Pressure sensing type asphalt cold patch construction equipment control method and system based on internet of things

Abstract

The application relates to the technical field of road construction equipment control, and discloses a pressure sensing type asphalt cold patch construction equipment control method and system based on the Internet of Things. The method comprises the following steps: acquiring a counterforce flow collected by a sensor at a roller of a road roller in a cold patch rolling area, and performing windowing pretreatment on the counterforce flow to obtain a counterforce sequence; performing compaction response characterization processing on the counterforce sequence to obtain a counterforce feature; performing pore state estimation and change characterization processing on the counterforce feature in a continuous window to obtain a pore change amount; generating a compaction stage according to the pore change amount, and performing vibration parameter mapping processing in combination with an energy coefficient and an equipment state to obtain a target frequency and a target amplitude; and generating a roller vibration instruction according to the target frequency and the target amplitude. The method can realize continuous linkage between cold patch compaction state identification and construction equipment vibration parameter control.

Description

Technical Field

[0001] This invention relates to the field of road construction equipment control technology, and in particular to a control method and system for pressure-sensitive asphalt cold patching construction equipment based on the Internet of Things. Background Technology

[0002] Currently, cold asphalt patch material is widely used in road repair construction. Construction equipment typically uses a roller to repeatedly compact the paved cold patch material, gradually transforming loose particles into a stable, dense structure. During compaction, the cold patch material undergoes continuous changes, including aggregate rearrangement, pore filling, and surface sealing. The requirements for vibration frequency and amplitude differ at each stage. With the development of IoT data acquisition technology, construction equipment can continuously collect force response data between the rollers and the material using smart sensors, providing a data foundation for identifying the material's compaction state.

[0003] In existing technologies, road rollers typically use fixed vibration frequencies and amplitudes for compaction control, or operators adjust the equipment vibration parameters based on their experience. While this method can achieve basic compaction, it's difficult to continuously adjust the equipment vibration state according to the stress response changes of the cold-mixed material at different compaction stages. When the cold-mixed material is in its initial loose stage, fixed vibration parameters may not adequately promote aggregate rearrangement; when the cold-mixed material approaches compaction, continuous strong vibration may disturb the already formed dense structure, resulting in a lack of stable correlation between the compaction state and the equipment output.

[0004] Existing technologies have the problem of difficulty in adaptively controlling the vibration parameters of construction equipment based on the real-time compaction status of cold-mixed materials. Summary of the Invention

[0005] This invention provides a control method and system for pressure-sensitive asphalt cold patching construction equipment based on the Internet of Things, in order to solve the problem in the prior art that it is difficult to adaptively control the vibration parameters of construction equipment according to the real-time compaction status of cold patching.

[0006] Firstly, to address the aforementioned technical problems, this invention provides a control method for pressure-sensitive asphalt cold patching construction equipment based on the Internet of Things, comprising: The reaction flow collected by the sensor at the roller of the road roller in the cold patch compaction area is obtained, and the reaction flow is preprocessed by windowing to obtain the reaction sequence; The reaction force sequence is subjected to compaction response characterization processing to obtain reaction force characteristics; The reaction force features within a continuous window are input into a pre-trained porosity mapping model to predict porosity, resulting in a pore sequence. The change trend of the pore sequence is then fitted to obtain the amount of pore change. Based on the porosity change, the interval assignment is determined to obtain the compaction stage, and the energy coefficient is calculated based on the reaction force characteristics according to the compaction stage to obtain the energy coefficient. The equipment status of the road roller during construction is obtained, and vibration parameter mapping is performed based on the compaction stage, the energy coefficient, and the equipment status to obtain the target frequency and target amplitude. Drive commands are converted based on the target frequency and the target amplitude to generate roller vibration commands.

[0007] Secondly, the present invention provides a control system for a pressure-sensitive asphalt cold patching construction equipment based on the Internet of Things, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the method described above.

[0008] Compared with the prior art, the present invention has the following beneficial effects: (1) This invention acquires the reaction flow collected by sensors at the rollers of the road roller within the compaction zone of the cold patch material, and performs windowing preprocessing and compaction response characterization processing on the reaction flow, transforming the continuous force change between the rollers and the cold patch material during compaction into reaction characteristics that can be used for state analysis. This method avoids relying solely on fixed vibration parameters or human experience to judge the compaction state of the cold patch material, enabling construction equipment to obtain a data basis that reflects the changes in the material's force response.

[0009] (2) This invention predicts porosity by inputting the reaction force features within a continuous window into a pre-trained porosity mapping model, obtaining a pore sequence. The change trend of the pore sequence is then fitted to obtain the pore change amount. Finally, the compaction stage is determined by classifying the pore change amount into intervals. This allows for the continuous differentiation of aggregate rearrangement, pore filling, and sealing compaction states during the rolling process of cold-mixed materials. This method establishes a correspondence between reaction force changes and the material pore evolution state, solving the problem of delayed or discontinuous identification of compaction stages.

[0010] (3) This invention calculates the energy coefficient based on the energy coefficient of the reaction force characteristics during the compaction stage, and performs vibration parameter mapping processing in conjunction with the equipment status during the construction process of the road roller to obtain the target frequency and target amplitude. Then, based on the target frequency and target amplitude, the drive command is converted to generate the roller vibration command, so that the exciter speed and eccentric block angle can be adjusted according to the changes in the compaction state of the cold-filled material. This method forms a closed-loop processing link from reaction force acquisition, compaction stage identification to roller vibration parameter control, reducing the adverse effects of insufficient vibration in the loose stage and continuous strong vibration in the compaction stage on the construction quality. Attached Figure Description

[0011] Figure 1This is a schematic diagram of the control method for a pressure-sensitive asphalt cold patching construction equipment based on the Internet of Things, provided in the first embodiment of the present invention. Detailed Implementation

[0012] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0013] Reference Figure 1 The first embodiment of the present invention provides a control method for pressure-sensitive asphalt cold patching construction equipment based on the Internet of Things, including the following steps: S1, acquire the reaction flow collected by the sensor at the roller of the road roller in the cold patch material compaction area, and perform windowing preprocessing on the reaction flow to obtain the reaction sequence; S2, Perform compaction response characterization processing on the reaction force sequence to obtain reaction force characteristics; S3, input the reaction force features within the continuous window into the pre-trained porosity mapping model to predict porosity, obtain a pore sequence, and fit the change trend according to the pore sequence to obtain the pore change amount; S4, determine the interval assignment based on the porosity change to obtain the compaction stage, and calculate the energy coefficient of the reaction force characteristic based on the compaction stage to obtain the energy coefficient; S5, acquire the equipment status of the road roller during construction, and perform vibration parameter mapping processing based on the compaction stage, the energy coefficient and the equipment status to obtain the target frequency and target amplitude; S6, drive command conversion is performed according to the target frequency and the target amplitude to generate roller vibration command.

[0014] In step S1, the reaction flow collected by the sensor at the roller of the road roller in the cold patch compaction area is obtained, and the reaction flow is preprocessed by windowing to obtain the reaction sequence.

[0015] The reaction flow is preprocessed by windowing to obtain a reaction sequence, including: The reaction flow is segmented according to a preset time window to obtain reaction segments; The reaction force segment is subjected to median filtering to obtain a smooth segment; The smoothed segment is subjected to Butterworth low-pass filtering to obtain the reaction force sequence.

