An edge-computing-based intelligent tire abnormal vibration signal feature extraction method

Abstract

This invention discloses a method for extracting features of abnormal vibration signals from intelligent tires based on edge computing, belonging to the field of edge computing technology. The extraction method acquires the transient angular displacement of a rotating object and the rate of change of potential along the energy transmission path. When the transient angular displacement and the rate of change of potential meet preset trigger conditions, the energy transmission path is controlled to switch to a physically disconnected state and a waiting sequence is executed. After the waiting sequence ends, the sensing channel is opened to acquire multi-axis transient vibration data. Based on the energy differences in different axes of the multi-axis transient vibration data, the orthotropic kinetic energy ratio coefficient is calculated. The orthotropic kinetic energy ratio coefficient is compared with a feature threshold to generate a judgment result and the energy transmission path is controlled to reset. This invention, through temporal disconnection and spatial decoupling, suppresses electromechanical crosstalk and centrifugal interference, reducing the probability of missed and false alarms.

Description

Technical Field

[0001] This invention relates to the field of edge computing technology, and in particular to a method for extracting features of abnormal vibration signals from intelligent tires based on edge computing. Background Technology

[0002] The intelligent tire edge sensing system relies on piezoelectric transducers to collect the mechanical energy of tire deformation to maintain the passive operation of the microcontroller, and monitors tire physical structural anomalies through multi-axis accelerometers deployed inside the tire. The standard engineering paradigm for extracting abnormal vibration characteristics in existing technologies is as follows: The microcontroller's built-in analog-to-digital converter performs continuous-time-domain digital sampling of the broadband vibration signal output from the multi-axis accelerometers; the microcontroller's built-in digital filtering algorithm performs noise reduction on the data sequence generated by the continuous-time-domain digital sampling; and then, the noise reduction output data sequence is subjected to absolute amplitude comparison or all-axial energy summation to output a judgment result.

[0003] Deploying an engineering paradigm relying on continuous-time domain digital sampling and purely digital domain filtering in a piezoelectric self-powered physical environment presents an engineering conflict due to the physical saturation of the underlying hardware acquisition circuitry. During the transient physical cycle of a tire rolling and contacting the ground, the piezoelectric transducer, affected by the compression deformation of the tire material matrix, instantaneously releases a transient electromagnetic inrush current of hundreds of volts. Under continuous-time domain digital sampling conditions, this hundreds-volt transient electromagnetic inrush current directly generates electromechanical coupling crosstalk to the microcontroller's reference voltage network through the common-ground physical network. This electromechanical coupling crosstalk causes broadband physical saturation distortion at the front-end sampling pins of the microcontroller's analog-to-digital converter (ADC). This broadband physical saturation distortion results in continuous full-scale digital gibberish output from the ADC, exceeding the effective recovery boundary of the digital filtering algorithm. Consequently, the microvolt-level stress wave signal from physical structural damage is masked by the underlying electromagnetic background noise.

[0004] In the dimension of physical signal spatial analysis, existing technologies relying on axial energy summation or absolute amplitude comparison face a three-dimensional spatial mismatch problem caused by centrifugal physical field interference. During tire rotation, the centrifugal physical field attached to the internal mass block of the multi-axis accelerometer generates parasitic mechanical resonance. The mechanical energy of this parasitic mechanical resonance is concentrated on the radial axis perpendicular to the tire's rotation plane. The high-frequency stress wave mechanical energy caused by tire structural tears or ply fractures is concentrated on the lateral and tangential axes parallel to the tire's rotation plane. Axial energy summation and absolute amplitude comparison confuse the parasitic mechanical resonance concentrated on the radial axis with the structural damage characteristics concentrated on the lateral and tangential axes. With increasing tire rotational speed, the parasitic mechanical resonance energy on the radial axis amplifies proportionally to the square of the tire's rotational speed, overwhelming the local vibration data on the lateral and tangential axes. This leads to missed events in the intelligent tire edge sensing system at low rotational speeds and widespread false alarms at high rotational speeds.

