiving coil converts these changing magnetic field signals into electrical signals and transmits them to the sensor mainboard 22. The signal generation and acquisition circuit and multi-channel analog front-end module on the sensor mainboard 22 perform preprocessing on the raw signal, including amplification, filtering, and analog-to-digital conversion. The signal is then demodulated by the phase-sensitive detection module to obtain a demodulated signal containing amplitude and phase information. The microcontroller calculates the inductive reactance of the excitation coil based on these demodulated signals, then obtains the absolute lift-off value through a preset calibration function, and performs differential combination calculations on the amplitude of the receiving coil to obtain preliminary damage characteristic signals. The host module 3 polls each eddy current sensor unit 2 via a serial bus to collect these preliminary damage characteristic signals, absolute lift-off values, and other data. Next, the host module 3 calculates the tilt angle of the eddy current sensor unit 2 based on the rail cross-section geometric model and the absolute lift-off value, and then calculates the differential signal influence coefficient based on this to correct the preliminary damage characteristic signals, obtaining corrected damage characteristic signals. Finally, by comparing with the preset damage characteristic signals, it determines whether the rail is damaged and marks the damage status. This non-contact design avoids the mechanical interference caused by traditional elastic clamping structures and auxiliary wheels, significantly improving the walking stability and detection accuracy of the monorail inspection robot under complex track conditions, and meeting the stringent requirements for rail inspection in special areas such as turnout areas, tunnels, and bridges.

[0046] like Figure 5 As shown, the present invention also provides a signal processing method for an array eddy current sensor device of a monorail inspection robot, applied to the array eddy current sensor device of a monorail inspection robot, the signal processing method comprising: S1. During the process of the monorail inspection robot walking along the rail, the eddy current sensor unit continuously collects the sensing signals to form the original signal matrix that is associated with the time series and the sensing signals.

[0047] In this embodiment of the invention, each eddy current sensor unit continuously acquires sensing signals at a preset sampling frequency (e.g., not less than 1 kHz) during the robot's movement. The sensing signals include the self-inductance signal generated by the excitation coil and the induced voltage signals generated by multiple receiving coils. Since the robot moves along the length of the rail, each sensor unit corresponds to a different position on the rail surface at different times over time. Therefore, the acquired raw signals not only contain the amplitude and phase information of each receiving coil but also implicitly include the time dimension, i.e., the sampling time. Arranging these time-varying sensing signals in chronological order and associating them with the corresponding sampling times forms a raw signal matrix. The rows of this matrix can represent different sampling times (corresponding to different positions of the robot's movement), the columns can represent different receiving coil channels, and the elements in the matrix are the signal amplitude or phase data acquired by each channel at the corresponding time. This raw signal matrix provides a complete spatiotemporal data foundation for subsequent signal processing and damage identification.

[0048] For example, as the robot moves along the rail, the host module sends acquisition commands to each eddy current sensor unit via an RS485 bus at a fixed frequency (1kHz). The microcontroller (MCU) inside each eddy current sensor unit controls its four ADCs to sample synchronously, simultaneously acquiring the self-induced voltage signal V from the excitation coil. exc (t) and the induced voltage signals VA(t), VB(t), VC(t), and VD(t) of the four receiving coils. In this embodiment, the excitation frequency is set to 100kHz, the ADC sampling rate is set to 1MHz (10 times the excitation frequency), and each sampling point is quantized into a 16-bit digital quantity. The acquired data is temporarily stored in the internal buffer of each unit and aggregated to the host module via the RS485 bus to form the original signal matrix V corresponding to the time series. raw (t).

[0049] S2. The original signal matrix is ​​preprocessed using the eddy current sensor unit to obtain the digital signal matrix.

[0050] In this embodiment of the invention, preprocessing includes signal amplification, bandpass filtering, and analog-to-digital conversion. Signal amplification involves differentially amplifying the weak voltage signal induced by the receiving coil using a preamplifier, converting it into a single-ended signal, and boosting it to a voltage range suitable for subsequent processing, thereby improving the signal-to-noise ratio and the accuracy of subsequent processing. Bandpass filtering uses a second-order bandpass filter to remove low-frequency noise below the excitation frequency (such as environmental electromagnetic interference and noise introduced by mechanical vibration) and high-frequency noise above the excitation frequency (such as thermal noise from electronic components), retaining the effective signal frequency band related to changes in the rail eddy current field. Analog-to-digital conversion uses a multi-channel ADC built into the microcontroller to convert the amplified and filtered analog signal into a 16-bit digital quantity, thus converting the continuous analog signal into a discrete digital signal for subsequent digital signal processing. After the above preprocessing steps, the original analog signal is converted into a digital signal matrix containing amplitude and phase information for each channel, laying the foundation for subsequent phase-sensitive detection and damage feature extraction.