[0016] In one implementation, the reaction flow is a time-series data set continuously collected by a sensor at the roller of the road roller within the compaction zone of the cold-mixed material. Each data point in the reaction flow includes a reaction force value and the acquisition time. The sensor is a piezoelectric sensor installed on the roller support structure or at the force transmission position inside the roller. During acquisition, the piezoelectric sensor converts the reaction force generated by the contact between the roller and the cold-mixed material into an electric charge signal. The signal conditioning circuit performs voltage conversion and analog-to-digital conversion on the electric charge signal to obtain the reaction flow arranged according to the acquisition time.

[0017] It should be noted that the sampling frequency of the reaction flow is set to 1000 Hz. This sampling frequency is determined by the vibration calibration data of the road roller. Specifically, the highest excitation frequency of the road roller under low-speed, medium-speed, and high-speed vibration states is read, and the highest excitation frequency is found to be 60 Hz. The lower limit of the sampling frequency is selected as 600 Hz according to the sampling condition of more than ten times the highest excitation frequency. Then, from the three sampling levels of 500 Hz, 1000 Hz, and 2000 Hz supported by the sensor acquisition module, 1000 Hz, which is not lower than 600 Hz and has low data overhead, is selected as the sampling frequency of the reaction flow.

[0018] This embodiment segments the reaction flow according to a preset time window. The preset time window is set to 500 milliseconds, which is determined through offline compaction calibration. Specifically, reaction segments are extracted on the same type of cold-mixed material sample using window lengths of 250 milliseconds, 500 milliseconds, and 750 milliseconds, respectively, and the coefficient of variation of the root mean square value of the reaction segment is calculated for each window length. When the window length is 250 milliseconds, the coefficient of variation is 4.6%; when the window length is 500 milliseconds, the coefficient of variation is 1.8%; and when the window length is 750 milliseconds, the coefficient of variation is 1.6%. Since 500 milliseconds reduces the coefficient of variation by 2.8 percentage points compared to 250 milliseconds, and 750 milliseconds only reduces it by 0.2 percentage points compared to 500 milliseconds, 500 milliseconds is determined as the preset time window. During segmentation, the reaction flow is read in ascending order of acquisition time, and every 500 consecutive data points are divided into a reaction segment. The start and end acquisition times of each reaction segment are written into the segment index.

[0019] It should be noted that the median filtering process uses a sliding window with a length of 5 sampling points. This length is determined by collecting data from an unloaded roller. Specifically, 10 sets of reaction flow data are collected with the roller away from the surface of the cold-mixed material. Sudden spikes are processed using sliding windows with 3, 5, and 7 sampling points, respectively. When the sliding window length is 3 sampling points, some single-point spikes are still retained. When the sliding window length is 5 sampling points, single-point spikes are replaced with the neighborhood median. When the sliding window length is 7 sampling points, the rising edge of the reaction force shows obvious passivation. Therefore, 5 sampling points are determined as the sliding window length for median filtering. During median filtering, two data points before and after the current data point within each reaction segment are read. The 5 data points are arranged in ascending order of reaction force value, and the value of the middle data point is used to replace the value of the current data point. When the current data point is located at the boundary of the reaction segment, the sliding window is repeatedly filled with boundary data points to obtain a smooth segment.

[0020] It is worth noting that the Butterworth low-pass filtering process uses a fourth-order Butterworth low-pass filter with a cutoff frequency set at 80 Hz. This cutoff frequency was determined through excitation frequency sweep calibration. Specifically, the roller was controlled to sweep frequencies within the excitation frequency range of 20 Hz to 60 Hz, and the reaction force flow at the roller was collected and its spectrum was statistically analyzed. This revealed that the main reaction force energy was concentrated below 80 Hz. After setting 80 Hz as the cutoff frequency, the amplitude of the main frequency after filtering decreased by less than 3% compared to the amplitude before filtering, and the high-frequency noise energy above 80 Hz decreased by more than 70%. During the Butterworth low-pass filtering process, fourth-order low-pass filter coefficients were first generated according to the 1000 Hz sampling frequency and the 80 Hz cutoff frequency. Then, recursive filtering was performed on the smoothed segment according to the acquisition time sequence to obtain the reaction force sequence after removing high-frequency jitter. The reaction force sequence retains the acquisition time of each data point and the filtered reaction force value for subsequent compaction response characterization processing.

[0021] For example, in a cold-compaction rolling operation, a piezoelectric sensor continuously collected data for 5 seconds at a sampling frequency of 1000 Hz, obtaining 5000 reaction force data points. In this embodiment, the reaction force flow is divided into 10 reaction force segments according to a preset time window of 500 milliseconds, with each reaction force segment containing 500 data points. A single-point mutation value of 18000 N exists within a certain reaction force segment, while the other four data points in the neighborhood are 8200 N, 8350 N, 8400 N, and 8500 N, respectively. After median filtering with a length of 5 sampling points, the single-point mutation value is replaced with 8400 N. Subsequently, a fourth-order Butterworth low-pass filter is applied to the smoothed segments to obtain the reaction force sequence.

[0022] In step S2, the reaction force sequence is subjected to compaction response characterization processing to obtain reaction force characteristics, including: The root mean square value and peak factor are calculated based on the reaction force sequence to obtain the time-domain characteristics; Perform a fast Fourier transform on the reaction force sequence to obtain a spectral sequence; Based on the spectral sequence, the dominant frequency component and the frequency band energy ratio are extracted to obtain the frequency domain features; The time-domain features and the frequency-domain features are standardized to obtain the reaction force features.

[0023] In one implementation, the time-domain feature is the statistical representation of the reaction force sequence over time. The root mean square (RMS) value reflects the overall intensity of the reaction force values ​​within the current time window, and the peak factor reflects the ratio of the instantaneous impact peak value to the overall intensity within the current time window. To calculate the RMS value, each reaction force value in the reaction force sequence is squared, the average of all squared results is calculated, and then the square root of the average is taken to obtain the RMS value. To calculate the peak factor, the reaction force value with the largest absolute value in the reaction force sequence is taken as the peak value, and the peak value is divided by the RMS value to obtain the peak factor; when the RMS value is 0, the peak factor is recorded as 0. The RMS value and the peak factor are written into the same feature record to obtain the time-domain feature.

[0024] It should be noted that the spectral sequence is a set of frequency amplitudes formed by processing the reaction force sequence through Fast Fourier Transform (FFT). Before performing the FFT, the average reaction force value of the reaction force sequence is first read, and the average reaction force value is subtracted from each reaction force value in the reaction force sequence to obtain a mean-reduced sequence. Then, the mean-reduced sequence is processed by FFT to convert the reaction force values ​​arranged by acquisition time into complex spectral results arranged by frequency. The modulus of the complex spectral result corresponding to each frequency point is then read to obtain the spectral sequence. Since the preset time window is 500 milliseconds and the sampling frequency is 1000 Hz, each reaction force sequence contains 500 data points, and the interval between adjacent frequency points in the spectral sequence is 2 Hz.

