Road construction state detection system based on dynamic data monitoring

Abstract

The invention relates to the technical field of data processing, in particular to a road construction state detection system based on dynamic data monitoring, which comprises an information fusion module, a feature extraction module, a state classification module, a state detection module and an optimization correction module. According to the method, comprehensive response calculation of multi-source data is realized through sequential alignment and weighted analysis of multi-equipment vibration frequency, hydraulic pressure and pavement compactness, and dynamic correlation features between equipment and a pavement are extracted in covariance relation analysis and feature normalization processing; state classification and trend judgment are carried out in a conditional probability distribution mode, so that a monitoring result can adapt to different time periods and working condition changes, the problems of misjudgment and lag are reduced, smooth correction and abnormal optimization of a fluctuation interval are completed in the state fluctuation analysis and data correction process, stability and continuity of a construction state recognition result are guaranteed, and the construction efficiency is improved. And the construction monitoring accuracy and the overall reliability are improved.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and in particular to a road construction status monitoring system based on dynamic data monitoring. Background Technology

[0002] The field of data processing technology encompasses a series of methodologies for collecting, organizing, calculating, analyzing, and expressing multi-source raw data. Its core aspects include data acquisition methods, data format standardization methods, data computation and statistical methods, and processing flows for generating structured information based on computation results. Typically, the representation and description of the operational state of an objective object are achieved through steps such as setting mathematical operation rules, filtering conditions, making logical judgments, and deriving state variables on continuous or discrete data.

[0003] Traditional road construction status monitoring systems based on dynamic data monitoring rely on time-series data generated at the road construction site. They acquire raw measurements such as vibration changes, displacement changes, environmental noise changes, and vehicle traffic changes by deploying measuring devices in the construction area. These raw measurements are then processed by comparing time-period values ​​with fixed thresholds to determine the mean difference and trend of the sections. This process determines the status of different work stages during construction. Traditional methods typically rely on manually setting time windows for segmented calculations of the collected data and performing status judgments based on fixed empirical thresholds to identify the status information of the construction site.

[0004] Existing technologies for road construction status monitoring rely on manually set time windows and fixed thresholds for status judgment. This approach cannot accurately identify the dynamic correlations between multiple sources of data at the construction site. Limited by the static threshold judgment method, it is prone to misjudgment and omission under different working conditions and equipment operation conditions. It lacks comprehensive analysis of the inherent relationships between multi-dimensional data such as vibration, pressure, and pavement compaction, resulting in monitoring results that cannot truly reflect the continuous changes in the construction process. In addition, the manual division of time segments causes the system to lag in response to abnormal fluctuations, making it impossible to detect short-term fluctuations and trend changes in the construction status in a timely manner. This results in poor stability of status identification results, affecting the accuracy and reliability of construction monitoring. Summary of the Invention

[0005] To address the technical problems existing in the prior art, embodiments of the present invention provide a road construction status monitoring system based on dynamic data monitoring. The technical solution is as follows: On the one hand, a road construction status monitoring system based on dynamic data monitoring is provided, which includes: The information fusion module collects vibration frequency, hydraulic pressure, construction speed and road surface compaction of multiple construction equipment, performs time-series alignment and weighted average analysis on all data to generate construction status monitoring dataset and transmits it to the feature extraction module. The feature extraction module analyzes equipment and road surface features based on the construction status monitoring dataset, calculates the covariance relationship between the vibration frequency of the construction equipment, hydraulic pressure, construction speed and compaction degree of the construction road surface, extracts key dynamic features and normalizes them, generates a construction dynamic feature set and transmits it to the status classification module. The status classification module classifies the construction status based on the construction dynamic feature set, calculates the conditional probability distribution between equipment and road surface features, determines the construction status category for multiple time periods, generates the construction status classification result, and transmits it to the status detection module. The status detection module analyzes the status changes in the time series based on the construction status classification results, calculates the status change trend of adjacent time periods and extracts the fluctuation amplitude and segment, judges the stability of the construction status based on the fluctuation amplitude and fluctuation threshold, and forms the construction status detection results. The optimization and correction module, based on the construction status detection results, recalibrates the vibration frequency, hydraulic pressure, and road surface compaction degree for the fluctuation range, and performs weighted smoothing analysis and anomaly correction to generate an optimized set of construction status detection results.