[0051] For example, the original signal matrix V is processed using an eddy current sensor unit. raw (t) Digital signal processing is performed. First, a bandpass filter is used to filter out high-frequency noise and power frequency interference. The filter parameters are: center frequency 100kHz, bandwidth ±10kHz, passband gain 0dB, stopband attenuation ≥40dB, and a 4th-order Butterworth filter is used. Then, the signal is amplified to make the signal amplitude suitable for the subsequent processing range. Finally, the preprocessed digital signal matrix V is obtained. pre (t), which contains the time-domain data of each channel.

[0052] S3. Demodulate the digital signal matrix using the eddy current sensor unit to obtain the demodulated signal of the excitation coil and the demodulated signal of the receiving coil. Calculate the inductive reactance of the excitation coil based on the demodulated signal of the excitation coil, and call the preset calibration function to calculate the absolute lift-off value based on the inductive reactance of the excitation coil.

[0053] In this embodiment of the invention, the preprocessed digital signal V pre (t) The input demodulation module performs lift-off calculation using phase-sensitive detection. The demodulation module employs phase-sensitive detection (PSD) technology to extract the lift-off value from the pre-processed digital signal V. pre The amplitude and phase information are extracted from (t), where the demodulated signal of the excitation coil includes the excitation amplitude and excitation phase, and the demodulated signal of the receiving coil includes the receiving amplitude and receiving phase.

[0054] Specifically, the signals of each channel are multiplied with the reference sine and cosine signals of the same frequency and then low-pass filtered to obtain the real part I and the imaginary part Q, and then the amplitude and phase are calculated.

[0055] During the lift-off value calculation, the inductive reactance of the excitation coil is calculated based on the demodulated signal of the excitation coil using the following formula: ; Among them, I exc The known excitation current amplitude is set to 50mA in this embodiment. Since the excitation coil covers a large area, local defects have little impact on its overall impedance. However, lift-off changes significantly alter the electromagnetic coupling between the coil and the rail, causing a regular change in the inductive reactance. Therefore, the inductive reactance value can accurately reflect the lift-off distance. A preset calibration function is called to calculate the absolute lift-off value based on the excitation coil's inductive reactance. This preset calibration function is obtained in advance through experiments: In a laboratory environment, the eddy current sensor unit is placed above a standard rail sample, and the lift-off distance is changed using a high-precision displacement platform (range 0.5mm-5mm, step 0.1mm), recording the corresponding inductive reactance values, and fitting them into a fourth-order polynomial curve.

[0056] S4. The host module polls each eddy current sensor unit and performs differential combination calculations on the received amplitude based on the spatial arrangement of multiple receiving coils in the eddy current sensor unit to obtain the preliminary damage characteristic signal.

[0057] In this embodiment of the invention, the host module establishes communication with each eddy current sensor unit sequentially via an RS485 bus at a preset polling period (e.g., 10ms) to read the received amplitude data from the demodulated signal of the receiving coil processed by its internal MCU. Considering that each eddy current sensor unit typically contains multiple receiving coils arranged according to a specific spatial pattern, the host module first determines the geometric spacing and orientation between these receiving coils based on their known spatial coordinates or relative positional relationships. Then, based on this spatial arrangement, differential combination operations are performed on the received amplitudes acquired by different receiving coils within the same eddy current sensor unit to highlight local electromagnetic field changes caused by surface or near-surface defects in the rail. This differential combination operation can effectively suppress common-mode interference caused by lift-off fluctuations, differences in rail material uniformity, etc., thereby extracting preliminary damage characteristic signals related to defects. These preliminary damage characteristic signals typically manifest as one or more voltage differences or combinations thereof related to the location and size of the defect.

[0058] Specifically, in this embodiment of the invention, there are four receiving coils, each with an area one-quarter the area of ​​the excitation coil. The receiving coils are identical in size and are distributed sequentially in the first, second, third, and fourth quadrants of the excitation coil from the bottom view direction. When calculating the preliminary damage characteristic signal, the difference between the first amplitude sum and the second amplitude sum is calculated to obtain the preliminary damage characteristic signal. The first amplitude sum is the sum of the received amplitudes of the receiving coils corresponding to the first quadrant and the third quadrant, and the second amplitude sum is the sum of the received amplitudes of the receiving coils corresponding to the second quadrant and the fourth quadrant.