[0025] This embodiment extracts the dominant frequency component and frequency band energy proportion based on the spectral sequence. The dominant frequency component is the effective frequency point with the largest amplitude in the spectral sequence. The effective frequency range is set to 10 Hz to 100 Hz, which is determined jointly by no-load vibration calibration and contact rolling calibration. Specifically, the exciter's base frequency range is read under no-load vibration conditions, and the reaction force spectral energy distribution is statistically analyzed under contact rolling conditions. Frequency points below 10 Hz mainly correspond to slow changes in roller attitude, while frequency points above 100 Hz mainly correspond to high-frequency vibrations of the structure. Therefore, 10 Hz to 100 Hz is determined as the effective frequency range. When extracting the dominant frequency component, all frequency points within the effective frequency range are traversed, and the frequency point with the largest amplitude is selected. The frequency value of this frequency point is determined as the dominant frequency component.

[0026] It is worth noting that the construction response frequency band in the frequency band energy proportion is set to 20 Hz to 80 Hz. This construction response frequency band is determined through reaction force sequence spectrum samples. Specifically, spectral statistics are performed on 50 reaction force sequences at different compaction stages, and the frequency points corresponding to the cumulative energy reaching 5% and 95% in each spectrum sequence are read. The low-end frequency points of all samples are rounded down to multiples of 10 Hz to obtain 20 Hz; the high-end frequency points of all samples are rounded up to multiples of 10 Hz to obtain 80 Hz. When calculating the frequency band energy proportion, the squares of the amplitudes of each frequency point in the 20 Hz to 80 Hz range are first summed to obtain the construction response frequency band energy; then the squares of the amplitudes of each frequency point in the 10 Hz to 100 Hz range are summed to obtain the effective total energy; the construction response frequency band energy is divided by the effective total energy to obtain the frequency band energy proportion. When the effective total energy is 0, the frequency band energy proportion is recorded as 0. The main frequency component and the frequency band energy ratio are written into the same feature record to obtain the frequency domain feature.

[0027] It should be noted that the standardization process uses standardized parameters formed from offline calibration samples. These standardized parameters include the root mean square (RMS) value, standard deviation of the RMS value, mean peak factor, standard deviation of the peak factor, mean of the dominant frequency component, standard deviation of the dominant frequency component, mean of the frequency band energy percentage, and standard deviation of the frequency band energy percentage. The formation process of these standardized parameters involves collecting no fewer than 120 reaction force sequence samples under the same type of road roller and the same type of cold-mixed material. The time-domain and frequency-domain characteristics of each sample are calculated using the method described above. The average of all sample values ​​for the same characteristic is calculated to obtain the mean of that characteristic. The dispersion of each sample value relative to the mean of that characteristic is then statistically analyzed to obtain the standard deviation of that characteristic. During standardization, the current value of each characteristic is subtracted from the corresponding characteristic mean, and then divided by the corresponding characteristic standard deviation to obtain the standardized characteristic value. When the standard deviation of a characteristic is 0, the standardized value of that characteristic is recorded as 0. The standardized RMS value, peak factor, dominant frequency component, and frequency band energy percentage are arranged in a fixed order to obtain the reaction force characteristics.

[0028] For example, a reaction force sequence contains 500 reaction force values. The calculated root mean square value is 8500 N, and the peak factor is 3.2. After mean-removal processing, a Fast Fourier Transform is performed on the reaction force sequence to obtain a spectral sequence. Within the range of 10 Hz to 100 Hz, the frequency point with the largest amplitude is 36 Hz, which is determined as the dominant frequency component. The construction response frequency band energy within the range of 20 Hz to 80 Hz is calculated, and the effective total energy within the range of 10 Hz to 100 Hz is calculated, yielding a frequency band energy proportion of 0.74. Subsequently, the standardized parameters formed by the offline calibration samples are read, and 8500 N, 3.2, 36 Hz, and 0.74 are standardized respectively, and arranged in the order of root mean square value, peak factor, dominant frequency component, and frequency band energy proportion to obtain the reaction force characteristics.

[0029] In step S3, the reaction force features within the continuous window are input into a pre-trained porosity mapping model to predict porosity, thereby obtaining a pore sequence. The change trend is then fitted based on the pore sequence to obtain the pore change amount.

[0030] Specifically, the reaction force features within a continuous window are input into a pre-trained porosity mapping model for porosity prediction, resulting in a pore sequence, including: The reaction force features are input into a pre-trained porosity mapping model for porosity prediction to obtain porosity values. The porosity values ​​corresponding to the continuous time windows are arranged in a time sequence to obtain the porosity sequence.

[0031] In one implementation, the porosity mapping model is a regression mapping model between reaction force characteristics and the porosity of the cold-fill material. The porosity mapping model employs a random forest regression model, with the reaction force characteristics as input and the porosity value as output. The reaction force characteristics include the standardized root mean square value, peak factor, dominant frequency component, and frequency band energy percentage. The porosity value is the proportion of the void volume within the cold-fill material to the total volume of the cold-fill material within the corresponding time window.

[0032] It should be noted that the porosity mapping model is obtained through offline calibration samples. The offline calibration samples are formed from the same type of cold patch material under different compaction states. When constructing the offline calibration samples, a 50 mm thick layer of cold patch material is first laid on the test bench, and three types of samples—loose, intermediate, and dense—are formed according to different compaction passes. After each compaction pass, the reaction flow generated by the sensor at the roller is collected, and the reaction force characteristics are obtained according to steps S1 and S2. Subsequently, the porosity of the current cold patch material sample layer is measured. The measurement process includes weighing the cold patch material sample layer, measuring its volume, and reading the theoretical maximum density of the cold patch material. The bulk density is calculated based on the sample layer's mass and volume, and the porosity label is calculated based on the difference between the bulk density and the material's theoretical maximum density. Each set of reaction force characteristics and the porosity label corresponding to the same compaction moment constitutes a training sample.

[0033] It is worth noting that the offline calibration sample size was set at 120 groups, determined through an incremental sample size experiment. Specifically, the random forest regression model was trained with 60, 90, 120, and 150 groups of samples, respectively, and the mean absolute error of porosity on the validation set was calculated. When the sample size increased from 60 to 90 groups, the mean absolute error of porosity decreased from 1.42 percentage points to 0.96 percentage points; when the sample size increased from 90 to 120 groups, the mean absolute error of porosity decreased from 0.96 percentage points to 0.63 percentage points; and when the sample size increased from 120 to 150 groups, the mean absolute error of porosity decreased from 0.63 percentage points to 0.60 percentage points. Since the error reduction of 150 groups compared to 120 groups was less than 0.05 percentage points, 120 groups were determined as the offline calibration sample size.

[0034] The random forest regression model comprises 100 regression trees, each with a maximum depth of 8 and a minimum number of samples per leaf node of 5. These model parameters were determined through five-fold cross-validation. During cross-validation, the offline calibration samples were divided into five subsets, with one subset used as the validation set and the remaining four subsets as the training set. Candidate values ​​for the number of regression trees were set to 50, 100, and 150; the maximum depth was set to 6, 8, and 10; and the minimum number of samples per leaf node was set to 3, 5, and 7. All candidate combinations were trained and validated separately. The candidate combination with the smallest mean absolute error of porosity in the five-fold cross-validation was selected, resulting in the model parameters for 100 regression trees, a maximum depth of 8, and a minimum number of samples per leaf node of 5.