[0006] As a further embodiment of the present invention, the construction status monitoring dataset includes time-aligned vibration frequency data, hydraulic pressure data, and construction pavement compaction data; the construction dynamic feature set includes the covariance of equipment vibration frequency and hydraulic pressure, key dynamic feature vectors, and normalized feature scales; the construction status classification results include a multi-time period construction status category sequence, conditional probability distribution, and time period classification results; the construction status detection results include status change trends, fluctuation amplitudes, and stability judgment indicators; and the construction status detection optimization result set includes a correction parameter set, a smoothing weighted sequence set, and anomaly correction records.

[0007] As a further aspect of the present invention, the information fusion module includes: The data analysis submodule acquires vibration frequency, hydraulic pressure, construction speed and road surface compaction data of multiple construction equipment, extracts the main frequency component from the vibration signal through frequency domain decomposition, extracts transient pressure from the hydraulic pressure, and synchronizes it with the compaction data in time to generate a multi-source construction raw dataset. The time-series alignment submodule calculates the time difference between sampling timestamps of multiple types of sensor data based on the multi-source construction original dataset, performs interpolation and resampling operations on multiple data sequences, reconstructs continuous time series according to a unified time step, calculates the mean square of synchronization error, and generates a multi-source synchronized time series dataset. The weighted analysis submodule calculates the response amplitudes of vibration frequency, hydraulic pressure, construction speed, and road surface compaction at multiple time steps based on the multi-source synchronous time series dataset, analyzes the weight ratios, combines multiple weighted terms into construction response quantities, and smooths them by moving average to generate a construction status monitoring dataset.

[0008] As a further aspect of the present invention, the feature extraction module includes: The covariance calculation submodule, based on the construction status monitoring dataset, extracts vibration frequency, hydraulic pressure, construction speed and construction pavement compaction degree for time synchronization, and calculates the covariance relationship between multiple time steps and integrates them with weights to generate a set of construction feature covariance values. The feature extraction submodule calls the construction feature covariance value group, performs difference analysis on the multi-covariance value sequence, extracts the rate of change and fluctuation amplitude within the time period, filters the main response components based on the amplitude change interval, and generates the key dynamic feature sequence of construction. The feature normalization submodule performs interval linear transformation on each component based on the key construction dynamic feature sequence, mapping the values ​​to a unified scale range and generating a construction dynamic feature set.

[0009] As a further aspect of the present invention, the state classification module includes: The conditional distribution submodule, based on the construction dynamic feature set, extracts equipment operation feature components and road surface response feature components and performs time step pairing, calculates the joint occurrence frequency of the two types of components at the same time step, analyzes the conditional probability by combining the occurrence frequency of single components, and generates a conditional probability distribution set of equipment and road surface. The state determination submodule calls the device and road surface condition probability distribution set, calculates and divides the probability fluctuation interval based on the amplitude of condition probability change over a continuous time period, compares the condition probability intervals corresponding to multiple time periods with the classification threshold, determines the construction state category of multiple intervals, and generates a construction state sequence for multiple time periods. The status result submodule performs continuity detection on the status categories of the multi-time period construction status sequence, calculates the transition frequency value between adjacent status categories, reconstructs the status time mapping relationship based on the transition frequency distribution, and generates construction status classification results.

[0010] As a further aspect of the present invention, the classification threshold is calculated by using statistical equipment and the range of conditional probability values ​​corresponding to multiple construction state categories in the road surface conditional probability distribution set, to calculate the mean and variance of conditional probabilities under multiple state categories, and to analyze the degree of difference in mean values.

[0011] As a further aspect of the present invention, the state detection module includes: The time segmentation submodule acquires and segments the time series data corresponding to the construction status classification results, extracts the proportion of status categories and the duration of status within multiple time periods, analyzes the continuity of categories in adjacent time periods, calculates the number of status changes and the duration within multiple segments, and generates segment distribution data. The state trend submodule, based on the segment distribution data, calls the state category sequence of adjacent time periods, judges the direction of state change between adjacent time periods, calculates the rate of change and duration of change corresponding to the direction of change, analyzes the fluctuation amplitude value sequence, and obtains the state fluctuation amplitude sequence. The stability determination submodule compares the difference between the fluctuation amplitude and the fluctuation threshold over multiple time periods based on the state fluctuation amplitude sequence, determines the stability interval distribution of state changes within multiple segments, calculates the ratio of the proportion of stable segments to the proportion of fluctuating segments, and obtains the construction state detection results.