[0059] For example, the received amplitude of receiving coil A is |VA|, the received amplitude of receiving coil B is |VB|, the received amplitude of receiving coil C is |VC|, and the received amplitude of receiving coil D is |VD|. The preliminary damage characteristic signal is then calculated using the following formula. 。

[0060] Among them, S raw This represents the initial damage characteristic signal. Thus, the above calculations can eliminate the common-mode signal caused by overall lift-up changes or the curved surface of the rail (the four coils change in the same direction), while amplifying the differential-mode signal caused by local damage (differences appear in diagonal or adjacent coils). When the eddy current sensor unit undergoes overall lift-up changes or the curved surface of the rail causes a uniform change in the magnetic field, the signals of the four coils increase or decrease synchronously, and the result of the above formula approaches zero, thereby eliminating common-mode interference; however, when there is local damage on the rail surface (such as cracks or spalling), the signal of the receiving coil directly below the damage will change abruptly, causing a difference between |VA|+|VC| and |VB|+|VD|, S raw Non-zero values ​​highlight the damage characteristics.

[0061] S5. Obtain the geometric model of the rail section and, in conjunction with the absolute lift-off value, calculate the tilt angle of the eddy current sensor unit.

[0062] In this embodiment of the invention, slight swaying may occur during robot movement, causing the sensor unit to tilt relative to the rail surface. This results in inconsistent lift-off values ​​for the four coils, leading to spurious changes in the differential signal. To compensate for this effect, this step introduces a rail cross-section geometric model. The rail cross-section profile can be obtained based on the standard geometric parameters of the rail cross-section geometric model.

[0063] The tilt angle of the eddy current sensor unit is calculated using the following formula based on the absolute lift-off value and the rail cross-sectional profile: in, This represents the difference in absolute lift-off values ​​between two adjacent eddy current sensor units. This indicates the spacing between adjacent units (e.g., 5mm).

[0064] S6. Calculate the influence coefficient of the differential signal based on the geometric model of the rail section and the tilt angle, and calculate the ratio of the influence coefficient of the preliminary damage characteristic signal to the influence coefficient of the differential signal to obtain the corrected damage characteristic signal.

[0065] In this embodiment of the invention, the influence coefficient of tilt on the differential signal (i.e., the differential signal influence coefficient) is calculated based on the geometric model of the rail cross-section and the tilt angle. Specifically, the differential signal influence coefficient is obtained in advance through finite element simulation: a sensor-rail electromagnetic field simulation model is established under different tilt angles, the amplitude ratio of the differential signal generated by the same damage is calculated, and the differential signal influence coefficient is calculated using the following formula: in, This represents the influence coefficient of the differential signal. Indicates the tilt angle, measured in degrees or radians. and The fitting coefficients are determined using the least squares method. It is 0.12. It is -0.005.

[0066] Furthermore, the damage characteristic signal is corrected using the following formula: Among them, S comp To correct the damage characteristic signal, S raw The initial damage characteristic signal is 1 when the tilt angle is 0. The corrected damage characteristic signal is the same as the initial damage characteristic signal.

[0067] S7. Based on the preset damage characteristic signal and the magnitude of the corrected damage signal, determine whether the rail surface is damaged, and upload the rail damage result to the robot main controller.

[0068] In this embodiment of the invention, the compensated corrected damage signal is compared with a preset damage characteristic signal. If the amplitude of the corrected damage characteristic signal is greater than or equal to the amplitude of the preset damage characteristic signal, it is determined that there is a relative amplitude and damage type of rail surface damage at that location, and the next detection cycle continues. If the amplitude of the corrected damage characteristic signal is less than the amplitude of the preset damage characteristic signal, it is determined that there is no damage, and the next detection cycle continues.

[0069] The preset damage feature signal is obtained by detecting a large number of known damage samples (such as cracks, spalling, etc. of different depths, widths, and lengths), extracting the mean and standard deviation of the corresponding modified damage feature signal, and calculating it using the following formula: Where Th is the preset damage characteristic signal, µ is the mean of the corrected damage characteristic signal, and σ is the standard deviation of the corrected damage characteristic signal. For example, if μ = 0.05V and σ = 0.02V, then Th = 0.11V.