[0035] In this embodiment, the reaction force characteristics of the current time window are input into the porosity mapping model during porosity prediction. Each regression tree in the porosity mapping model is divided into nodes according to each feature item in the reaction force characteristics, and a porosity prediction sub-value is output at the leaf node. The 100 porosity prediction sub-values ​​output by the 100 regression trees are read, and the average of these 100 sub-values ​​is calculated to obtain the porosity value corresponding to the current time window. Each porosity value retains its corresponding window number, window start time, and window end time.

[0036] This embodiment performs time-series arrangement processing based on the porosity values ​​corresponding to continuous time windows. During time-series arrangement, the start time of each window corresponding to each porosity value is read, and the porosity values ​​are arranged in ascending order of their start times to obtain a porosity sequence. Each sequence element in the porosity sequence includes a window number, a window start time, and a porosity value; the porosity sequence serves as input for subsequent trend fitting.

[0037] For example, porosity values ​​of 14.8%, 13.9%, 13.1%, 12.6%, and 12.3% were obtained for five consecutive time windows. In this embodiment, the porosity values ​​are arranged in ascending order according to the start time of the five time windows to obtain a porosity sequence. The first sequence element in the porosity sequence records the porosity value of 14.8% for the first time window, the second sequence element records the porosity value of 13.9% for the second time window, and the remaining sequence elements are written in the same way.

[0038] The porosity change is obtained by fitting the change trend based on the pore sequence, including: The porosity change value is calculated based on the porosity value corresponding to adjacent time windows to obtain the change sequence; The trend of change is obtained by performing least-squares fitting on the change sequence. The slope of change is extracted based on the changing trend to obtain the amount of porosity change.

[0039] In one implementation, the change sequence is a sequence formed by arranging the porosity change rates between adjacent time windows in the pore sequence in chronological order. The porosity change rate is a first-order quantity, measured in percentage points per second. When calculating the porosity change rate, this embodiment reads the porosity values ​​and the start time of two adjacent time windows in the pore sequence. The porosity value of the preceding time window is subtracted from the porosity value of the following time window to obtain the porosity difference between adjacent windows. Then, the porosity difference is divided by the time interval between the start times of the two windows to obtain the porosity change rate. Since the preset time window in step S1 is 500 milliseconds, the time interval between adjacent time windows is also 500 milliseconds. Arranging the porosity change rates obtained from consecutive adjacent windows in chronological order yields the change sequence.

[0040] It should be noted that the porosity change is defined as the fitting slope obtained by fitting the porosity sequence through adjacent differences and least squares. It is a trend of the porosity change rate, and the unit is percentage points per second per window number. This quantity is different from the porosity change rate directly calculated between two adjacent windows. The latter is in percentage points per second and is used to construct the change sequence. The former is used to determine the interval assignment in the compaction stage.

[0041] It should be noted that the variation sequence is fitted using the least squares method based on the porosity change rate of the most recent eight adjacent windows. These eight adjacent windows were determined through offline compaction calibration, specifically by constructing variation sequences using 4, 8, and 12 adjacent windows respectively. The standard deviation of porosity change under the same stable compaction state and the response delay to compaction stage switching were statistically analyzed. When using 4 adjacent windows, the standard deviation of porosity change was 0.92 percentage points per second per window number; when using 8 adjacent windows, the standard deviation was 0.38 percentage points per second per window number; and when using 12 adjacent windows, the standard deviation was 0.34 percentage points per second per window number, but the response delay to compaction stage switching increased from 1.0 second to 2.5 seconds. Since 8 adjacent windows significantly reduced fluctuations compared to 4 adjacent windows, and 12 adjacent windows only slightly reduced fluctuations and increased the response delay compared to 8 adjacent windows, the most recent eight adjacent windows were determined as the data length for the least squares fitting.

[0042] This embodiment performs least squares fitting based on the change sequence. During least squares fitting, the window number corresponding to each porosity change rate in the change sequence is used as the independent variable, and the porosity change rate is used as the dependent variable. A straight line is constructed to minimize the sum of squares of the longitudinal deviations of each porosity change rate from the straight line in the change sequence. This straight line represents the change trend. The solution process for the least squares fitting is as follows: first, calculate the average value of all window numbers and the average value of all porosity change rates; then calculate the deviation of each window number relative to the average value of the window numbers, and the deviation of each porosity change rate relative to the average value of the porosity change rates; sum the products of the two types of deviations and divide by the sum of squares of the window number deviations to obtain the slope of the change trend; then calculate the intercept of the straight line based on the slope and the two average values ​​to obtain the complete change trend.

[0043] It is worth noting that the porosity change is the slope of the change trend. When the porosity change is negative, it indicates that the rate of porosity change gradually decreases over time; when the porosity change is close to zero, it indicates that the rate of porosity change tends to stabilize. Subsequent compaction stage mapping processing directly reads the porosity change and determines the compaction stage based on it.

[0044] For example, the porosity values ​​corresponding to nine consecutive time windows in a certain pore sequence are 14.8%, 13.9%, 13.1%, 12.6%, 12.3%, 12.1%, 12.0%, 11.95%, and 11.92%, respectively. The time interval between adjacent time windows is 500 milliseconds. In this embodiment, the porosity change rate between adjacent time windows is calculated sequentially to obtain eight porosity change rates, which are then arranged into a change sequence in chronological order. Subsequently, the change sequence is fitted with least squares to obtain the change trend. If the slope of the change trend is -0.31 percentage points per second per window number, then -0.31 percentage points per second per window number is determined as the porosity change.

[0045] In step S4, the interval assignment is determined based on the porosity change to obtain the compaction stage, and the energy coefficient is calculated based on the reaction force characteristics according to the compaction stage to obtain the energy coefficient.

[0046] The compaction stage is determined by classifying the intervals based on the porosity changes, including: The porosity change is compared with a preset stage interval. When the change in porosity corresponds to the initial compaction range, the compaction stage is determined to be the aggregate rearrangement stage; When the porosity change corresponds to the intermediate compaction interval, the compaction stage is determined to be the pore filling stage; When the change in porosity corresponds to the final compaction interval, the compaction stage is determined to be the sealing compaction stage.

[0047] In one implementation, the compaction stage is a material state marker corresponding to the cold patch material within the current time window. The compaction stage includes an aggregate rearrangement stage, a pore-filling stage, and a sealing compaction stage. The aggregate rearrangement stage corresponds to a state where the position of coarse particles inside the cold patch material is still significantly adjusted; the pore-filling stage corresponds to a state where fine particles enter the gaps between coarse particles, causing a continuous decrease in porosity; and the sealing compaction stage corresponds to a state where the surface gradually stabilizes and the change in porosity slows down. The porosity change is output from step S3 and directly enters the interval assignment judgment.

[0048] It should be noted that the preset stage intervals include the initial compaction interval, the intermediate compaction interval, and the final compaction interval. These preset stage intervals are constructed using offline rolling calibration data. During construction, 120 sets of cold-mixed material rolling samples were collected. For each set of cold-mixed material rolling samples, a porosity change was obtained according to steps S1 to S3. Subsequently, the k-means clustering algorithm was used to perform three-class clustering on the 120 porosity changes. The k-means clustering algorithm first randomly selects three porosity changes as initial cluster centers, then assigns each porosity change to the nearest cluster center. Next, the average value of all porosity changes in each cluster is recalculated as the new cluster center. This assignment and update process is repeated until the difference between two consecutive cluster center values ​​is less than 0.005 percentage points per second per window number. The 0.005 percentage point per second per window number was determined through repeated clustering experiments. Specifically, 0.01, 0.005, and 0.002 were used as stopping differences for clustering. Compared to 0.01, 0.005 reduced the cluster boundary fluctuation by more than 40%, and 0.002 reduced the cluster boundary fluctuation by less than 5% compared to 0.005. Therefore, 0.005 was selected as the stopping difference.