[0012] As a further aspect of the present invention, the fluctuation threshold is obtained by extracting the fluctuation amplitude distribution of multiple state categories in the time series, calculating the mean and standard deviation of the fluctuation amplitude, and then analyzing the difference between the mean and standard deviation.

[0013] As a further aspect of the present invention, the optimization and correction module includes: The data correction submodule acquires the fluctuation interval monitoring data in the construction status detection results, calculates the time change rate for vibration frequency, hydraulic pressure and construction pavement compaction degree, performs linear translation and amplitude redistribution within the interval on the data points, and generates a set of correction parameters. The weighted smoothing submodule, based on the correction parameter set, assigns weights to the multi-parameter sequence, analyzes the weight coefficients according to the ratio of vibration frequency change rate, hydraulic pressure difference value and compaction deviation, calls the average weight of adjacent time series to perform sliding smoothing on the parameters, and generates a smoothed weighted sequence set; The anomaly correction submodule calculates the difference rate between multiple sequence points and the mean of the same interval based on the smooth weighted sequence set, compares the difference rate with the anomaly threshold, filters out anomaly points and replaces them with the mean within the interval, and generates a construction status detection optimization result set through difference rate correction and reconstruction of the parameter sequence. The abnormal threshold is set by statistically analyzing the distribution characteristics of the difference rate of multiple parameters in the construction monitoring data.

[0014] The beneficial effects of the technical solutions provided in the embodiments of the present invention include at least the following: By performing time-series alignment and weighted analysis of vibration frequencies, hydraulic pressures, and road compaction degrees of multiple devices, comprehensive response calculations of multi-source data are achieved. Dynamic correlation features between equipment and road surface are extracted through covariance relationship analysis and feature normalization processing. Conditional probability distribution is used for state classification and trend judgment, enabling monitoring results to adapt to changes in different time periods and working conditions, reducing misjudgments and lag issues. In the process of state fluctuation analysis and data correction, smoothing correction and anomaly optimization of fluctuation ranges are completed, ensuring the stability and continuity of construction state identification results, and improving the accuracy and overall reliability of construction monitoring. Attached Figure Description

[0015] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0016] Figure 1 This is a schematic diagram of the system of the present invention; Figure 2 This is a schematic diagram of the system framework of the present invention; Figure 3 This is a flowchart of the information fusion module in this invention; Figure 4 This is a flowchart of the feature extraction module in this invention; Figure 5 This is a flowchart of the state classification module in this invention; Figure 6 This is a flowchart of the state detection module in this invention; Figure 7 This is a flowchart of the optimized correction module in this invention. Detailed Implementation

[0017] The technical solution of the present invention will now be described with reference to the accompanying drawings.

[0018] In embodiments of the present invention, words such as "exemplarily," "for example," etc., are used to indicate that something is an example, illustration, or description. Any embodiment or design described as "exemplary" in the present invention should not be construed as being more preferred or advantageous than other embodiments or designs. Specifically, the use of the word "exemplary" is intended to present the concept in a concrete manner. Furthermore, in embodiments of the present invention, the meaning expressed by "and / or" can be both, or either one.

[0019] In the embodiments of this invention, the terms "image" and "picture" may sometimes be used interchangeably. It should be noted that, without emphasizing the distinction between them, they convey the same meaning. Similarly, the terms "of," "corresponding (relevant)," and "corresponding" may sometimes be used interchangeably. It should be noted that, without emphasizing the distinction between them, they convey the same meaning.

[0020] In this embodiment of the invention, sometimes a subscript such as W1 may be written in a non-subscript form such as W1. When the difference is not emphasized, the meaning they express is the same.

[0021] To make the technical problems, technical solutions and advantages of the present invention clearer, a detailed description will be given below in conjunction with the accompanying drawings and specific embodiments.