[0005] Based on the above physical failure analysis, existing abnormal vibration signal feature extraction techniques face engineering limitations due to wideband physical saturation distortion at the front end of the analog-to-digital converter and large-scale false alarm events caused by centrifugal physical field interference. Summary of the Invention

[0006] This invention provides a method for extracting features of abnormal tire vibration signals based on edge computing. This method addresses the problem that, under piezoelectric self-powered conditions, the electromechanical coupling crosstalk caused by a transient electromagnetic inrush current of hundreds of volts and the parasitic mechanical resonance driven by the centrifugal physical field superimpose to cause irreversible broadband saturation distortion at the front end of the analog-to-digital converter and disrupt the spatial distribution consistency of the multi-axis vibration signal. As a result, weak structural damage features are masked by background noise during the acquisition and analysis stages, leading to engineering defects such as missed low-speed and false high-speed detections.

[0007] In view of the above problems, the present invention provides a method for feature extraction of abnormal vibration signals of smart tires based on edge computing, comprising the following steps: Space electrical association anchoring steps: acquire motion vector data of the rotating object to calculate transient angular displacement, and acquire the potential change rate of the energy transmission path associated with the rotating object; When the transient angular displacement is within a preset range and the rate of change of potential is greater than a preset trigger threshold, a first trigger signal is issued. In response to the first trigger signal, the energy transmission path is controlled to switch to a first state, and a waiting sequence is executed; After the waiting sequence ends, a second trigger signal is sent to open the signal sensing channel and acquire multi-axis transient vibration data; Based on the energy distribution differences of the multi-axis transient vibration data in different spatial axes, the orthotropic kinetic energy ratio coefficient is calculated, and the orthotropic kinetic energy ratio coefficient is compared with a preset feature threshold to generate a judgment result. After generating the judgment result, the energy transmission path is controlled to switch to the second state.

[0008] Furthermore, the first state is the physically disconnected state of the energy transmission path, and the second state is the physically connected state of the energy transmission path.

[0009] Furthermore, the duration of the waiting sequence is positively correlated with the residual energy amplitude at the instant the energy transmission path switches to the first state; wherein, as the residual energy amplitude increases, the growth slope of the waiting sequence duration tends to flatten.

[0010] Furthermore, the positive correlation mapping is configured using the following formula: in, The duration of the waiting sequence. The preset attenuation constant, For adjustment coefficients, The residual energy amplitude, As the normalization factor, , To preset weights, Based on the bias time.

[0011] Furthermore, the values ​​of the adjustment coefficient, the preset weight, and the basic bias time are configured to ensure that the duration of the waiting sequence is not less than three times the decay constant.

[0012] Furthermore, the process of calculating the orthotropic kinetic energy ratio coefficient includes: using the energy of the first axis in the multi-axis transient vibration data as the denominator and the sum of the energies of the other axes as the numerator, to calculate the ratio.

[0013] Furthermore, the orthotropic kinetic energy ratio coefficient is configured using the following formula: in, The orthotropic kinetic energy ratio coefficient is... The tangential vibration component is located in the multi-axis transient vibration data. The transverse vibration component is located in the multi-axis transient vibration data. The radial vibration component is the first axial direction. These are preset non-zero constants.

[0014] Furthermore, the preset feature threshold is updated based on the average value of the orthogonal kinetic energy ratio coefficient over historical periods.

[0015] Compared with existing technologies, the intelligent tire abnormal vibration signal feature extraction method based on edge computing of the present invention has the following beneficial effects: By using spatial electrical linkage anchoring and disconnecting and delaying the energy transmission path, this invention suppresses electromechanical coupling crosstalk and broadband physical saturation distortion of analog-to-digital converters caused by transient electromagnetic inrush currents. During the transient extreme value phase of signal acquisition, this invention utilizes physical network disconnection and a nonlinear delay algorithm to dissipate inrush charge. The physical network disconnection and nonlinear delay algorithm ensure that the signal sensing channel opens within a preset clearance range after the electromagnetic noise floor decreases, thereby reducing the probability of physical damage stress waves being masked by the background electric field at the acquisition front end.