[0049] In this embodiment, the three cluster centers obtained from clustering are arranged in ascending order of their numerical values. The three cluster centers obtained from the offline compaction calibration data are -0.36 percentage points per second per window, -0.14 percentage points per second per window, and -0.03 percentage points per second per window, respectively. The midpoint of adjacent cluster centers is read as the interval boundary, resulting in -0.25 percentage points per second per window and -0.09 percentage points per second per window. Therefore, the initial compaction interval is set as a porosity change less than or equal to -0.25 percentage points per second per window; the intermediate compaction interval is set as a porosity change greater than -0.25 percentage points per second per window and less than or equal to -0.09 percentage points per second per window; and the final compaction interval is set as a porosity change greater than -0.09 percentage points per second per window.

[0050] During the comparison process, this embodiment reads the current porosity change and sequentially determines its classification against the initial compaction interval, the intermediate compaction interval, and the final compaction interval. If the current porosity change falls within the initial compaction interval, the compaction stage is recorded as the aggregate rearrangement stage; if it falls within the intermediate compaction interval, it is recorded as the pore filling stage; and if it falls within the final compaction interval, it is recorded as the sealing compaction stage. The compaction stage and the current time window number are written together into the stage record for subsequent energy coefficient calculation.

[0051] For example, the porosity change corresponding to a certain current time window is -0.31 percentage points per second per window number. Since -0.31 percentage points per second per window number is less than -0.25 percentage points per second per window number, the porosity change falls within the initial compaction range. In this embodiment, the compaction stage is defined as the aggregate rearrangement stage. As another example, the porosity change corresponding to another current time window is -0.06 percentage points per second per window number. Since -0.06 percentage points per second per window number is greater than -0.09 percentage points per second per window number, the porosity change falls within the final compaction range. In this embodiment, the compaction stage is defined as the sealing compaction stage.

[0052] The energy coefficient is obtained by calculating the energy coefficient of the reaction force characteristics based on the compaction stage, including: The target frequency band is determined based on the compaction stage; Based on the reaction force characteristics, extract the frequency band energy and total frequency band energy corresponding to the target frequency band; The energy coefficient is obtained by calculating the ratio between the energy of the specified frequency band and the total energy of the frequency band.

[0053] In one implementation, the target frequency band is the frequency range used in the energy coefficient calculation during the current compaction stage. The target frequency band includes the 20 Hz to 40 Hz band corresponding to the aggregate rearrangement stage, the 40 Hz to 60 Hz band corresponding to the pore filling stage, and the 60 Hz to 80 Hz band corresponding to the sealing compaction stage. The reaction force feature is a window feature record output in step S2. The window feature record includes the standardized root mean square value, peak factor, dominant frequency component, and frequency band energy percentage, and also includes a spectral sequence index retained when generating the dominant frequency component and the frequency band energy percentage. The spectral sequence index records the frequency point and the corresponding squared amplitude value.

[0054] It should be noted that the target frequency band is constructed using offline compaction calibration samples. During construction, 120 sets of cold-mix compaction samples, after stage marking, are read, and the spectral sequence index of each set is read. The effective frequency range of 10 Hz to 100 Hz is divided into 9 continuous frequency bands, each with a width of 10 Hz. For all samples under the same compaction stage, the proportion of the sum of the squared amplitude values ​​within each continuous frequency band to the sum of the squared amplitude values ​​across all samples within the effective frequency range is calculated, and then the median proportion of the same continuous frequency band in all samples of that compaction stage is calculated. Subsequently, the two adjacent continuous frequency bands with the largest median proportion are merged into the target frequency band corresponding to that compaction stage. Through the above construction process, the aggregate rearrangement stage corresponds to the 20 Hz to 40 Hz frequency band, the pore filling stage corresponds to the 40 Hz to 60 Hz frequency band, and the sealing compaction stage corresponds to the 60 Hz to 80 Hz frequency band.

[0055] In this embodiment, the target frequency band is determined according to the compaction stage. If the compaction stage is the aggregate rearrangement stage, the 20 Hz to 40 Hz frequency band is determined as the target frequency band; if the compaction stage is the pore filling stage, the 40 Hz to 60 Hz frequency band is determined as the target frequency band; if the compaction stage is the sealing compaction stage, the 60 Hz to 80 Hz frequency band is determined as the target frequency band.

[0056] This embodiment extracts the frequency band energy and total frequency band energy corresponding to the target frequency band based on the reaction force characteristics. During extraction, the spectral sequence index in the reaction force characteristics is read, and the squared amplitude values ​​of all frequencies falling within the target frequency band are located. These squared amplitude values ​​are then summed to obtain the frequency band energy. Subsequently, the squared amplitude values ​​of all frequencies falling within the effective frequency range of 10 Hz to 100 Hz are located, and these squared amplitude values ​​are summed to obtain the total frequency band energy. The ratio between the frequency band energy and the total frequency band energy is calculated to obtain the energy coefficient. When the total frequency band energy is 0, the energy coefficient is recorded as 0, and the current time window is marked as a low response window.

[0057] It is worth noting that the energy coefficient represents the proportion of energy in the target frequency band corresponding to the current compaction stage within the effective frequency range. This energy coefficient, along with the compaction stage, is written into the state record. Subsequent vibration parameter mapping processing reads the compaction stage and energy coefficient from the state record.

[0058] For example, the current compaction stage is the pore filling stage, and in this embodiment, the 40 Hz to 60 Hz frequency band is determined as the target frequency band. After reading the spectral sequence index in the current reaction force characteristics, the sum of the squared amplitude values ​​of all frequency points in the 40 Hz to 60 Hz range is calculated to obtain a frequency band energy of 4600; the sum of the squared amplitude values ​​of all frequency points in the 10 Hz to 100 Hz range is calculated to obtain a total frequency band energy of 10000. Dividing 4600 by 10000 yields an energy coefficient of 0.46, and 0.46 is written into the status record corresponding to the current time window.

[0059] In step S5, the equipment status of the road roller during construction is obtained, and vibration parameter mapping is performed based on the compaction stage, the energy coefficient, and the equipment status to obtain the target frequency and target amplitude.

[0060] The process of mapping vibration parameters based on the compaction stage, the energy coefficient, and the equipment status to obtain the target frequency and target amplitude includes: The compaction stage, the energy coefficient, and the equipment status are input into a pre-trained support vector regression model for parameter prediction processing to obtain the prediction frequency and prediction amplitude. Based on the vibration requirements corresponding to the compaction stage, the predicted frequency and the predicted amplitude are subjected to boundary correction processing to obtain the target frequency and the target amplitude.

[0061] In one implementation, the equipment status is a set of operating parameters during the roller's construction process. The equipment status includes travel speed, vibratory motor current, and material surface temperature. The travel speed is collected by the roller's speed sensor within the current time window, the vibratory motor current is collected by a current acquisition unit within the current time window, and the material surface temperature is collected by an infrared temperature sensor within the current time window. After collection, this embodiment reads the median travel speed, median vibratory motor current, and median material surface temperature within the same time window, and writes these three values ​​into the equipment status.