[0022] This invention provides a road construction status monitoring system based on dynamic data monitoring, such as... Figures 1-2 The diagram shown illustrates a road construction status monitoring system based on dynamic data monitoring. The system includes: The information fusion module collects vibration frequency, hydraulic pressure, construction speed and road surface compaction of multiple construction equipment, performs time-series alignment and weighted average analysis on all data to generate construction status monitoring dataset and transmits it to the feature extraction module. The feature extraction module analyzes equipment and road surface features based on the construction status monitoring dataset, calculates the covariance relationship between the vibration frequency of construction equipment, hydraulic pressure, construction speed and compaction degree of construction road surface, extracts key dynamic features and normalizes them, generates a construction dynamic feature set and passes it to the status classification module. The status classification module classifies construction status based on the construction dynamic feature set, calculates the conditional probability distribution between equipment and road features, determines the construction status category in multiple time periods, generates construction status classification results, and transmits them to the status detection module. The status detection module analyzes the status changes in the time series based on the construction status classification results, calculates the status change trend of adjacent time periods and extracts the fluctuation amplitude and segment, judges the stability of the construction status based on the fluctuation amplitude and fluctuation threshold, and forms the construction status detection results. The optimization and correction module, based on the construction status detection results, recalibrates the vibration frequency, hydraulic pressure, and road surface compaction degree for the fluctuation range, and performs weighted smoothing analysis and anomaly correction to generate an optimized set of construction status detection results.

[0023] The construction status monitoring dataset includes time-aligned vibration frequency data, hydraulic pressure data, and pavement compaction data. The construction dynamic feature set includes the covariance of equipment vibration frequency and hydraulic pressure, key dynamic feature vectors, and normalized feature scales. The construction status classification results include multi-time period construction status category sequences, conditional probability distributions, and time period classification results. The construction status detection results include status change trends, fluctuation amplitudes, and stability judgment indicators. The construction status detection optimization result set includes a set of correction parameters, a set of smoothed weighted sequences, and anomaly correction records.

[0024] Specifically, such as Figure 2 , 3 As shown, the information fusion module includes: The data analysis submodule acquires vibration frequency, hydraulic pressure, construction speed and road surface compaction data of multiple construction equipment, extracts the main frequency component from the vibration signal through frequency domain decomposition, extracts transient pressure from the hydraulic pressure, and synchronizes it with the compaction data in time to generate a multi-source construction raw dataset. The process involves acquiring raw measurement signals from the accelerometer sensor on the vibrating drum of the road roller, the pressure sensor of the hydraulic system, the overall speed sensor, and a non-contact pavement compaction tester. The process of extracting the dominant frequency component from the vibration signal through frequency domain decomposition is as follows: Continuous vibration acceleration signals are acquired from the accelerometer sensor installed on the vibrating drum. The sampling frequency is set to 1024 Hz, and continuous acquisition is performed for 1 second to obtain a vibration signal sample array consisting of 1024 acceleration value points. For example, a sample array might be [-0.2, -0.5, 0.1, ..., 1.5, 0.9]. A Fast Fourier Transform (FFT) is then performed on this sequence of 1024 data points to transform it from the time domain to the frequency domain, obtaining a set of frequency components and their corresponding amplitudes. This transformation produces an amplitude spectrum, where the horizontal axis of each point represents the frequency, and the vertical axis represents the amplitude of that frequency component. For example, within the frequency range of 0 to 512 Hz, the calculated amplitude at 35 Hz is 1.2g, while at other frequencies, such as 30 Hz, the amplitude is 0.4g, and at 40 Hz, it is 0.6g. By iterating through the amplitude values ​​at all frequencies, the frequency with the largest amplitude is found. In this example, the amplitude of 1.2g corresponding to 35 Hz is the maximum value among all amplitudes; therefore, 35 Hz is determined as the dominant frequency component for this time period. The process of extracting transient pressure from hydraulic pressure is as follows: pressure signals are continuously acquired from pressure sensors installed in the hydraulic system, with the sampling frequency set to 10 Hz, i.e., a pressure value is recorded every 0.1 seconds. For each newly acquired transient pressure value, the arithmetic mean of the 10 data points (comprising the previous 9 acquisition points) is calculated. For example, the transient pressure value collected at time point T is 25.5 MPa, while the nine pressure values ​​from time point T-0.9 seconds to T-0.1 seconds are [25.1, 25.2, 25.0, 25.1, 25.3, 25.2, 25.4, 25.3, 25.2] MPa, with an average value of 25.2 MPa. Subtracting this average value of 25.2 MPa from the current transient pressure value of 25.5 MPa yields a difference of 0.3 MPa, which is the transient pressure. Subsequently, the dominant vibration frequency component, the transient pressure, along with the construction speed data (e.g., 2.5 km / h) obtained from the velocity sensor and the compaction data (e.g., 92.5%) obtained from the compaction tester, are aligned according to the unified timestamp recorded by the data acquisition equipment. For example, the vibration dominant frequency component of 35 Hz, transient pressure of 0.3 MPa, construction speed of 2.5 km / h, and compaction degree of 92.5% obtained at the timestamp "14:22:05.100" are combined into a single data record. This process is repeated to continuously generate data records at multiple time points, forming a multi-source construction raw dataset.