[0016] Based on the orthogonal kinetic energy ratio coefficient configuration logic, this invention achieves orthogonal decoupling of the superimposed physical fields in three-dimensional space and mathematical normalization of the vehicle speed common-mode factor. In the feature analysis stage, this invention utilizes the orthogonal physical law that parasitic mechanical resonances are concentrated on the radial axis and intrinsic damage stress waves are distributed in the tangential and transverse planes to construct a cross-axis ratio operator. The cross-axis ratio operator uses a division command to remove and cancel the centrifugal common-mode interference variable that increases exponentially with the square of the rotational speed, thus avoiding the defects of low-speed underreporting and high-speed false alarms caused by the all-axial energy summation method.

[0017] By integrating hardware and software mechanisms, this invention resolves the engineering contradiction of digital noise reduction operations relying on floating-point computing resources while maintaining the basic computing power overhead of the microcontroller. This invention combines physical disconnection in the timing dimension with mathematical decoupling in the spatial dimension, sinking the frequency domain filtering logic down to the physical hardware topology and basic arithmetic instructions. The timing disconnection and spatial decoupling mechanisms prevent abnormal execution branches caused by computing power overload and register deadlock, establishing the robustness of the system's underlying operation in a rotating electromechanical coupling environment. Attached Figure Description

[0018] Figure 1 The following is a flowchart of a method for extracting abnormal vibration signals of smart tires based on edge computing in an embodiment of the present invention. Detailed Implementation

[0019] The above technical solutions will now be described in detail with reference to the accompanying drawings and specific embodiments to provide a better understanding of them. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. It should be understood that the present invention is not limited to the exemplary embodiments used only to explain the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention. Furthermore, it should be noted that, for ease of description, only the parts related to the present invention are shown in the drawings, not all of them.

[0020] Example: Please refer to Figure 1This invention provides a method for extracting features of abnormal vibration signals from intelligent tires based on edge computing, comprising the following steps: Space electrical association anchoring steps: acquire motion vector data of the rotating object to calculate transient angular displacement, and acquire the potential change rate of the energy transmission path associated with the rotating object; When the transient angular displacement is within a preset range and the rate of change of potential is greater than a preset trigger threshold, a first trigger signal is issued. In response to the first trigger signal, the energy transmission path is controlled to switch to a first state, and a waiting sequence is executed; After the waiting sequence ends, a second trigger signal is sent to open the signal sensing channel and acquire multi-axis transient vibration data; Based on the energy distribution differences of the multi-axis transient vibration data in different spatial axes, the orthotropic kinetic energy ratio coefficient is calculated, and the orthotropic kinetic energy ratio coefficient is compared with a preset feature threshold to generate a judgment result. After generating the judgment result, the energy transmission path is controlled to switch to the second state.

[0021] The physical platform for implementing the edge computing-based intelligent tire abnormal vibration signal feature extraction method is deployed inside the tire. The edge computing nodes operate using piezoelectric energy harvesting circuits for self-powered operation, and the processing unit consists of a microcontroller operating at extremely low computing power. The entire operating environment is constantly within a centrifugal and electromagnetically coupled physical field. To ensure deterministic command issuance, the mapping relationship between the energy transmission path and the signal sensing channel on the physical hardware is predefined. The energy transmission path is mapped as a closed-loop physical circuit for charge charging and discharging, including transient voltage suppression diodes and energy storage capacitors. The signal sensing channel is mapped as the communication bus between the microcontroller's analog-to-digital converter pins and the multi-axis accelerometer.

[0022] With the microcontroller powering on and initializing, the main process of the abnormal vibration signal feature extraction method enters its execution cycle. The main process aims to establish a physical isolation time window as its initial stage objective. The microcontroller first performs a spatial electrical correlation anchoring step to acquire environmental physical quantities. After acquiring these quantities, the microcontroller cuts off the energy transmission path, putting the circuit in a physically disconnected state. Once in the physically disconnected state, the microcontroller executes a delay compensation sequence based on RC discharge characteristics. After the delay compensation sequence is completed, the microcontroller activates the signal sensing channel to perform multi-axis transient vibration data decoupling. After obtaining the determination result from the decoupling, the microcontroller resets the energy transmission path to a physically connected state.