[0062] It should be noted that the pre-trained support vector regression model adopts a radial basis function (RBF) kernel support vector regression structure. The model input vector consists of the compaction stage encoding value, energy coefficient, travel speed, vibratory motor current, and material surface temperature. The model output is the predicted frequency and predicted amplitude. The compaction stage encoding value is formed according to the compaction stage: the aggregate rearrangement stage corresponds to encoding value 1, the pore filling stage corresponds to encoding value 2, and the sealing compaction stage corresponds to encoding value 3. The energy coefficient is output in step S4. The travel speed, the vibratory motor current, and the material surface temperature are derived from the equipment status.

[0063] The support vector regression model is trained using offline calibration samples. The offline calibration samples are formed from rolling records of the same type of cold patch material at different compaction stages. In constructing the offline calibration samples, this embodiment applies candidate frequency and candidate amplitude combinations to the cold patch material sample layer. The candidate frequencies are 35 Hz, 40 Hz, 45 Hz, 50 Hz, 55 Hz, 60 Hz, and 65 Hz. The candidate amplitudes are 0.4 mm, 0.8 mm, 1.2 mm, 1.6 mm, and 2.0 mm. Each candidate frequency and each candidate amplitude combination completes one rolling record. Each rolling record obtains the compaction stage and energy coefficient according to steps S1 to S4, and simultaneously records the equipment status. Subsequently, the final porosity after rolling is measured, and the rolling power consumption is calculated based on the vibratory motor current and drive voltage within the same rolling record.

[0064] It should be noted that the target porosity is set at 8.0%. This 8.0% is derived from the design porosity reading in the cold-mix material mix proportion test report. The compaction power consumption is the average output power consumption of the vibratory motor within the same compaction record. During calculation, the vibratory motor current and drive voltage corresponding to each sampling point are read. The instantaneous power consumption is obtained by multiplying the vibratory motor current and drive voltage at the same sampling point, and then averaging all instantaneous power consumptions within the same compaction record to obtain the compaction power consumption. The rated power consumption of the vibratory motor is obtained by multiplying the rated voltage and rated current on the vibratory motor nameplate.

[0065] It is worth noting that the upper limit of the compaction power consumption is set at 90% of the rated power consumption of the vibratory motor. The 90% is determined through thermal stability calibration, specifically by running the vibratory motor continuously for 10 minutes at 80%, 90%, and 100% of the rated power consumption, respectively, and recording the temperature rise of the vibratory motor windings and the fluctuation of the output frequency. The output margin is relatively small at the 80% setting, the winding temperature rise is lower than the rated allowable temperature rise and the output frequency fluctuation does not exceed 1 Hz at the 90% setting, and there are records of the winding temperature rise being close to the rated allowable temperature rise at the 100% setting. Therefore, 90% is determined as the upper limit of the compaction power consumption.

[0066] For all candidate frequency and amplitude combinations under the same compaction stage, this embodiment first eliminates candidate combinations whose rolling power consumption is greater than 90% of the rated power consumption of the vibratory motor, and then calculates the calibration score for the remaining candidate combinations:

[0067] in, The calibration score for candidate combinations, The final porosity corresponding to the candidate combination. The absolute value of the maximum difference between the final porosity and 8.0% for the candidate combinations retained under the same compaction stage. The power consumption corresponding to the candidate combinations is the crushing power. This is the rated power consumption of the vibration motor. When... When the value is 0, the porosity deviation is recorded as 0. The candidate combination with the smallest calibration score is determined as the optimal combination, and the candidate frequency and candidate amplitude in the optimal combination are respectively used as the annotation frequency and annotation amplitude corresponding to the compaction record. If the calibration scores of two candidate combinations are the same, the candidate combination with the smaller compaction power consumption is selected; if the compaction power consumption is still the same, the candidate combination with the smaller candidate amplitude is selected; if the candidate amplitude is still the same, the candidate combination with the smaller candidate frequency is selected. Each offline calibration sample includes the compaction stage code value, energy coefficient, equipment status, annotation frequency, and annotation amplitude.

[0068] It is worth noting that the offline calibration sample size was set at 180 sets, determined through an incremental sample size experiment. Specifically, the support vector regression model was trained using 90, 120, 150, and 180 sets of samples respectively, and the mean absolute error of prediction frequency and mean absolute error of prediction magnitude were statistically analyzed on the validation set. When the sample size increased from 150 to 180 sets, the mean absolute error of prediction frequency decreased from 1.9 Hz to 1.3 Hz, and the mean absolute error of prediction magnitude decreased from 0.16 mm to 0.11 mm. When the sample size continued to increase to 210 sets, the mean absolute error of prediction frequency decreased to 1.2 Hz, and the mean absolute error of prediction magnitude decreased to 0.10 mm. Since the error reduction of 210 sets compared to 180 sets was less than 0.02 mm and less than 0.2 Hz, 180 sets were determined as the offline calibration sample size.

[0069] In training the support vector regression model, this embodiment divides 180 offline labeled samples into five subsets, selecting one subset as the validation set and the remaining four subsets as the training set in turn. The candidate values ​​for the penalty parameter are set to 1, 10, and 100; the candidate values ​​for the radial basis function kernel parameter are set to 0.05, 0.1, and 0.2; and the candidate values ​​for the insensitive loss width are set to 0.05, 0.1, and 0.2. The penalty parameter controls the intensity of the training error penalty, the radial basis function kernel parameter controls the rate of sample similarity decay, and the insensitive loss width controls the width of the interval where the error does not participate in the penalty. Five-fold cross-validation is performed on each candidate combination, and the candidate combination with the smallest normalized sum of the mean absolute error of the prediction frequency and the mean absolute error of the prediction magnitude is selected. In this embodiment, the penalty parameter is 10, the radial basis function kernel parameter is 0.1, and the insensitive loss width is 0.1. The training objective is to reduce the error between the prediction frequency and the labeled frequency, and to reduce the error between the prediction magnitude and the labeled magnitude, under the constraint of the penalty parameter.

[0070] In this embodiment, during parameter prediction, the current compaction stage, current energy coefficient, and current equipment status are converted into model input vectors and input into the pre-trained support vector regression model. The support vector regression model calculates the similarity between the current input vector and the support vectors of each training sample based on the radial basis function kernel function, and then performs a weighted summation based on the support vector weights obtained during the training phase, outputting the predicted frequency and predicted amplitude. The predicted frequency is the vibration frequency value before boundary correction processing, and the predicted amplitude is the vibration amplitude value before boundary correction processing.

[0071] It should be noted that the vibration requirements include the frequency and amplitude boundaries corresponding to each compaction stage. These vibration requirements are obtained through statistical analysis of the labeled frequencies and amplitudes in the offline calibration samples. For all labeled frequencies under the same compaction stage, the 10th and 90th percentiles are taken as frequency boundaries; for all labeled amplitudes under the same compaction stage, the 10th and 90th percentiles are taken as amplitude boundaries. Statistically, the frequency boundary for the aggregate rearrangement stage is 35 Hz to 45 Hz, and the amplitude boundary is 1.5 mm to 2.0 mm; the frequency boundary for the pore filling stage is 45 Hz to 55 Hz, and the amplitude boundary is 1.0 mm to 1.5 mm; the frequency boundary for the sealing compaction stage is 55 Hz to 65 Hz, and the amplitude boundary is 0.4 mm to 1.0 mm.