[0025] The time-series alignment submodule, based on the original multi-source construction dataset, calculates the time difference between sampling timestamps of data from multiple types of sensors, performs interpolation and resampling operations on multiple data sequences, reconstructs continuous time series with a uniform time step, calculates the mean square of synchronization error, and generates a multi-source synchronized time series dataset. Based on the generated multi-source construction raw dataset, the time difference between sampling timestamps of various sensor data is first calculated. Specifically, the timestamps of each data sequence in the raw dataset are read sequentially. For example, the timestamp sequence for vibration data is [10.000s, 10.010s], with a sampling interval of 10 milliseconds; while the timestamp sequence for compaction data is [10.050s, 10.150s], with a sampling interval of 100 milliseconds. The calculated time difference between two adjacent vibration data points is 10 milliseconds, and the time difference between two adjacent compaction data points is 100 milliseconds. Subsequently, interpolation and resampling operations are performed on these data sequences with different sampling rates. A unified time step is determined, and this step is set based on experimental verification. To set this step step, the following experiment was conducted: ten typical road sections, each 100 meters long, were selected, and data processing and compaction quality analysis were performed using unified time steps of 50 milliseconds, 100 milliseconds, 200 milliseconds, and 500 milliseconds, respectively. Experimental data show that when using a 100-millisecond time step, the error between the compaction degree prediction and the actual core sampling verification results is the smallest, with an average absolute error of 0.21%, while the error is greater than 0.35% when using other time steps. Therefore, a unified time step of 100 milliseconds is determined. Based on this time step, a continuous time series [10.000s, 10.100s, 10.200s] with 100-millisecond intervals is reconstructed. For data sequences with sampling intervals less than 100 milliseconds, such as vibration data, downsampling is performed. For example, for a new time point of 10.100 seconds, the arithmetic mean of all vibration dominant frequency components between 10.050 seconds and 10.150 seconds of the original timestamp is selected as the new data for that point. For data sequences with sampling intervals greater than 100 milliseconds, such as construction speed (one point per second), linear interpolation is performed. If the speed at 10.000 seconds is 2.5 km / h and the speed at 11.000 seconds is 2.7 km / h, then the interpolated speed at 10.100 seconds is calculated as 2.5 + (2.7 - 2.5) * (10.100 - 10.000) / (11.000 - 10.000) = 2.52 km / h. In this way, the corresponding vibration frequency, hydraulic pressure, construction speed, and compaction degree are calculated for each new unified timestamp. Finally, the mean square of the synchronization error is calculated by interpolating the original data to its original timestamp and comparing it with the original value. For example, for a vibration frequency of 35.2 Hz at the original time point of 10.010 seconds, which does not exist in the resampled sequence, the value obtained through reverse interpolation is 35.21 Hz, then the squared error is (35.21 - 35.2)^2 = 0.0001. The squared error is calculated for all original data points, and then the average value is taken to obtain the mean square of the synchronization error, such as 0.0025. This process ultimately generates a multi-source synchronized time series dataset.

[0026]

[0027] As shown in Table 1, this table lists some data records under a unified time step after time alignment, showing the synchronization status of each physical quantity at continuous time points.