[0023] The space electrical correlation anchoring step aims to capture the physical extrema of the tire's contact patch. The microcontroller acquires the motion vector data of the rotating object as the input for space anchoring. Guided by the data stream, the microcontroller applies a low-pass filtering algorithm to extract the radially pointing low-frequency DC component from the motion vector data. The low-frequency DC component represents the transient physical projection of the Earth's gravitational field onto the radial axis of the rotating object. After extracting the low-frequency DC component, the microcontroller uses an inverse trigonometric function formula to convert it into a transient angular displacement variable representing the rotation phase. As an alternative approach, the microcontroller outputs the transient angular displacement variable by querying an arcsine mapping data table pre-programmed into non-volatile memory, thereby reducing the computational overhead of floating-point operations.

[0024] To compensate for geometric projection deviations caused by physical deformation, an electrical measurement and verification phase is initiated. While outputting the transient angular displacement variable, the microcontroller simultaneously acquires discrete potential sampling points along the energy transmission path associated with the rotating object via an analog-to-digital converter. The microcontroller performs first-order differential time-varying operations on the continuously acquired discrete potential sampling points, outputting the potential change rate. The potential change rate directly corresponds to the extreme value of the charge accumulation rate generated by the compressive deformation of the piezoelectric transducer material matrix. After both the transient angular displacement variable and the potential change rate are extracted, a threshold Boolean judgment logic is executed. The microcontroller sends the transient angular displacement variable to a first logic register for inclusion comparison with a preset interval, and simultaneously sends the potential change rate to a second logic register for magnitude determination against a preset trigger threshold. When both conditions are met—the transient angular displacement variable being within the preset interval and the potential change rate being greater than the preset trigger threshold—the microcontroller's general-purpose input / output port sends a first trigger signal.

[0025] With the issuance of the first trigger signal, the abnormal vibration signal feature extraction method enters the static window dynamic scheduling stage based on charge relaxation. The static window dynamic scheduling stage aims to cut off the electromechanical coupling noise source and establish a physical isolation time window. After the microcontroller's general-purpose input / output port outputs the first trigger signal, the power management control pin of the energy transmission path receives the first trigger signal. In response to the first trigger signal, the microcontroller pulls down the gate voltage of the field-effect transistor inside the energy transmission path, forcing the energy transmission path to switch from the physically on state to the first state. The first state is the physically off state of the energy transmission path. Synchronously with the energy transmission path switching to the first state, the microcontroller executes the working mode state machine switching instruction, switching the current operating mode from low-power monitoring mode to feature extraction mode. In low-power monitoring mode, the microcontroller only maintains the lowest frequency spatial electrical parameter acquisition task; in feature extraction mode, the microcontroller activates the full-speed operation clock of the digital signal processing unit and analog-to-digital converter to meet the high-frequency acquisition requirements of subsequent multi-axis transient vibration data.

[0026] The technique for forcing the energy transfer path into a physically disconnected state involves the following: When the piezoelectric transducer compresses at its deformation extreme point, it generates a transient electromagnetic inrush current of several hundred volts. This inrush current is directly coupled to the microcontroller's ground terminal through the energy transfer path, thus submerging the microvolt-level physical vibration signal. Cutting off the energy transfer path effectively suppresses the coupling of the electromagnetic inrush current at the hardware physical topology level. As an alternative technique, the microcontroller controls the energy transfer path to switch to the first state by triggering the disconnect contact of an external solid-state relay.

[0027] When the energy transfer path switches to the first state, some undissipated charge remains inside the energy storage capacitor network of the energy transfer path. To ensure that the remaining charge is completely dissipated, the microcontroller immediately triggers its internal hardware timer to execute a waiting sequence after controlling the energy transfer path to switch to the first state. The waiting sequence has a duration. To dynamically match the charge dissipation requirements under different impact physical field intensities, the arithmetic logic unit of the microcontroller calculates the duration based on the residual energy amplitude at the instant the energy transfer path switches to the first state. The arithmetic logic unit executes a positive correlation mapping algorithm to adjust the duration. It is positively correlated with the residual energy amplitude. As the residual energy amplitude increases, the arithmetic logic unit limits the duration by controlling the algorithm parameters. The growth rate tends to flatten out.

[0028] Arithmetic Logic Unit (ALU) computation time The calculation process is configured using the following formula: in, The duration of the waiting sequence. The preset attenuation constant, For adjustment coefficients, This represents the residual energy amplitude. As the normalization factor, , To preset weights, Based on the bias time.