[0072] This embodiment performs boundary correction processing on the predicted frequency and the predicted amplitude based on the vibration requirements corresponding to the compaction stage. If the predicted frequency is lower than the lower boundary of the frequency corresponding to the current compaction stage, the predicted frequency is corrected to the lower boundary; if the predicted frequency is higher than the upper boundary of the frequency corresponding to the current compaction stage, the predicted frequency is corrected to the upper boundary; if the predicted frequency is within the frequency boundary range corresponding to the current compaction stage, the predicted frequency is retained. The same boundary correction processing is performed on the predicted amplitude to obtain the target amplitude. After the boundary correction processing is completed, this embodiment writes the corrected frequency into the target frequency and the corrected amplitude into the target amplitude.

[0073] For example, the current compaction stage is the sealing compaction stage, with an energy coefficient of 0.28, a travel speed of 1.2 m / s, a vibration motor current of 32 amps, and a material surface temperature of 28 degrees Celsius. In this embodiment, the sealing compaction stage is encoded as 3, and the encoded values ​​3, 0.28, 1.2 m / s, 32 amps, and 28 degrees Celsius form the model input vector. The support vector regression model outputs a predicted frequency of 53 Hz and a predicted amplitude of 1.3 mm. The frequency boundary corresponding to the sealing compaction stage is 55 Hz to 65 Hz, and the amplitude boundary is 0.4 mm to 1.0 mm. In this embodiment, 53 Hz is corrected to 55 Hz, and 1.3 mm is corrected to 1.0 mm, resulting in a target frequency of 55 Hz and a target amplitude of 1.0 mm.

[0074] In step S6, a drive command conversion is performed based on the target frequency and the target amplitude to generate a roller vibration command, including: Determine the target rotational speed of the exciter based on the target frequency; Determine the target angle of the eccentric block based on the target amplitude; The roller vibration command is generated by encoding the command based on the target rotational speed of the exciter and the target angle of the eccentric block.

[0075] In one implementation, the target rotational speed of the vibrator is the rotational speed corresponding to the target frequency of the road roller vibrator, and the vibrator adopts a structure that generates one excitation cycle in a single rotation. In this embodiment, the target frequency is read, and the target frequency is multiplied by 60 to obtain the target rotational speed of the vibrator. The unit of the target rotational speed of the vibrator is revolutions per minute. If the target frequency is 55 Hz, then 55 is multiplied by 60 to obtain a target rotational speed of 3300 revolutions per minute.

[0076] It should be noted that the target angle of the eccentric block is the relative angle between the eccentric blocks of the road roller, and the target angle of the eccentric block is determined by the angle calibration table. The angle calibration table is constructed through static vibration calibration. In this embodiment, the relative angles of the eccentric blocks are set sequentially to 0 degrees, 30 degrees, 60 degrees, 90 degrees, 120 degrees, 150 degrees, and 180 degrees. At each relative angle, the vibrator is controlled to run stably at 50 Hz for 3 seconds, and the vibration amplitude at the roller is collected by the vibration sensor. Each relative angle is collected 5 times, and the median of the 5 vibration amplitudes is taken as the calibration amplitude corresponding to that relative angle. The relative angle and calibration amplitude are written into the angle calibration table in pairs. Different roller models have different eccentric block structures, and the angle calibration table needs to be calibrated for each model according to the above method. For example, the example values ​​obtained by calibrating a certain 13 kN single steel drum roller are as follows: 0 degrees corresponds to 2.2 mm, 30 degrees corresponds to 1.9 mm, 60 degrees corresponds to 1.6 mm, 90 degrees corresponds to 1.2 mm, 120 degrees corresponds to 0.8 mm, 150 degrees corresponds to 0.5 mm, and 180 degrees corresponds to 0.3 mm. The 50 Hz is determined by the median of the labeled frequencies in the offline calibration samples, and the 3 seconds is determined by the vibration amplitude stabilization time, specifically by counting the time required for the vibration amplitude change rate to be continuously lower than 2% after each start-up, and taking the 90th percentile of all records to obtain 3 seconds.

[0077] When determining the target angle of the eccentric block, this embodiment reads the target amplitude and searches for two adjacent calibration amplitudes in the angle calibration table. If the target amplitude is equal to a certain calibration amplitude, the relative angle corresponding to that calibration amplitude is read as the target angle of the eccentric block. If the target amplitude is between two calibration amplitudes, the target angle of the eccentric block is calculated based on the linear ratio between the two calibration amplitudes. When calculating the linear ratio, first calculate the difference between the target amplitude and the smaller calibration amplitude, then calculate the difference between the larger calibration amplitude and the smaller calibration amplitude. Divide the first difference by the second difference to obtain the ratio value. Then multiply the angle difference between the relative angles corresponding to the two calibration amplitudes by the ratio value, and add the relative angle corresponding to the smaller calibration amplitude to obtain the target angle of the eccentric block.

[0078] It is worth noting that the roller vibration command is a digital command sent to the vibration drive circuit of the road roller. The roller vibration command includes a frame header field, a speed field, an angle field, and a check field. The frame header field is fixed at 0xA5. The speed field occupies two bytes and contains the integer value of the target speed of the vibrator. The angle field occupies two bytes and contains the integer value of the target angle of the eccentric block magnified tenfold. The check field occupies one byte and takes the lower eight bits of the sum of the frame header field, speed field, and angle field. During command encoding, this embodiment first converts the target speed of the vibrator into a speed field, then converts the target angle of the eccentric block into an angle field, then generates the check field, and finally concatenates them in the order of the frame header field, speed field, angle field, and check field to generate the roller vibration command.

[0079] It should be noted that this embodiment runs on an ARM Cortex-A53 embedded controller with a main frequency of 800 MHz. The preset time window is 500 milliseconds, and the typical time consumption of each processing step within a single window is obtained by statistically analyzing the running logs of 1000 consecutive windows. The median time consumption for windowed preprocessing is 18 milliseconds, the median time consumption for compaction response characterization is 22 milliseconds, the median time consumption for porosity prediction is 31 milliseconds, the median time consumption for trend fitting is 5 milliseconds, the median time consumption for interval assignment judgment and energy coefficient calculation is 7 milliseconds, the median time consumption for vibration parameter mapping is 28 milliseconds, the median time consumption for drive command conversion is 3 milliseconds, and the median time consumption for communication transmission is 6 milliseconds. The median total time consumption of the above processing steps is 120 milliseconds, and the maximum total time consumption in 1000 consecutive windows is 168 milliseconds, which is less than the 500-millisecond window period.