[0028] The weighted analysis submodule calculates the response amplitudes of vibration frequency, hydraulic pressure, construction speed and pavement compaction at multiple time steps based on the multi-source synchronous time series dataset, analyzes the weight ratios, combines multiple weighted terms into construction response quantities and smooths them by moving average, and generates a construction status monitoring dataset. Based on the generated multi-source synchronous time-series dataset, a sliding time window with a width of 5 time steps (500 milliseconds) is first set. For the data within this window, the response amplitudes of vibration frequency, hydraulic pressure, and construction speed are calculated. The response amplitude is calculated by subtracting the minimum value from the maximum value of the data sequence within the window. Taking the data in Table 1 as an example, for a time window starting at 10.000 seconds [10.000s, 10.100s, 10.200s, 10.300s, 10.400s], the vibration frequency data is [35.1, 35.0, 35.2, 35.3, 35.1] Hz, and its response amplitude is 35.3 - 35.0 = 0.3 Hz. Similarly, the transient pressure data is [0.32, 0.28, 0.35, 0.41, 0.38] MPa, and the response amplitude is 0.41 - 0.28 = 0.13 MPa. The construction speed data were [2.50, 2.52, 2.54, 2.56, 2.58] km / h, with a response amplitude of 2.58 - 2.50 = 0.08 km / h. Subsequently, the weighting ratio of each response amplitude was analyzed. The weighting was set according to the following experimental verification process: In a standard test field, a single construction parameter (vibration frequency, hydraulic pressure, speed) was changed, and its impact on the compaction rate was recorded. Through multiple regression analysis of 100 sets of experimental data, the standardized regression coefficients of each parameter's contribution to compaction were obtained, and weights were set based on these coefficients. The experimental results showed that the standardized regression coefficient for transient pressure was 0.81, for vibration frequency it was 0.58, and for construction speed it was 0.23. To obtain the weights, these coefficient values ​​were normalized. The total coefficient sum was calculated as 0.81 + 0.58 + 0.23 = 1.62. Based on this, the weights are calculated as follows: transient pressure weight is 0.81 / 1.62 = 0.50; vibration frequency weight is 0.58 / 1.62 = 0.36; and construction speed weight is 0.23 / 1.62 = 0.14. Subsequently, the weighted terms are combined to form the construction response quantity. The calculation process is as follows: multiply the response amplitude of each physical quantity by its corresponding weight, and then sum them up to obtain the construction response quantity. =(0.3*0.36)+(0.13*0.50)+(0.08*0.14)=0.108+0.065+0.0112=0.1842. This construction response quantity is recorded at the end of the current time window, i.e., 10.400 seconds. Next, the time series of construction response quantities generated in this way is smoothed by a moving average. A moving average window length of 3 is set. For the construction response quantity at 10.400 seconds, its smoothed value is the arithmetic mean of itself and the construction response quantities at the two previous time points (10.300 seconds and 10.200 seconds). Assuming the construction response quantities at the first two points are 0.1790 and 0.1815 respectively, the smoothed value at 10.400 seconds is (0.1842+0.1790+0.1815) / 3=0.1816. This smoothed data series is the construction status monitoring dataset.

