City wisdom guardrail real-time monitoring method based on internet of things
By applying compressed sensing and sparse reconstruction algorithms to IoT sensor terminals at urban guardrail nodes, the communication load and power consumption issues in high-frequency vibration signal monitoring are resolved, enabling efficient monitoring and diagnosis of impact events and the health status of guardrail structures.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- SHENZHEN DECIMETER INTERNET TECH CO LTD
- Filing Date
- 2026-02-28
- Publication Date
- 2026-05-29
Smart Images
Figure CN122120838A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of Internet of Things (IoT) technology, and in particular to a method for real-time monitoring of smart urban guardrails based on IoT. Background Technology
[0002] In practical applications of IoT technology for municipal facility monitoring, the common approach is to directly transmit raw sampled data or only upload alarm information when a threshold is triggered. When dealing with high-frequency vibration signal acquisition scenarios, directly transmitting uncompressed raw data leads to prolonged high-load operation of the wireless communication link, consuming terminal device power, reducing battery life, requiring frequent battery replacements or manual maintenance, and increasing overall operation and maintenance costs. Furthermore, concurrent transmission of large amounts of data can easily cause channel congestion in areas at the edge of network signal coverage, resulting in data packet loss or transmission delays, leading to incomplete data during critical monitoring periods. Therefore, improvements are needed. Summary of the Invention
[0003] The purpose of this invention is to address the shortcomings of existing technologies by proposing a real-time monitoring method for smart urban guardrails based on the Internet of Things.
[0004] To achieve the above objectives, the present invention adopts the following technical solution: a real-time monitoring method for smart urban guardrails based on the Internet of Things, comprising the following steps:
[0005] The original vibration signals of the guardrails are collected by IoT sensor terminals deployed at the nodes of the city guardrails. The original vibration signals are then projected and dimension-reduced using a preset compressed sensing observation matrix to obtain a compressed measurement data packet.
[0006] The compressed measurement data packet is transmitted to a cloud server via a wide area network, and the compressed measurement data packet is solved using a sparse reconstruction algorithm to obtain the reconstructed time-domain vibration waveform.
[0007] Time-frequency features are extracted from the reconstructed time-domain vibration waveform, and the extracted feature data is used to determine whether a collision event has occurred with the guardrail and to identify the specific type of the collision event.
[0008] When it is determined that no impact event has occurred and the reconstructed time-domain vibration waveform has a free decay trend, the attenuation tail wave data segment in the reconstructed time-domain vibration waveform is extracted, the fastening integrity assessment index of the guardrail structure is calculated based on the attenuation tail wave data segment, and the fastening integrity assessment index is compared with the preset safety benchmark threshold to determine the loosening state of the guardrail connectors.
[0009] Preferably, the step of using a preset compressed sensing observation matrix to perform projection dimensionality reduction processing on the original vibration signal to obtain a compressed measurement value data packet specifically includes:
[0010] A random observation matrix following a Bernoulli distribution is pre-set in the microcontroller of the IoT sensing terminal. The number of columns of the random observation matrix is consistent with the number of sampling points of the original vibration signal, and the number of rows of the random observation matrix is much smaller than the number of columns.
[0011] The original vibration signal is constructed into a digital signal vector in column vector form. The random observation matrix is multiplied by the digital signal vector through matrix multiplication to obtain an observation vector with reduced dimension.
[0012] The observation vector is quantized and encoded, and timestamp tags and node identification are added according to a preset data packet format to generate the compressed measurement value data packet;
[0013] The data volume of the compressed measurement value data packet is 20% to 30% of the original vibration signal data volume, and the IoT sensing terminal only needs to perform a wireless transmission operation on the compressed measurement value data packet.
[0014] Preferably, the specific steps for solving the compressed measurement data packet using a sparse reconstruction algorithm to obtain the reconstructed time-domain vibration waveform are as follows:
[0015] After receiving the compressed measurement data packet, the cloud server parses out the observation vector;
[0016] Construct or call a pre-stored sensing matrix, which is obtained by multiplying the random observation matrix used by the IoT sensing terminal with a preset sparse transformation basis matrix;
[0017] Initialize the residual vector to the observation vector, initialize the index set to an empty set, and enter the orthogonal matching pursuit iteration loop;
[0018] In each iteration, the inner product of the residual vector and all column vectors in the sensing matrix is calculated, the column index corresponding to the inner product with the largest absolute value is found, and the column index is added to the index set.
[0019] Extract the corresponding submatrix from the sensing matrix according to the index set, solve the projection coefficient of the observation vector on the space spanned by the submatrix using the least squares method, and update the residual vector to the difference between the observation vector and the estimated observation value corresponding to the projection coefficient;
[0020] Determine whether the norm of the residual vector is less than a preset reconstruction accuracy threshold. If it is less, stop the iteration and use the finally obtained sparse coefficient vector and the sparse transformation basis matrix to recover the reconstructed time-domain vibration waveform.