[0029] To simulate the physical curve of residual energy decay in a real circuit, the arithmetic logic unit (ALU) uses the hyperbolic tangent function as the activation kernel of the nonlinear mapping operator. The hyperbolic tangent function's calculated output strictly converges to a constant 1 as the input variable approaches infinity. The ALU utilizes the convergence characteristics of the hyperbolic tangent function to approximate the limiting decay law of a physical circuit with clamping protection diodes when encountering a limiting voltage surge. The hyperbolic tangent function ensures that regardless of the magnitude of the residual energy amplitude, the duration calculated by the ALU is consistent. All convergences towards a preset physical upper limit, preventing the microcontroller from falling into register deadlock due to calculating infinitely large time values ​​at the instruction level. As a degradation alternative for low-performance microcontrollers, the arithmetic logic unit, when disabling the floating-point unit, outputs a duration approximating the above formula configuration by addressing and querying a pre-stored exponentially damped piecewise linear mapping table. .

[0030] Duration After the waiting sequence, the abnormal vibration signal feature extraction method enters the spatial eigenstate decoupling and feature extraction stage. The microcontroller's timer interrupt flag flips after the countdown of the waiting sequence reaches zero. This flip triggers the microcontroller to send a second trigger signal. In response to the second trigger signal, the microcontroller's serial peripheral interface bus sends a wake-up command to the sensor array to activate the signal sensing channel. Once the signal sensing channel is activated, the microcontroller acquires multi-axis transient vibration data through the analog-to-digital converter interface. This acquisition of multi-axis transient vibration data occurs after the charge relaxation operation of the waiting sequence, ensuring that the multi-axis transient vibration data is captured within a physical window where the electromagnetic noise floor is lower than the least significant bit quantization error.

[0031] After acquiring multi-axis transient vibration data, the microcontroller's digital signal processing unit analyzes the differences in energy distribution along different spatial axes. Inside the rotating object, limited by the two-dimensional extensibility of the physical material matrix, the intrinsic damage signal of the high-frequency stress wave caused by the fracture of the fabric layer results in the mechanical energy of the intrinsic damage signal being concentrated in the tangential and transverse axes parallel to the axis of rotation. Conversely, driven by the centrifugal physical field generated by the rotation of the object, the non-intrinsic interference signal attached to the sensor's own mass block produces a periodic stretching phenomenon along the normal of the rotating object, causing the mechanical energy of the non-intrinsic interference signal to be concentrated in the radial axis.

[0032] Based on the aforementioned orthogonal distribution law of the physical field, the digital signal processing unit (DSP) performs multidimensional spatial decoupling calculations. The DSP uses the energy along the first axis of the multi-axis transient vibration data as the denominator, and this first axis is designated as the radial axis. Simultaneously, the DSP uses the sum of the energies along the other axes as the numerator, and these other axes are designated as the tangential and transverse axes. The DSP executes a division instruction to calculate the ratio, thereby generating the orthogonal kinetic energy ratio coefficient.

[0033] The digital signal processing unit calculates the orthogonal kinetic energy ratio coefficient using the following formula: in, The kinetic energy ratio coefficient is the coefficient for orthogonal anisotropy. The tangential vibration component in multi-axis transient vibration data. This refers to the lateral vibration component in multi-axis transient vibration data. The radial vibration component is the first axial direction. These are preset non-zero constants.

[0034] During integration, the digital signal processing unit (DSP) calculates and adds the square integrals of the tangential and transverse vibration components within the sampling time window to determine the total fracture kinetic energy in the parallel plane. The DSP also calculates the square integral of the radial vibration component to determine the total resonant energy of the radial normal plane. When the rotational speed of the rotating object increases dramatically, the physical amplitudes of the tangential, transverse, and radial vibration components are all linearly amplified proportionally by the centrifugal field. The DSP achieves physical normalization of the rotational speed amplification effect by placing the square integral of the radial vibration component (including centrifugal interference) in the denominator of the division instruction and using ratio calculation. The division instruction forcibly cancels the rotational speed interference variable as a common-mode factor from the expression, ensuring that the orthotropic kinetic energy ratio coefficient output by the DSP is only correlated with the intrinsic damage state of the physical structure.