[0070] Furthermore, the host module (3) records the time point of the damage occurrence and, combined with the robot's walking speed and mileage information, calculates the specific location coordinates of the damage on the rail. The host module (3) packages the damage location, damage characteristic signal amplitude, and other information and uploads them to the main controller of the monorail detection robot via CAN bus or Ethernet. The main controller then performs subsequent damage level assessment, data storage, and reporting operations.

[0071] As can be seen from the above technical solution, this application provides an array eddy current sensing device and its signal processing method for a monorail detection robot. It abandons the existing method of relying on auxiliary wheels or elastic clamping structures for physical constraint lifting, and proposes a collaborative anti-interference architecture of hardware spatial filtering + software dynamic compensation. This architecture utilizes an excitation coil and multiple receiving coils arranged in quadrants within a single sensor unit, combined with the real-time absolute lift-off value calculated based on the inductive reactance of the excitation coil, to construct a dynamic reference benchmark. This fundamentally eliminates the interference of dynamic lift-off changes on the detection signal without any mechanical contact. Simultaneously, the bottom of the fixture is equipped with a contoured structure that conforms to the top of the rail and the rail gauge angle, ensuring that each sensor unit automatically becomes perpendicular to the rail surface after installation, reducing system errors introduced by installation angle deviations.

[0072] Furthermore, within a single eddy current sensor unit, the four receiving coils are one-quarter the size of the excitation coil and are arranged strictly in quadrants directly below the excitation coil. This arrangement ensures that lift-off changes are represented as common-mode signals from the four coils, while local damage is represented as differential-mode signals, providing the hardware foundation for subsequent differential calculations. Unlike traditional methods that merely eliminate lift-off as interference, this invention actively utilizes the signal from the excitation coil, demodulating its inductive reactance and combining it with a pre-stored calibration function to calculate the absolute lift-off value of the sensor unit relative to the rail surface in real time and accurately. This lift-off value is not only used for dynamic compensation but also provides a crucial input for attitude calculation of the rail cross-section model.

[0073] To address the sensor tilting issue caused by robot movement and swaying, this invention introduces a geometric model of the rail cross-section. By utilizing the distribution characteristics of lift-off values ​​calculated from multiple adjacent elements, the current tilt angle of the sensor is deduced, and the differential signal influence coefficient is calculated to correct the differential signal. This algorithm enables the sensor to maintain detection accuracy even under minute swaying, significantly improving the robot's maneuverability.

[0074] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. An array eddy current sensing device for a monorail inspection robot, characterized in that, include: The sensor fixture has a contoured structure at its bottom that adapts to the profile of the rail top and gauge angle. The surface of the sensor fixture has multiple mounting holes. The sensor fixture is detachably connected to the monorail inspection robot. An eddy current sensor unit, installed in each of the mounting holes, includes a sensor housing. Inside the sensor housing are a sensor main board, an excitation coil, and multiple receiving coils. The excitation coil is located at the bottom of the sensor housing, facing the bottom surface of the mounting hole when the eddy current sensor unit is installed in the mounting hole. The multiple receiving coils are adjacent to each other along their long or wide sides and are located below the excitation coil. The sensor main board provides communicative connections between the excitation coil and the receiving coils. The host module is detachably connected to the side of the sensor fixture and communicates with the sensor motherboard of each eddy current sensor unit via a serial bus.

2. The array eddy current sensing device according to claim 1, characterized in that, The number of receiving coils is four, and the area of ​​each receiving coil is one-quarter of the area of ​​the excitation coil. The receiving coils are the same size and are distributed sequentially in the first quadrant, second quadrant, third quadrant, and fourth quadrant of the excitation coil from the bottom view direction.

3. The array eddy current sensing device according to claim 1, characterized in that, The mounting holes are staggered and equidistantly distributed on the surface of the sensor fixture, and multiple mounting holes cover the top of the rail and the gauge angle area. The bottom surface of the mounting hole is tangent to the arc surface of the rail corresponding to the mounting hole, so that each eddy current sensor unit is perpendicular to the corresponding rail surface after installation.

4. The array eddy current sensing device according to claim 3, characterized in that, The mounting holes are arranged in an array of 4 rows and 10 columns, wherein the rows are along the length of the rail and the columns are along the tangent of the rail. The row-direction offset distance between two adjacent mounting holes in the same row is within a first interleaving distance range, and the column-direction offset distance is within a second interleaving distance range; The row-direction offset distance between two adjacent mounting holes in the same column is within the third interleaving distance range, and the column-direction offset distance is within the fourth interleaving distance range.