[0021] Preferably, the steps of extracting time-frequency features from the reconstructed time-domain vibration waveform and determining whether a collision event has occurred on the guardrail and identifying the specific type of the collision event based on the extracted feature data are as follows:
[0022] Perform a fast Fourier transform on the reconstructed time-domain vibration waveform to obtain frequency-domain power spectral density distribution data;
[0023] Calculate the short-time energy value and zero-crossing rate value of the reconstructed time-domain vibration waveform within a preset time window;
[0024] The dominant peak frequency in the frequency domain power spectral density distribution data is matched with a preset wind-induced vibration frequency range, and the short-time energy value is compared with a preset vehicle impact energy threshold.
[0025] If the peak frequency of the dominant frequency is within the wind-induced vibration frequency range and the short-time energy value is less than the vehicle impact energy threshold, then the current event is determined to be a natural vibration caused by environmental wind pressure.
[0026] If the short-time energy value is greater than the vehicle impact energy threshold and the amplitude decay rate of the reconstructed time-domain vibration waveform is greater than the preset rigid collision decay threshold, then the current event is determined to be a vehicle impacting the guardrail event.
[0027] Preferably, when it is determined that no impact event has occurred and the reconstructed time-domain vibration waveform exhibits a free decay trend, the step of extracting the decayed tailwave data segment from the reconstructed time-domain vibration waveform specifically involves:
[0028] Envelope detection is performed on the reconstructed time-domain vibration waveform to obtain the waveform amplitude envelope;
[0029] Traverse the waveform amplitude envelope to find waveform segments that satisfy the monotonically decreasing condition and whose duration exceeds the preset minimum decay period;
[0030] The start time of the waveform segment is defined as the time when the vibration excitation disappears, and the time when the amplitude of the waveform segment decreases to a preset noise floor level is defined as the time when the decay ends.
[0031] The waveform data between the start time and the end time of attenuation is extracted, and the extracted data is bandpass filtered to remove high-frequency noise interference, resulting in the denoised attenuation tailwave data segment.
[0032] Preferably, the step of calculating the fastening integrity assessment index of the guardrail structure based on the attenuated wake data segment is as follows:
[0033] Identify all local maxima in the attenuated tailwave data segment and construct a set of peak sequences;
[0034] Select the amplitude of the first peak at the initial moment and the amplitude of the final peak after a preset M vibration cycles from the set of peak sequences.
[0035] The logarithmic decay rate is calculated based on the amplitude of the first peak and the amplitude of the final peak, and a calculation model for the fastening integrity assessment index is constructed by combining it with a preset nonlinear amplitude correction factor.
[0036] The fastening integrity assessment index is derived using the aforementioned calculation model, and the formula corresponding to the calculation model is as follows: ;in, This represents the fastening integrity assessment index. This represents the preset normalized weighting coefficients. This indicates the number of vibration periods between the two selected wave peaks. This represents the amplitude of the first peak. This indicates the amplitude of the termination peak. Let represent the logarithmic function with base e. This represents the baseline logarithmic decay rate of the guardrail under standard factory tightening conditions. This represents the nonlinear amplitude correction factor. This indicates the maximum allowable elastic deformation amplitude of the guardrail material.
[0037] Preferably, the step of comparing the fastening integrity assessment index with a preset safety benchmark threshold to determine the loosening state of the guardrail connector specifically includes:
[0038] Read the historical monitoring database stored in the cloud server to obtain the historical fastening integrity assessment index sequence of the guardrail nodes within a preset time period;
[0039] The average value of the historical fastening integrity assessment index sequence is calculated as a dynamic reference benchmark, and the safety benchmark threshold is set to 80% of the dynamic reference benchmark.
[0040] Compare the current fastening integrity assessment index with the safety benchmark threshold.
[0041] If the fastening integrity assessment index is greater than or equal to the safety benchmark threshold, the guardrail connector is determined to be in a fastened state.
[0042] If the fastening integrity assessment index is less than the safety benchmark threshold, the guardrail connector is determined to be loose, and a maintenance warning signal of the corresponding level is generated based on the difference between the fastening integrity assessment index and the safety benchmark threshold.
[0043] Preferably, the specific steps for collecting the raw vibration signals of the guardrails through IoT sensor terminals deployed at urban guardrail nodes are as follows:
[0044] The mechanical vibration of the guardrail is sensed by a piezoelectric accelerometer installed at the connection between the guardrail post and the crossbeam.
[0045] The charge signal output by the piezoelectric accelerometer is converted into a voltage signal by a charge amplifier;
[0046] The voltage signal is subjected to high-frequency filtering using an anti-aliasing filter.
[0047] The processed voltage signal is discretized by an analog-to-digital converter at a preset Nyquist sampling frequency to generate a time-series discrete digital signal, which serves as the original vibration signal.
[0048] The IoT sensing terminal automatically enters sleep mode when it detects that the amplitude of the original vibration signal is lower than a preset silence threshold, and triggers the projection dimensionality reduction process only when the amplitude exceeds the silence threshold.