[0080] For example, the target frequency is 55 Hz and the target amplitude is 1.0 mm. In this embodiment, 55 is multiplied by 60 to obtain a target speed of 3300 rpm for the exciter. In the angle calibration table, 0.8 mm corresponds to 90 degrees and 1.2 mm corresponds to 120 degrees. The target amplitude of 1.0 mm is between 0.8 mm and 1.2 mm. First, the difference of 0.2 mm between 1.0 mm and 0.8 mm is calculated, and then the difference of 0.4 mm between 1.2 mm and 0.8 mm is calculated. Dividing 0.2 mm by 0.4 mm gives a ratio of 0.5. Then, the angle difference of 30 degrees between 120 degrees and 90 degrees is multiplied by 0.5 to obtain 15 degrees. Adding 90 degrees gives a target angle of 105 degrees for the eccentric block. During instruction encoding processing, 3300 is written into the speed field, 1050 is written into the angle field, and a check field is calculated to finally generate the roller vibration instruction. The system executes steps S1 to S6 in a 500-millisecond cycle to form a complete closed-loop control of reaction force acquisition, compaction stage identification, vibration parameter adjustment, and roller vibration command issuance. After the roller vibration command is sent, the sensor continues to acquire the reaction force flow in the next time window and generates the next roller vibration command according to the same processing flow.

[0081] In summary, this invention acquires the reaction flow collected by sensors at the rollers of a road roller within the compaction zone of cold-mixed material. The reaction flow undergoes windowing preprocessing and compaction response characterization processing sequentially to obtain reaction force characteristics. These characteristics within a continuous window are then input into a pre-trained porosity mapping model for porosity prediction, resulting in a pore sequence. The pore sequence is then fitted with a trend to obtain the pore change amount that characterizes the compaction state of the cold-mixed material. Subsequently, the pore change amount is used to determine the compaction stage through interval assignment, and the energy coefficient is calculated based on the compaction stage. Furthermore, vibration parameter mapping processing is performed in conjunction with the equipment status during the road roller's construction process to obtain the target frequency and target amplitude. Finally, the target frequency and target amplitude are used to convert drive commands and generate roller vibration commands, thus achieving continuous linkage between cold-mixed material compaction state identification and construction equipment roller vibration parameter control.

[0082] The second embodiment of the present invention provides a control system for a pressure-sensitive asphalt cold patching construction equipment based on the Internet of Things, including: a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, it implements the method described above.

[0083] It should be noted that the IoT-based pressure-sensitive asphalt cold patching equipment control system provided in this embodiment of the invention is used to execute all process steps of the IoT-based pressure-sensitive asphalt cold patching equipment control method described in the above embodiment. The working principles and beneficial effects of the two are one-to-one, so they will not be described again.

[0084] It should be noted that the system embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0085] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the scope of protection of the present invention. In particular, it should be noted that any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention for those skilled in the art.

Claims

1. A control method for pressure-sensitive asphalt cold patching construction equipment based on the Internet of Things, characterized in that, include: The reaction flow collected by the sensor at the roller of the road roller in the cold patch compaction area is obtained, and the reaction flow is preprocessed by windowing to obtain the reaction sequence; The reaction force sequence is subjected to compaction response characterization processing to obtain reaction force characteristics; The reaction force features within a continuous window are input into a pre-trained porosity mapping model to predict porosity, resulting in a pore sequence. The change trend of the pore sequence is then fitted to obtain the amount of pore change. Based on the porosity change, the interval assignment is determined to obtain the compaction stage, and the energy coefficient is calculated based on the reaction force characteristics according to the compaction stage to obtain the energy coefficient. The equipment status of the road roller during construction is obtained, and vibration parameter mapping is performed based on the compaction stage, the energy coefficient, and the equipment status to obtain the target frequency and target amplitude. Drive commands are converted based on the target frequency and the target amplitude to generate roller vibration commands.

2. The control method for pressure-sensitive asphalt cold patching construction equipment based on the Internet of Things according to claim 1, characterized in that, The step of windowing preprocessing the reaction flow to obtain the reaction force sequence includes: The reaction flow is segmented according to a preset time window to obtain reaction segments; The reaction force segment is subjected to median filtering to obtain a smooth segment; The smoothed segment is subjected to Butterworth low-pass filtering to obtain the reaction force sequence.

3. The control method for pressure-sensitive asphalt cold patching construction equipment based on the Internet of Things according to claim 1, characterized in that, The compaction response characterization process performed on the reaction force sequence to obtain reaction force characteristics includes: The root mean square value and peak factor are calculated based on the reaction force sequence to obtain the time-domain characteristics; Perform a fast Fourier transform on the reaction force sequence to obtain a spectral sequence; Based on the spectral sequence, the dominant frequency component and the frequency band energy ratio are extracted to obtain the frequency domain features; The time-domain features and the frequency-domain features are standardized to obtain the reaction force features.

4. The control method for pressure-sensitive asphalt cold patching construction equipment based on the Internet of Things according to claim 1, characterized in that, The step of inputting the reaction force features within a continuous window into a pre-trained porosity mapping model for porosity prediction to obtain a pore sequence includes: The reaction force features are input into a pre-trained porosity mapping model to predict porosity and obtain porosity values. The porosity values ​​corresponding to the continuous time windows are arranged in a time sequence to obtain the porosity sequence.

5. The control method for pressure-sensitive asphalt cold patching construction equipment based on the Internet of Things according to claim 4, characterized in that, The step of fitting the change trend based on the pore sequence to obtain the pore change amount includes: The porosity change value is calculated based on the porosity value corresponding to adjacent time windows to obtain the change sequence; The trend of change is obtained by performing least-squares fitting on the change sequence. The slope of change is extracted based on the changing trend to obtain the amount of porosity change.

6. The control method for pressure-sensitive asphalt cold patching construction equipment based on the Internet of Things according to claim 1, characterized in that, The step of determining the compaction stage based on the porosity change includes: The porosity change is compared with a preset stage interval. When the change in porosity corresponds to the initial compaction range, the compaction stage is determined to be the aggregate rearrangement stage; When the porosity change corresponds to the intermediate compaction interval, the compaction stage is determined to be the pore filling stage; When the change in porosity corresponds to the final compaction interval, the compaction stage is determined to be the sealing compaction stage.

7. The control method for pressure-sensitive asphalt cold patching construction equipment based on the Internet of Things according to claim 1, characterized in that, The step of calculating the energy coefficient based on the reaction force characteristics during the compaction stage to obtain the energy coefficient includes: The target frequency band is determined based on the compaction stage; Based on the reaction force characteristics, extract the frequency band energy and total frequency band energy corresponding to the target frequency band; The energy coefficient is obtained by calculating the ratio between the energy of the specified frequency band and the total energy of the frequency band.

8. The control method for pressure-sensitive asphalt cold patching construction equipment based on the Internet of Things according to claim 1, characterized in that, The step of mapping vibration parameters based on the compaction stage, the energy coefficient, and the equipment status to obtain the target frequency and target amplitude includes: The compaction stage, the energy coefficient, and the equipment status are input into a pre-trained support vector regression model for parameter prediction processing to obtain the prediction frequency and prediction amplitude. Based on the vibration requirements corresponding to the compaction stage, the predicted frequency and the predicted amplitude are subjected to boundary correction processing to obtain the target frequency and the target amplitude.

9. The control method for pressure-sensitive asphalt cold patching construction equipment based on the Internet of Things according to claim 1, characterized in that, The step of converting the drive command based on the target frequency and the target amplitude to generate a roller vibration command includes: Determine the target rotational speed of the exciter based on the target frequency; Determine the target angle of the eccentric block based on the target amplitude; The roller vibration command is generated by encoding the command based on the target rotational speed of the exciter and the target angle of the eccentric block.

10. A control system for pressure-sensitive asphalt cold patching construction equipment based on the Internet of Things, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the method according to any one of claims 1 to 9.

Images