[0035] To prevent the risk of abnormal system termination, the digital signal processing unit forcibly adds a preset non-zero constant to the denominator accumulator register before executing the floating-point division instruction. This blocks the execution path that would cause the microcontroller to crash due to the radial vibration component being zero under static conditions. After generating the orthotropic kinetic energy ratio coefficient, the microcontroller's numerical comparator extracts the orthotropic kinetic energy ratio coefficient and compares it with a preset feature threshold by performing a difference operation. Based on the difference operation comparison result, a Boolean type judgment result is output.

[0036] With the generation of the judgment result, the abnormal vibration signal feature extraction method enters the adaptive closed-loop and hardware reset stage. Using the orthogonal kinetic energy ratio coefficient as the input source, the microcontroller's direct memory access controller extracts the historical periodic orthogonal kinetic energy ratio coefficient sequence stored in the static random access memory. After extracting the historical periodic orthogonal kinetic energy ratio coefficient sequence, the microcontroller's arithmetic logic unit executes the sliding window mean calculation instruction. The arithmetic logic unit directly overwrites the latest value output by the sliding window mean calculation into the preset feature threshold block inside the non-volatile register. After completing the dynamic overwriting update of the preset feature threshold, the microcontroller sends a high-level closing instruction to the power management control pin of the energy transmission path. The internal field-effect transistor of the energy transmission path receives the high-level closing instruction and restores the physical conduction structure, causing the energy transmission path to switch from the first state to the second state. The second state is the physical conduction state of the energy transmission path. The establishment of the physical conduction state reopens the physical circuit for charge injection from the piezoelectric transducer to the energy storage capacitor network, and the system state machine converges to the energy harvesting ready state before the next event is triggered.

[0037] To ensure that the core parameter calibration of the abnormal vibration signal feature extraction method has a physical hardware basis and achieves dimensional consistency in the underlying calculation, the arithmetic logic unit loads a series of preset constant parameters before performing the mapping calculation of the waiting sequence duration. The value of the attenuation constant is substantially limited by the physical product of the actual capacitance value of the energy storage capacitor network and the equivalent impedance value of the energy transmission path at the physical hardware level. The value of the basic bias time is substantially limited by the physical time constant of the cold start wake-up of the analog-to-digital converter inside the signal sensing channel. To meet the dimensionless mathematical requirement of the hyperbolic tangent function algorithm for verifying the input independent variable, the physical dimension of the normalization factor is strictly configured as the reciprocal of the physical dimension of the residual energy amplitude. With the execution of the multiplication instruction of the normalization factor and the residual energy amplitude, the input parameters of the hyperbolic tangent function are transformed into dimensionless pure numbers. At the same time, the adjustment coefficient and the preset weights are both configured by the system as dimensionless pure numerical weight parameters. After the dimensionless constant term is added to and multiplied by the dimensionless function output value, it is finally multiplied by the decay constant with time dimension to ensure that the final dimension of the waiting sequence duration output by the arithmetic logic unit is strictly equal to the time dimension of the decay constant.

[0038] Under extreme physical conditions, the above parameter configurations collectively construct a robust defense boundary for system logic convergence. When the piezoelectric transducer experiences an extreme physical impact causing the residual energy amplitude to approach infinity, the microcontroller enters a deadlock-prevention limit verification state. As the residual energy amplitude approaches infinity, the mathematical output limit of the hyperbolic tangent function algorithm kernel strictly approximates and converges to the constant 1. After the arithmetic logic unit executes addition and multiplication instructions, the calculation result of the waiting sequence duration does not exhibit an infinite divergence, but converges to the physical time upper limit value jointly constituted by the product of the adjustment coefficient, the decay constant, and the preset weight, plus the basic bias time. The existence of the physical time upper limit value forcibly blocks the abnormal branch of the microcontroller's hardware timer that would fall into an instruction deadlock state due to loading infinite values.