[0049] Compared with the prior art, the advantages and positive effects of the present invention are as follows:
[0050] In this invention, IoT sensor terminals deployed at urban guardrail nodes collect raw vibration signals from the guardrails. A preset compressed sensing observation matrix is used to perform projection dimensionality reduction on the raw vibration signals, generating compressed measurement value data packets. This reduces the size of the data to be transmitted, alleviates network bandwidth pressure, and effectively reduces power consumption at the edge. The compressed measurement value data packets are transmitted to a cloud server via a wide area network. A sparse reconstruction algorithm is used to solve and reconstruct the reconstructed time-domain vibration waveform with high fidelity. This transfers complex signal processing and high-computing-power tasks from low-power terminals to the cloud, optimizing the allocation of computing resources. Time-frequency features are extracted from the reconstructed time-domain vibration waveform, and the impact event type is identified based on the feature data, ensuring rapid response and handling of traffic accidents. Even when no impact has occurred and the waveform exhibits a free decay trend, the decay tailwave data segment is extracted. By calculating the guardrail structure's tightness integrity assessment index and comparing it with a safety benchmark threshold, the looseness of the connectors can be quantitatively determined. This allows for simultaneous external impact monitoring and internal structural health diagnosis using a single vibration source without the need for additional stress sensors. Attached Figure Description
[0051] Figure 1 This is a schematic diagram of the steps of the present invention. Detailed Implementation
[0052] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.
[0053] Please see Figure 1 This invention provides a technical solution: a real-time monitoring method for smart urban guardrails based on the Internet of Things, comprising the following steps:
[0054] The original vibration signals of the guardrails are collected by IoT sensor terminals deployed at the nodes of the city guardrails. The original vibration signals are then projected and dimension-reduced using a preset compressed sensing observation matrix to obtain a compressed measurement data packet.
[0055] The compressed measurement data packet is transmitted to the cloud server via a wide area network. The sparse reconstruction algorithm is used to solve the compressed measurement data packet to obtain the reconstructed time-domain vibration waveform.
[0056] Time-frequency features are extracted from the reconstructed time-domain vibration waveform. Based on the extracted feature data, it is determined whether a collision event has occurred with the guardrail and the specific type of the collision event is identified.
[0057] When it is determined that no impact event has occurred and the reconstructed time-domain vibration waveform shows a free decay trend, the attenuated tail wave data segment in the reconstructed time-domain vibration waveform is extracted. The fastening integrity assessment index of the guardrail structure is calculated based on the attenuated tail wave data segment, and the fastening integrity assessment index is compared with the preset safety benchmark threshold to determine the loosening state of the guardrail connectors.
[0058] In this embodiment, the steps of using a preset compressed sensing observation matrix to project and reduce the dimensionality of the original vibration signal to obtain a compressed measurement value data packet are as follows: A random observation matrix following a Bernoulli distribution is preset in the microcontroller of the IoT sensing terminal. The number of columns of the random observation matrix is the same as the number of sampling points of the original vibration signal, and the number of rows of the random observation matrix is much smaller than the number of columns. The original vibration signal is constructed into a digital signal vector in column vector form. The random observation matrix is multiplied by the digital signal vector through matrix multiplication to obtain the observation vector with reduced dimensionality. The observation vector is quantized and encoded. A timestamp tag and node identity identifier are added according to the preset data packet format to generate a compressed measurement value data packet. The data volume of the compressed measurement value data packet is 20% to 30% of the data volume of the original vibration signal. The IoT sensing terminal only needs to perform a wireless transmission operation on the compressed measurement value data packet.
[0059] Specifically, a random observation matrix following a Bernoulli distribution is pre-configured in the microcontroller of the IoT sensing terminal. This matrix is not generated through real-time computation but is constructed using a pseudo-random sequence generated by a pre-defined linear feedback shift register algorithm combined with a specific initial seed. The number of columns in the random observation matrix is set to match the single sampling window length N of the original vibration signal; for example, N can be 1024 or 2048. The number of rows M is set to 20% to 30% of the number of columns based on the compression ratio requirement. For example, if N is 1024 and the compression ratio is 0.25, then M is set to 256. The elements in this matrix contain only two values: +1 and -1, or 0 and 1, depending on the Bernoulli distribution parameters. During the data acquisition phase, the original vibration signal after analog-to-digital conversion is temporarily stored in the microcontroller's circular buffer. Once the buffer is filled with N data points, these data are read and arranged in chronological order to construct a column vector of digital signals. Then, the matrix multiplication unit is activated to execute the compressed sensing observation matrix. With digital signal vector Multiplication operation Due to the sparsity or binary nature of the observation matrix, this multiplication operation is quickly performed inside the microcontroller using accumulators and shift registers, eliminating the need for complex floating-point operations. The result is the observation vector with reduced dimensionality. Its length is M, followed by the observation vector Each floating-point value in the data is quantized and mapped to a 16-bit or 32-bit fixed-point integer to reduce storage space. At the same time, the current timestamp is obtained from the real-time clock register inside the microcontroller, and the unique MAC address of the node, which is pre-programmed from non-volatile memory, is read as an identification identifier. According to the custom communication protocol frame format, the frame header, node identification identifier, timestamp, quantized observation vector data, and checksum are sequentially concatenated to generate the final compressed measurement value data packet. The effective payload size of this data packet is only about one-quarter of the original signal data size. Then, the RF transmission circuit in low-power mode is woken up, and the data packet is transmitted through the LoRa or NB-IoT wireless network to complete a data reporting process.