[0029] Specifically, such as Figure 2 , 4 As shown, the feature extraction module includes: The covariance calculation submodule, based on the construction status monitoring dataset, extracts vibration frequency, hydraulic pressure, construction speed and construction pavement compaction degree for time synchronization, and calculates the covariance relationship between multiple time steps and integrates them with weights to generate a set of construction feature covariance values. Based on the generated construction status monitoring dataset, synchronous time-series data for four dimensions—vibration frequency, hydraulic pressure, construction speed, and pavement compaction—are first extracted from the dataset. Next, a sliding calculation window with a width of 10 time steps (i.e., 1 second) is set, and the covariance relationship between multiple time steps is calculated for the data within this window. Specifically, the covariance between vibration frequency and pavement compaction is first calculated. Ten vibration frequency data points are extracted from the window; for example, a sequence of [35.1, 35.0, 35.2, 35.3, 35.1, 35.0, 34.9, 35.1, 35.2, 35.1] Hz, with an arithmetic mean of 35.1 Hz. Simultaneously, 10 corresponding compaction data points were extracted, with the sequence [92.5, 92.6, 92.7, 92.8, 92.9, 93.0, 93.1, 93.2, 93.3, 93.4]%, whose arithmetic mean was 92.95%. For each pair of time-synchronized data points, the product of the differences between each point and the mean was calculated. For example, the calculation for the first point was (35.1-35.1)*(92.5-92.95)=0. The calculation results of the 10 points were summed, resulting in a total of 0.45. Finally, this total was divided by (time step number - 1), i.e., 0.45 / (10-1)=0.05, which is the covariance between vibration frequency and compaction degree. Using the exact same calculation steps, the covariance between hydraulic pressure and compaction degree, as well as the covariance between construction speed and compaction degree, were calculated, yielding values ​​of 0.12 and -0.08, respectively. Subsequently, the three covariance values ​​were weighted and integrated. The weights were set based on the following experiment: compaction operations were carried out on 10 test road sections with different base materials, the covariance of each parameter and the degree of compaction was recorded, and the correlation with the uniformity of the final core sample compaction was analyzed.

[0030]

[0031] As shown in Table 2, the standardized coefficients of the influence of each covariance value on the uniformity of compaction were determined by performing multiple linear regression analysis on the data in the table. The calculated coefficients for hydraulic pressure, vibration frequency, and construction speed were 0.65, 0.25, and 0.10, respectively. These three coefficient values ​​were used as weights. The weighted integration calculation is as follows: (0.12*0.65)+(0.05*0.25)+(-0.08*0.10)=0.078+0.0125-0.008=0.0825. This calculation result is the construction feature covariance value at the center point of the current time window. The above calculation is repeated every time the sliding window advances by one time step (100 milliseconds), thereby continuously generating a set of construction feature covariance values.

[0032] The feature extraction submodule calls the construction feature covariance value group, performs difference analysis on the multi-covariance value sequence, extracts the rate of change and fluctuation amplitude within the time period, filters the main response components based on the amplitude change interval, and generates the key dynamic feature sequence of construction. The generated covariance value set for construction features is called. This set is a time series, for example, values ​​at five consecutive time points: [0.0825, 0.0850, 0.0835, 0.0910, 0.0880]. First, difference analysis is performed on this covariance value series to extract the rate of change within the time period. The difference between two adjacent value points is calculated; for example, the rate of change at the second time point is 0.0850 - 0.0825 = 0.0025, and the rate of change at the third time point is 0.0835 - 0.0850 = -0.0015. This yields a rate of change sequence: [0.0025, -0.0015, 0.0075, -0.0030]. Next, the fluctuation amplitude is extracted. A sliding window with a width of five time steps is set, and the difference between the maximum and minimum covariance values ​​within the window is calculated. For the sequence [0.0825, 0.0850, 0.0835, 0.0910, 0.0880], the maximum value is 0.0910, the minimum value is 0.0825, and the fluctuation range is 0.0910-0.0825=0.0085. Subsequently, the main response components were screened based on the amplitude variation range. The amplitude variation range was set based on statistical analysis of historical construction data. Statistical data showed that when the fluctuation amplitude value exceeded a certain range, it often corresponded to a significant change in the compaction process or an anomaly in the materials. The experimental process was as follows: The covariance fluctuation range within 10 seconds before and after the occurrence of 200 historical construction anomaly events (such as a sudden drop in compaction degree or abnormal equipment parameter alarms) was analyzed. It was found that the fluctuation amplitude corresponding to 95% of the anomaly events was greater than 0.0080. Therefore, the amplitude variation range was set into two levels: the normal response range (less than 0.0080) and the main response range (greater than or equal to 0.0080). In the aforementioned example, the calculated fluctuation amplitude was 0.0085, which falls within the principal response range. The filtering operation determines whether the fluctuation amplitude is greater than or equal to 0.0080. If this condition is met, the peak value of the covariance sequence within that time window (i.e., 0.0910) is extracted as the principal response component. If the fluctuation amplitude is 0.0075 (less than 0.0080), no component is extracted. By continuously performing this judgment and extraction operation on the entire set of construction feature covariance values, a sequence of key dynamic features for construction, composed of the filtered peak values, is generated.