[0039] To mitigate the risk of division-by-zero core crashes, the microcontroller employs a division-by-zero overflow prevention mechanism. When a rotating object is in a steady-state reference physical state, the radial vibration component in the multi-axis transient vibration data output by the analog-to-digital converter approaches a digital zero value. Before executing the division instruction for the orthogonal kinetic energy ratio coefficient, the microcontroller's arithmetic logic unit forcibly pushes a preset non-zero constant into the denominator accumulation register of the division instruction. This preset non-zero constant acts as a denominator bias term when the square integral of the radial vibration component is zero, significantly reducing the probability of the underlying hardware throwing a division-by-zero exception.

[0040] To address the high-frequency electromagnetic inrush current dissipation requirements, the system performs physical clearance boundary verification. According to the natural logarithmic decay law of first-order RC discharge circuits in physics, when the energy transmission path is in the first state for a time span reaching three times the decay constant, the residual energy amplitude decays exponentially to 4.98% of the initial peak value at the moment of disconnection. By forcibly configuring adjustment coefficients and preset weights to ensure the waiting sequence duration is no less than three times the decay constant, 95.02% of the electromagnetic inrush current interference charge is completely dissipated by the physical grounding loop. The residual background electric field intensity after dissipating 95.02% of the electromagnetic inrush current interference charge is stably lower than the minimum effective bit quantization error amplitude of the signal sensing channel data acquisition pin. These charge decay physical phenomena verify the theoretical optimality of using three times the decay constant as a hard physical constraint on the waiting sequence duration.

[0041] For 8-bit microcontrollers lacking floating-point arithmetic units, an abnormal vibration signal feature extraction method provides an operator-degraded execution path. When the microcontroller's arithmetic logic unit (ALU) performs the positive correlation mapping calculation for the waiting sequence duration, it bypasses the hyperbolic tangent function calculation instruction. As a specific implementation of the degraded execution path, the microcontroller's direct memory access controller reads a pre-programmed nonlinear mapping discrete data table from read-only memory. After reading the nonlinear mapping discrete data table, the ALU extracts the high-order data byte of the residual energy amplitude as an address index to obtain two adjacent basic discrete mapping values ​​from the nonlinear mapping discrete data table. After obtaining the two adjacent basic discrete mapping values, the ALU uses the low-order data byte of the residual energy amplitude to perform piecewise linear interpolation and shift operations on the two adjacent basic discrete mapping values. The combined execution of piecewise linear interpolation and shift operations approximates and outputs a waiting sequence duration value with physical saturation convergence characteristics while avoiding computationally expensive floating-point multiplication and division instructions.

[0042] The abnormal vibration signal feature extraction method uses a pre-defined mathematical substitution operator. In the calculation of the waiting sequence duration, the microcontroller's arithmetic logic unit (ALU) calls an exponential damping function to replace the hyperbolic tangent function. The ALU uses the natural constant as the base and the negative of the product of the normalization factor and the residual energy amplitude as the exponent to execute the natural exponential decay calculation instruction. After executing the natural exponential decay calculation instruction, the ALU performs a difference operation between the constant 1 and the output result of the natural exponential decay calculation. The output result of the difference operation exhibits a sharp linear tracking slope in the low residual energy amplitude range and strictly converges to the preset upper limit of the constant in the high residual energy amplitude range, producing a silent window physical scheduling effect completely equivalent to the hyperbolic tangent function.

[0043] In the process of calculating the orthogonal kinetic energy ratio coefficient, the microcontroller's digital signal processing unit (DSP) replaces continuous-time integration instructions with discrete sampling comparison logic. The DSP extracts discrete tangential vibration component sampling sequences and discrete transverse vibration component sampling sequences within a preset time window, and calculates the discrete root mean square (RMS) values ​​of both sequences. After calculating the RMS values, the DSP performs a digital division ratio calculation using the RMS value of the discrete radial vibration component sampling sequence along the first axial direction as the denominator. This ratio calculation based on the RMS values ​​also achieves the physical normalization goal of removing the common-mode interference factor of the rotating object's rotation.