[0060] In this embodiment, the steps for solving the compressed measurement data packet using the sparse reconstruction algorithm to obtain the reconstructed time-domain vibration waveform are as follows: After receiving the compressed measurement data packet, the cloud server parses the observation vector, constructs or calls a pre-stored sensing matrix. The sensing matrix is obtained by multiplying the random observation matrix used by the IoT sensing terminal with a preset sparse transformation basis matrix. The residual vector is initialized as the observation vector, and the index set is initialized as an empty set. The orthogonal matching pursuit iteration loop is entered. In each iteration, the inner product of the residual vector and all column vectors in the sensing matrix is calculated. The column index corresponding to the inner product with the largest absolute value is found and added to the index set. The corresponding sub-matrix is extracted from the sensing matrix according to the index set. The projection coefficients of the observation vector on the space spanned by the sub-matrix are solved using the least squares method. The residual vector is updated to the difference between the estimated observation value corresponding to the observation vector and the projection coefficient. It is determined whether the norm of the residual vector is less than the preset reconstruction accuracy threshold. If it is less, the iteration is stopped. The reconstructed time-domain vibration waveform is recovered using the finally obtained sparse coefficient vector and sparse transformation basis matrix.
[0061] Specifically, after receiving the compressed measurement data packet, the cloud server first verifies the validity of the data packet according to the protocol frame header and extracts the observation vector. The system then allocates computing space in the server's memory, along with the corresponding timestamp and node ID, to construct or call a pre-stored sensor matrix. The construction process of this sensing matrix is based on compressed sensing theory, which transforms the random observation matrix used by IoT sensing terminals. With the preset sparse transformation basis matrix Perform matrix multiplication, i.e. , where the sparse transformation basis matrix Typically, a discrete cosine transform basis or a wavelet transform basis is chosen to ensure that the vibration signal has sparsity in this transform domain. The solution process then proceeds to the orthogonal matching pursuit (OMP) algorithm, first initializing the residual vector. equal to the observation vector Initialize the index set Set an iteration counter for an empty set. Enter the iteration loop, and in the k-th iteration, calculate the current residual vector. With sensing matrix Find the inner product of all column vectors in the array, resulting in a column inner product value. Iterate through these inner product values and find the column index corresponding to the inner product with the largest absolute value. Add the index to the index set According to the updated index set From the sensor matrix Extract the corresponding column vectors to form a submatrix The least squares method is used to solve for the observation vector. Projection coefficients across the space spanned by the submatrix The calculation formula is: Then update the residual vector. Calculate the updated residual vector norm and compare it with the preset reconstruction accuracy threshold. The threshold is compared. Typically, the residual norm is set to 1% to 5% of the energy of the observation vector. If the residual norm is less than the threshold or the number of iterations reaches a preset upper limit (e.g., M / 2), the iteration stops, and the final sparse coefficient vector is output. Finally, the sparse transformation basis matrix is used. with sparse coefficient vector Perform inverse transformation operation This allows for the high-fidelity recovery of the reconstructed time-domain vibration waveform.
[0062] In this embodiment, the steps of extracting time-frequency features from the reconstructed time-domain vibration waveform and determining whether a collision event has occurred with the guardrail and identifying the specific type of the collision event based on the extracted feature data are as follows: A fast Fourier transform is performed on the reconstructed time-domain vibration waveform to obtain frequency domain power spectral density distribution data. The short-time energy value and zero-crossing rate of the reconstructed time-domain vibration waveform within a preset time window are calculated. The dominant peak frequency in the frequency domain power spectral density distribution data is matched with a preset wind-induced vibration frequency range. Simultaneously, the short-time energy value is compared with a preset vehicle collision energy threshold. If the dominant peak frequency is within the wind-induced vibration frequency range and the short-time energy value is less than the vehicle collision energy threshold, the current event is determined to be natural vibration caused by environmental wind pressure. If the short-time energy value is greater than the vehicle collision energy threshold and the amplitude decay rate of the reconstructed time-domain vibration waveform is greater than a preset rigid collision decay threshold, the current event is determined to be a vehicle collision with the guardrail.