5. The array eddy current sensing device according to claim 1, characterized in that, The sensor housing has a guide structure and a clamping structure on both sides. The guide structure is located below the clamping structure. The guide structure is a convex strip, and the clamping structure is an elastic buckle.

6. The array eddy current sensing device according to claim 1, characterized in that, The sensor motherboard integrates a microcontroller, a signal generation and acquisition circuit, a multi-channel analog front-end module, a phase-sensitive detection module, and a serial communication interface. The microcontroller is communicatively connected to the signal generation and acquisition circuit, the multi-channel analog front-end module, and the phase-sensitive detection module, respectively; the microcontroller is communicatively connected to the host module via the serial communication interface. The signal generation and acquisition circuit is communicatively connected to the excitation coil; The multi-channel analog front-end module includes multiple analog front-end channels, and the multiple analog front-end channels are communicatively connected to the multiple receiving coils. The phase-sensitive detection module is communicatively connected to the signal generation and acquisition circuit and the multi-channel analog front-end module, respectively.

7. The array eddy current sensing device according to claim 6, characterized in that, The analog front-end channel includes a preamplifier, a bandpass filter, and an ADC driver circuit. The input terminal of the preamplifier is communicatively connected to its corresponding receiving coil, the input terminal of the bandpass filter is communicatively connected to the output terminal of the preamplifier, the input terminal of the ADC driver circuit is communicatively connected to the output terminal of the bandpass filter, and the output terminal of the ADC driver circuit is communicatively connected to the microcontroller.

8. The array eddy current sensing device according to claim 6, characterized in that, The top of the sensor fixture is provided with a robot connection hole, the eddy current sensing unit is encapsulated in the mounting hole, and the serial communication interface is exposed.

9. A signal processing method for an array eddy current sensor device of a monorail inspection robot, applied to the array eddy current sensor device of the monorail inspection robot according to any one of claims 1-8, characterized in that, include: During the process of the monorail inspection robot walking along the rail, the eddy current sensor unit continuously collects the sensing signals to form a time series and an original signal matrix associated with the sensing signals. The sensing signals include the self-inductance signal generated by the excitation coil and the induced voltage signal generated by multiple receiving coils. The original signal matrix is ​​preprocessed using the eddy current sensor unit to obtain a digital signal matrix. The preprocessing includes signal amplification, signal bandpass filtering, and signal analog-to-digital conversion. The digital signal matrix is ​​demodulated using the eddy current sensor unit to obtain the demodulated signal of the excitation coil and the demodulated signal of the receiving coil. The inductive reactance of the excitation coil is calculated based on the demodulated signal of the excitation coil, and a preset calibration function is called to calculate the absolute lift-off value based on the inductive reactance of the excitation coil. The demodulated signal of the excitation coil includes the excitation amplitude and the excitation phase, and the demodulated signal of the receiving coil includes the receiving amplitude and the receiving phase. The host module polls each of the eddy current sensor units and performs differential combination calculations on the received amplitudes based on the spatial arrangement of the multiple receiving coils in the eddy current sensor units to obtain preliminary damage characteristic signals. Obtain the geometric model of the rail cross section, and calculate the tilt angle of the eddy current sensor unit by combining the absolute lift-off value. The tilt angle is caused by the swaying of the monorail detection robot as it walks along the rail. The differential signal influence coefficient is calculated based on the rail cross-section geometric model and the tilt angle, and the ratio of the preliminary damage characteristic signal to the differential signal influence coefficient is calculated to obtain the corrected damage characteristic signal. Based on the preset damage feature signal and the magnitude of the corrected damage signal, the robot determines whether the rail surface is damaged and uploads the rail damage result to the robot's main controller.

10. The signal processing method according to claim 9, characterized in that, The number of receiving coils is four, and the area of ​​each receiving coil is one-quarter of the area of ​​the excitation coil. The receiving coils are the same size and are distributed sequentially in the first quadrant, second quadrant, third quadrant and fourth quadrant of the excitation coil from the bottom view direction. The step of performing differential combination calculations on the received amplitude based on the spatial arrangement of multiple receiving coils in the eddy current sensor unit to obtain a preliminary damage characteristic signal includes: The difference between the first amplitude sum and the second amplitude sum is calculated to obtain the preliminary damage characteristic signal. The first amplitude sum is the sum of the received amplitude of the receiving coil corresponding to the first quadrant and the received amplitude of the receiving coil corresponding to the third quadrant. The second amplitude sum is the sum of the received amplitude of the receiving coil corresponding to the second quadrant and the received amplitude of the receiving coil corresponding to the fourth quadrant.

Images