[0044] The abnormal vibration signal feature extraction method is implemented within an edge sensing node hardware architecture. This hardware architecture is encapsulated and deployed inside a rotating object. It comprises a microcontroller, a multi-axis accelerometer, a power management integrated circuit (power management IC), and a piezoelectric transducer. The electrode ports of the piezoelectric transducer are physically connected to the charge input ports of the power management IC, forming a charge collection loop. The voltage output pins of the power management IC are connected to the power supply port of the microcontroller via a power bus. The microcontroller's general purpose input / output (GPIO) ports are connected to the logic control enable pins of the power management IC. The microcontroller sends a first trigger signal through the GPIO ports to force the switching between open and closed states of the energy transfer path. The microcontroller's serial peripheral interface (SPI) bus is physically connected to the data communication pins of the multi-axis accelerometer, forming a signal sensing channel. The microcontroller sends a second trigger signal via the SPI bus and relies on the SPI bus to read multi-axis transient vibration data.

[0045] The abnormal vibration signal feature extraction method, which includes a cross-domain cooperative scheduling algorithm and spatial decoupling computation logic, is compiled into low-level machine execution instructions by a cross-compiler. These instructions are then programmed and stored in the microcontroller's internal non-volatile memory. The non-volatile memory is specifically instantiated as an electrically erasable programmable read-only memory chip or a flash memory chip. The microcontroller's central processing unit (CPU) reads the low-level machine execution instructions from the non-volatile memory clockwise via its internal system data bus. After reading the instructions, the CPU triggers the arithmetic logic unit (ALU), direct memory access controller (DRAM), and digital signal processing unit (DSP) to perform corresponding physical level switching, logic gate state transitions, and hardware register value transfer operations. The storage and execution of these low-level machine execution instructions within the microcontroller enables the abnormal vibration signal feature extraction method to run as a computer program within an electronic device with physical spatiotemporal consumption attributes.

[0046] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for extracting features of abnormal vibration signals from intelligent tires based on edge computing, characterized in that, Includes the following steps: Space electrical association anchoring steps: acquire motion vector data of the rotating object to calculate transient angular displacement, and acquire the potential change rate of the energy transmission path associated with the rotating object; When the transient angular displacement is within a preset range and the rate of change of potential is greater than a preset trigger threshold, a first trigger signal is issued. In response to the first trigger signal, the energy transmission path is controlled to switch to a first state, and a waiting sequence is executed; After the waiting sequence ends, a second trigger signal is sent to open the signal sensing channel and acquire multi-axis transient vibration data; Based on the energy distribution differences of the multi-axis transient vibration data in different spatial axes, the orthotropic kinetic energy ratio coefficient is calculated, and the orthotropic kinetic energy ratio coefficient is compared with a preset feature threshold to generate a judgment result. After generating the judgment result, the energy transmission path is controlled to switch to the second state.

2. The method according to claim 1, characterized in that, The first state is the physical disconnection state of the energy transmission path, and the second state is the physical connection state of the energy transmission path.

3. The method according to claim 1, characterized in that, The duration of the waiting sequence is positively correlated with the residual energy amplitude at the instant the energy transmission path switches to the first state; wherein, as the residual energy amplitude increases, the growth slope of the waiting sequence duration tends to flatten.

4. The method according to claim 3, characterized in that, The positive correlation mapping is configured using the following formula: in, The duration of the waiting sequence. The preset attenuation constant, For adjustment coefficients, The residual energy amplitude, As the normalization factor, , To preset weights, Based on the bias time.

5. The method according to claim 4, characterized in that, The values ​​of the adjustment coefficient, the preset weight, and the basic bias time are configured to ensure that the duration of the waiting sequence is not less than three times the decay constant.

6. The method according to claim 1, characterized in that, The process of calculating the orthotropic kinetic energy ratio coefficient includes: using the energy of the first axis in the multi-axis transient vibration data as the denominator and the sum of the energies of the other axes as the numerator, to calculate the ratio.

7. The method according to claim 6, characterized in that, The orthotropic kinetic energy ratio coefficient is configured using the following formula: in, The orthotropic kinetic energy ratio coefficient is... The tangential vibration component is located in the multi-axis transient vibration data. The transverse vibration component is located in the multi-axis transient vibration data. The radial vibration component is the first axial direction. These are preset non-zero constants.

8. The method according to claim 1, characterized in that, The preset feature threshold is updated based on the average value of the orthotropic kinetic energy ratio coefficient over historical periods.

Images