[0063] Specifically, a Fast Fourier Transform (FFT) is performed on the reconstructed time-domain vibration waveform, with the number of transform points matching the signal length. This converts the time-domain signal to a frequency-domain signal, and the squared magnitude of each frequency component is calculated. This squared magnitude is then divided by the frequency resolution to obtain the frequency-domain power spectral density distribution data. Subsequently, a sliding time window of 50 to 100 milliseconds is set in the time domain. Within this window, the squared amplitude of the reconstructed waveform is integrated or summed to calculate the short-time energy value. Simultaneously, the number of times the waveform crosses the zero-level axis is counted, and this number is divided by the window duration to obtain the zero-crossing rate. In the event discrimination logic, a preset wind-induced vibration frequency range is established. This range is set based on the natural frequency of the guardrail structure and the Karman vortex street principle. For example, if the first-order natural frequency of the guardrail is measured to be 3Hz through on-site modal testing, then the range is set to 2.5Hz to 3.5Hz. A preset vehicle impact energy range is also established. The energy threshold is obtained by conducting controlled pendulum impact experiments on guardrails of the same model, recording the minimum energy value at which plastic deformation occurs, and taking 60% of this value as the threshold. The peak frequency of the dominant frequency obtained from frequency domain analysis is compared with the frequency range of wind-induced vibration. If the dominant frequency falls within this range and the calculated short-time energy value is consistently lower than the vehicle impact energy threshold, it indicates that the vibration is resonance or shaking caused by wind load, and is judged as natural vibration. Conversely, if the short-time energy value is detected to rise sharply at a certain moment and exceed the vehicle impact energy threshold, and the amplitude decay rate of the waveform exceeds the preset rigid collision decay threshold, this decay threshold is set based on the damping characteristics of the metal material, for example, set to decay the peak amplitude by 50% per second, then it is determined that a vehicle collision with the guardrail has occurred, and further identification of whether it is a minor scratch or a frontal collision is made in combination with the spectral characteristics.
[0064] In this embodiment, when it is determined that no impact event has occurred and the reconstructed time-domain vibration waveform exhibits a free decay trend, the specific steps for extracting the decayed tailwave data segment from the reconstructed time-domain vibration waveform are as follows: Envelope detection is performed on the reconstructed time-domain vibration waveform to obtain the waveform amplitude envelope. The waveform amplitude envelope is traversed to find waveform segments that satisfy the monotonically decreasing condition and whose duration exceeds a preset minimum decay period. The start time of the waveform segment is positioned as the moment when the vibration excitation disappears, and the moment when the amplitude of the waveform segment decreases to a preset noise floor level is positioned as the decay end time. Waveform data between the start time and the decay end time is extracted, and the extracted data is bandpass filtered to remove high-frequency noise interference, resulting in a denoised decayed tailwave data segment.
[0065] Specifically, envelope detection processing is performed on the reconstructed time-domain vibration waveform. Hilbert transform is used to calculate the analytic signal, and the modulus of the analytic signal is taken as the waveform amplitude envelope. Alternatively, a peak-holding algorithm is used to connect all local maxima to form the envelope. The generated envelope data is traversed, and a monotonically decreasing criterion is set. For example, the envelope value of 10 consecutive sampling points must show a decreasing trend, or minor fluctuations caused by noise are allowed to be within 5%. Waveform segments that meet this condition and whose duration exceeds a preset minimum decay period are searched. The minimum decay period is set according to the first-order natural period of the guardrail structure, typically 3 to 5 vibration periods. Once a segment that meets the conditions is identified, the time corresponding to the starting point with the largest amplitude in that segment is used. The interval is marked as the moment when the vibration excitation disappears, that is, the instant when the impact or external force is removed. Then, the envelope is traced backward until the amplitude first decreases to the preset noise floor level. This noise floor level is statistically derived based on the average amplitude of the main environmental noise (such as the sound of distant traffic), for example, three times the standard deviation of the sensor readings under static conditions. This point is positioned as the end of the decay. The original reconstructed waveform data between the start time and the end time of decay is extracted. A bandpass digital filter is configured, whose passband range covers the first three modal frequencies of the guardrail structure. For example, the passband is set to 2Hz to 20Hz. The extracted data is sent to the filter for convolution operation to filter out high-frequency electromagnetic interference and low-frequency drift, and the denoised attenuated tailwave data segment is obtained.
[0066] In this embodiment, the specific steps for calculating the fastness integrity assessment index of the guardrail structure based on the attenuated tailwave data segment are as follows: Identify all local maxima points in the attenuated tailwave data segment, construct a set of peak sequences, select the amplitude of the first peak at the initial moment and the amplitude of the final peak after a preset M vibration cycles from the peak sequence set, calculate the logarithmic attenuation rate based on the first peak amplitude and the final peak amplitude, and construct a calculation model for the fastness integrity assessment index by combining it with a preset nonlinear amplitude correction factor. The fastness integrity assessment index is then derived using the calculation model. The formula corresponding to the calculation model is: ,in, This represents the fastening integrity assessment index. This represents the preset normalized weighting coefficients. This indicates the number of vibration periods between the two selected wave peaks. Indicates the amplitude of the first peak. Indicates the final peak amplitude. Let represent the logarithmic function with base e. This represents the baseline logarithmic decay rate of the guardrail under standard factory tightening conditions. This represents the nonlinear amplitude correction factor. This indicates the maximum allowable elastic deformation amplitude of the guardrail material.
[0067] Specifically, all local maxima in the denoised attenuated wake wave data segment are identified by comparing the magnitudes of three adjacent data points. and To determine the peak location, construct a set of peak sequences, and extract the amplitude of the first peak from this set as... and the next number Each peak is used to extract its amplitude as... ,in The value is typically set to 5 to 10 to ensure sufficient attenuation samples are not overwhelmed by noise, and then the formula is used. Calculate the fastening integrity assessment index, where The tightness integrity assessment index is a dimensionless value used to quantify the tightness of the guardrail connection structure. The closer the value is to 1, the better the condition. This represents the preset normalization weighting coefficient, used to balance the influence of the logarithmic decay term and the amplitude term. This coefficient is determined by the analytic hierarchy process (AHP) based on expert scoring, and is usually between 0.6 and 0.8. This represents the number of vibration periods between the two selected wave peaks. It is a positive integer, and its value is derived from the interval selected in the wave peak sequence above. The amplitude of the first peak at the start of the attenuation phase is expressed in volts or gravitational acceleration g and is directly obtained by envelope detection and peak extraction algorithms. This represents the amplitude of the final peak after M cycles, in units of 1 and 2. Consistent, also obtained by the peak extraction algorithm; This represents a logarithmic function with the natural constant e (approximately 2.71828) as its base, used to calculate the logarithmic decay rate; This represents the baseline logarithmic decay rate of the guardrail under factory standard tightening conditions. This value is obtained by manually vibrating and calculating the average logarithmic decay rate after the guardrail is installed and confirmed to be fully tightened. It serves as a reference zero point for the health status and is stored in the device's non-volatile memory. This represents the nonlinear amplitude correction factor, used to correct the nonlinear damping effect under large amplitude. This factor is obtained by fitting the guardrail material through shaking table tests with different amplitudes, and is usually a positive number less than 1, such as 0.15. This indicates the maximum allowable elastic deformation amplitude of the guardrail material, in units of 1 / 2 and 2 / 3. Consistent, this value is calculated by finite element analysis based on the yield strength of the guardrail material and the structural geometric parameters, representing the critical amplitude before the structure enters plastic deformation.
[0068] In this embodiment, the step of comparing the fastening integrity assessment index with a preset safety benchmark threshold to determine the loose state of the guardrail connector is as follows: read the historical monitoring database stored in the cloud server, obtain the historical fastening integrity assessment index sequence of the guardrail node in the past preset time period, calculate the average value of the historical fastening integrity assessment index sequence as a dynamic reference benchmark, set the safety benchmark threshold to 80% of the dynamic reference benchmark, compare the current fastening integrity assessment index with the safety benchmark threshold, if the fastening integrity assessment index is greater than or equal to the safety benchmark threshold, the guardrail connector is determined to be in a fastened state; if the fastening integrity assessment index is less than the safety benchmark threshold, the guardrail connector is determined to be in a loose state, and a corresponding level of maintenance warning signal is generated based on the difference between the fastening integrity assessment index and the safety benchmark threshold.
[0069] Specifically, the system reads the historical tightness integrity assessment index records of the node stored in the cloud server database over the past 30 days or other preset time periods. Outliers caused by sensor malfunctions or extreme weather are removed. The remaining valid historical data are averaged to obtain an average value reflecting the recent structural trend, which serves as a dynamic reference benchmark. To set reasonable alarm limits, this dynamic reference benchmark is multiplied by 0.8 (i.e., 80%) to obtain the current safety benchmark threshold. This proportional coefficient is determined according to the guardrail structure safety redundancy design specifications. The currently calculated... The value is compared with the security baseline threshold. If the current value is... If the value is greater than or equal to the threshold, it indicates that the guardrail damping characteristics have not significantly deteriorated compared to the historical average level, the connectors remain secure, and no alarm is triggered. If the value is less than the threshold, it indicates a significant increase in structural damping or a decrease in stiffness, indicating that the guardrail connectors are loose. At this point, the difference between the two is calculated, and the warning level is determined based on the magnitude of the difference. For example, if the difference is within 10% of the threshold, it is marked as a yellow warning, indicating that it needs to be included in the routine inspection plan. If the difference exceeds 10%, it is marked as a red warning, and the message is directly pushed to the maintenance personnel's handheld terminal, requiring immediate repair.
[0070] In this embodiment, the steps of collecting the original vibration signal of the guardrail by the IoT sensing terminal deployed at the urban guardrail node are as follows: the mechanical vibration of the guardrail is sensed by a piezoelectric accelerometer installed at the connection between the guardrail post and the beam; the charge signal output by the piezoelectric accelerometer is converted into a voltage signal by a charge amplifier; the voltage signal is filtered out at high frequency by an anti-aliasing filter; and the processed voltage signal is discretized by an analog-to-digital converter at a preset Nyquist sampling frequency to generate a time-series discrete digital signal as the original vibration signal. The IoT sensing terminal automatically enters a sleep mode when it detects that the amplitude of the original vibration signal is lower than a preset silence threshold, and triggers projection dimensionality reduction processing only when the amplitude exceeds the silence threshold.
[0071] Specifically, a piezoelectric accelerometer installed at the connection between the guardrail post and the beam utilizes a piezoelectric ceramic plate inside the sensor. When the guardrail structure vibrates mechanically, the piezoelectric effect generates a weak charge signal. This charge signal is input to a charge amplifier circuit, where a high-input-impedance operational amplifier and a feedback capacitor convert the high-resistance charge source into a low-resistance voltage signal. This voltage signal is then fed into an anti-aliasing filter, typically a fourth-order or higher Butterworth low-pass filter. Its cutoff frequency is set to 1.2 to 1.5 times the effective bandwidth of the signal. For example, if the upper limit of the vibration frequency of interest is 200Hz, the cutoff frequency is set to 240Hz to filter out high-frequency noise above the Nyquist frequency and prevent spectral aliasing. The processed analog voltage signal then enters an analog-to-digital converter (ADC). The ADC, triggered by the microcontroller's timer, discretizes the voltage signal at a preset Nyquist sampling frequency (e.g., 1kHz), quantizing the continuous analog voltage into digital code to generate a time-series discrete digital signal stream as the original vibration signal. A low-power monitoring thread runs in the microcontroller software to continuously monitor the absolute value or root mean square value of the ADC output data. When the calculated amplitude is lower than a preset silence threshold, which is set to 1.5 times the ambient background noise amplitude, the microcontroller shuts down all peripheral clocks except for the timer and wake-up pin, and enters deep sleep mode to save power. Only when the monitored vibration amplitude exceeds the silence threshold is the main processor woken up by an external interrupt or level change interrupt, triggering the subsequent compressed sensing projection dimensionality reduction processing program.
[0072] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments that can be applied to other fields. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.
Claims
1. A real-time monitoring method for smart urban guardrails based on the Internet of Things, characterized in that, Includes the following steps: The original vibration signals of the guardrails are collected by IoT sensor terminals deployed at the nodes of the city guardrails. The original vibration signals are then projected and dimension-reduced using a preset compressed sensing observation matrix to obtain a compressed measurement data packet. The compressed measurement data packet is transmitted to a cloud server via a wide area network, and the compressed measurement data packet is solved using a sparse reconstruction algorithm to obtain the reconstructed time-domain vibration waveform. Time-frequency features are extracted from the reconstructed time-domain vibration waveform, and the extracted feature data is used to determine whether a collision event has occurred with the guardrail and to identify the specific type of the collision event. When it is determined that no impact event has occurred and the reconstructed time-domain vibration waveform has a free decay trend, the attenuation tail wave data segment in the reconstructed time-domain vibration waveform is extracted, the fastening integrity assessment index of the guardrail structure is calculated based on the attenuation tail wave data segment, and the fastening integrity assessment index is compared with the preset safety benchmark threshold to determine the loosening state of the guardrail connectors.
2. The real-time monitoring method for smart urban guardrails based on the Internet of Things according to claim 1, characterized in that, The specific steps for projecting and reducing the dimensionality of the original vibration signal using a preset compressed sensing observation matrix to obtain a compressed measurement value data packet are as follows: A random observation matrix following a Bernoulli distribution is pre-set in the microcontroller of the IoT sensing terminal. The number of columns of the random observation matrix is consistent with the number of sampling points of the original vibration signal, and the number of rows of the random observation matrix is much smaller than the number of columns. The original vibration signal is constructed into a digital signal vector in column vector form. The random observation matrix is multiplied by the digital signal vector through matrix multiplication to obtain an observation vector with reduced dimension. The observation vector is quantized and encoded, and timestamp tags and node identification are added according to a preset data packet format to generate the compressed measurement value data packet; The data volume of the compressed measurement value data packet is 20% to 30% of the original vibration signal data volume, and the IoT sensing terminal only needs to perform a wireless transmission operation on the compressed measurement value data packet.
3. The real-time monitoring method for smart urban guardrails based on the Internet of Things according to claim 1, characterized in that, The specific steps for solving the compressed measurement data packet using the sparse reconstruction algorithm to obtain the reconstructed time-domain vibration waveform are as follows: After receiving the compressed measurement data packet, the cloud server parses out the observation vector; Construct or call a pre-stored sensing matrix, which is obtained by multiplying the random observation matrix used by the IoT sensing terminal with a preset sparse transformation basis matrix; Initialize the residual vector to the observation vector, initialize the index set to an empty set, and enter the orthogonal matching pursuit iteration loop; In each iteration, the inner product of the residual vector and all column vectors in the sensing matrix is calculated, the column index corresponding to the inner product with the largest absolute value is found, and the column index is added to the index set. Extract the corresponding submatrix from the sensing matrix according to the index set, solve the projection coefficient of the observation vector on the space spanned by the submatrix using the least squares method, and update the residual vector to the difference between the observation vector and the estimated observation value corresponding to the projection coefficient; Determine whether the norm of the residual vector is less than a preset reconstruction accuracy threshold. If it is less, stop the iteration and use the finally obtained sparse coefficient vector and the sparse transformation basis matrix to recover the reconstructed time-domain vibration waveform.
4. The real-time monitoring method for smart urban guardrails based on the Internet of Things according to claim 1, characterized in that, The steps for extracting time-frequency features from the reconstructed time-domain vibration waveform, and determining whether a collision event has occurred on the guardrail and identifying the specific type of collision event based on the extracted feature data are as follows: Perform a fast Fourier transform on the reconstructed time-domain vibration waveform to obtain frequency-domain power spectral density distribution data; Calculate the short-time energy value and zero-crossing rate value of the reconstructed time-domain vibration waveform within a preset time window; The dominant peak frequency in the frequency domain power spectral density distribution data is matched with a preset wind-induced vibration frequency range, and the short-time energy value is compared with a preset vehicle impact energy threshold. If the peak frequency of the dominant frequency is within the wind-induced vibration frequency range and the short-time energy value is less than the vehicle impact energy threshold, then the current event is determined to be a natural vibration caused by environmental wind pressure. If the short-time energy value is greater than the vehicle impact energy threshold and the amplitude decay rate of the reconstructed time-domain vibration waveform is greater than the preset rigid collision decay threshold, then the current event is determined to be a vehicle impacting the guardrail event.
5. The real-time monitoring method for smart urban guardrails based on the Internet of Things according to claim 1, characterized in that, When it is determined that no impact event has occurred and the reconstructed time-domain vibration waveform exhibits a free decay trend, the specific steps for extracting the attenuated tailwave data segment from the reconstructed time-domain vibration waveform are as follows: Envelope detection is performed on the reconstructed time-domain vibration waveform to obtain the waveform amplitude envelope; Traverse the waveform amplitude envelope to find waveform segments that satisfy the monotonically decreasing condition and whose duration exceeds the preset minimum decay period; The start time of the waveform segment is defined as the time when the vibration excitation disappears, and the time when the amplitude of the waveform segment decreases to a preset noise floor level is defined as the time when the decay ends. The waveform data between the start time and the end time of attenuation is extracted, and the extracted data is bandpass filtered to remove high-frequency noise interference, resulting in the denoised attenuation tailwave data segment.
6. The real-time monitoring method for smart urban guardrails based on the Internet of Things according to claim 1, characterized in that, The specific steps for calculating the fastening integrity assessment index of the guardrail structure based on the attenuated tailwave data segment are as follows: Identify all local maxima in the attenuated tailwave data segment and construct a set of peak sequences; Select the amplitude of the first peak at the initial moment and the amplitude of the final peak after a preset M vibration cycles from the set of peak sequences. The logarithmic decay rate is calculated based on the amplitude of the first peak and the amplitude of the final peak, and a calculation model for the fastening integrity assessment index is constructed by combining it with a preset nonlinear amplitude correction factor. The fastening integrity assessment index is derived using the aforementioned calculation model, and the formula corresponding to the calculation model is as follows: ； in, This represents the fastening integrity assessment index. This represents the preset normalized weighting coefficients. This indicates the number of vibration periods between the two selected wave peaks. This represents the amplitude of the first peak. This indicates the amplitude of the termination peak. Let represent the logarithmic function with base e. This represents the baseline logarithmic decay rate of the guardrail under standard factory tightening conditions. This represents the nonlinear amplitude correction factor. This indicates the maximum allowable elastic deformation amplitude of the guardrail material.
7. The real-time monitoring method for smart urban guardrails based on the Internet of Things according to claim 1, characterized in that, The specific steps for determining the looseness of guardrail connectors by comparing the fastening integrity assessment index with a preset safety benchmark threshold are as follows: Read the historical monitoring database stored in the cloud server to obtain the historical fastening integrity assessment index sequence of the guardrail nodes within a preset time period; The average value of the historical fastening integrity assessment index sequence is calculated as a dynamic reference benchmark, and the safety benchmark threshold is set to 80% of the dynamic reference benchmark. Compare the current fastening integrity assessment index with the safety benchmark threshold. If the fastening integrity assessment index is greater than or equal to the safety benchmark threshold, the guardrail connector is determined to be in a fastened state. If the fastening integrity assessment index is less than the safety benchmark threshold, the guardrail connector is determined to be loose, and a corresponding level of maintenance warning signal is generated based on the difference between the fastening integrity assessment index and the safety benchmark threshold.
8. The real-time monitoring method for smart urban guardrails based on the Internet of Things according to claim 1, characterized in that, The specific steps for collecting the raw vibration signals of the guardrails using IoT sensor terminals deployed at urban guardrail nodes are as follows: The mechanical vibration of the guardrail is sensed by a piezoelectric accelerometer installed at the connection between the guardrail post and the crossbeam. The charge signal output by the piezoelectric accelerometer is converted into a voltage signal by a charge amplifier; The voltage signal is subjected to high-frequency filtering using an anti-aliasing filter. The processed voltage signal is discretized by an analog-to-digital converter at a preset Nyquist sampling frequency to generate a time-series discrete digital signal, which serves as the original vibration signal. The IoT sensing terminal automatically enters sleep mode when it detects that the amplitude of the original vibration signal is lower than a preset silence threshold, and triggers the projection dimensionality reduction process only when the amplitude exceeds the silence threshold.