[0033] The feature normalization submodule performs interval linear transformation on each component based on the key dynamic feature sequence of construction, mapping the values ​​to a unified scale range and generating a construction dynamic feature set. Based on the generated key dynamic feature sequence of construction, an interval linear transformation is performed on each component of the sequence. This sequence consists of a series of discontinuous covariance peaks extracted only when certain conditions are met; for example, a sequence segment might be [0.0910, 0.0955, 0.0890]. The goal of the interval linear transformation is to map these values ​​to a uniform scale range of [0, 1]. This transformation process requires a maximum reference value and a minimum reference value. The determination of these two reference values ​​is based on experimental verification using a large amount of historical data. Specifically, the experimental process involves collecting and processing construction status monitoring datasets from 50 different highway projects, extracting all key dynamic feature sequences, and performing statistical analysis on over 1 million data points. The statistical results show that the minimum value of all these dynamic feature values ​​is 0.0815, and the maximum value is 0.1215. Therefore, the minimum reference value is set to 0.0800, and the maximum reference value is set to 0.1250. This range covers most of the historical data and allows for a certain margin. For the first component 0.0910 in the sequence, the normalization calculation process is as follows: First, calculate the difference between this value and the minimum reference value, that is, 0.0910-0.0800=0.0110; Then calculate the difference between the maximum and minimum reference values, i.e., 0.1250 - 0.0800 = 0.0450; finally, divide the two differences, 0.0110 / 0.0450, the result is 0.244. For the second component in the sequence... For 0.0955, the same steps are used for calculation: (0.0955-0.0800) / (0.1250-0.0800)=0.0155 / 0.0450, resulting in 0.344. For the third component 0.0890, the calculation result is: (0.0890-0.0800) / (0.1250-0.0800)=0.0090 / 0.0450, the calculation result is 0.200. After performing the above linear transformation on each value in the construction key dynamic feature sequence, the new sequence [0.244, 0.344, 0.200] is the final generated construction dynamic feature set.

[0034] Specifically, such as Figure 2 , 5 As shown, the state classification module includes: The conditional distribution submodule, based on the construction dynamic feature set, extracts the equipment operation feature components and the road surface response feature components and performs time step pairing, calculates the joint occurrence frequency of the two types of components at the same time step, combines the occurrence frequency of single components to analyze the conditional probability, and generates the equipment and road surface conditional probability distribution set. Based on the generated construction dynamic feature set, the first step is to associate this single feature with the multi-dimensional original states of equipment and road surface. Specifically, from the "multi-source synchronized time-series dataset" synchronized with the construction dynamic feature set, equipment operation feature components and road surface response feature components are extracted and time-stepped. The equipment operation feature component is defined as vibration frequency, and the road surface response feature component is defined as road surface compaction degree. To calculate the joint occurrence frequency, the continuous feature components need to be discretized. The equipment operation feature component (vibration frequency) is divided into two states: when the frequency is below 35 Hz, it is denoted as "low frequency"; when the frequency is greater than or equal to 35 Hz, it is denoted as "high frequency". This 35 Hz benchmark value was obtained by analyzing the construction data of 100 excellent compaction sections, and statistics showed that 96% of the excellent sections had a stable vibration frequency above 35 Hz during construction. The road surface response feature component (compaction degree) is also divided into two states: when the compaction degree is below 93%, it is denoted as "weak compaction"; when the compaction degree is greater than or equal to 93%, it is denoted as "strong compaction". The 93% threshold is set according to the minimum requirements for the compaction degree of the highway subgrade in the "Specifications for Field Testing of Highway Subgrade and Pavement" (JTG3450-2019). Subsequently, within a calculation cycle containing 1000 consecutive time steps, the number of times the two components jointly occur at the same time step is counted. For example, the number of times "high frequency" and "strong compaction" occur simultaneously is counted, yielding a count of 680. Next, the conditional probability is analyzed by combining the occurrence frequency of each component. In these 1000 time steps, the total number of occurrences of the "high frequency" state is counted as 800. The conditional probability is calculated as follows: the number of times "high frequency" and "strong compaction" occur jointly (680) is divided by the total number of times "high frequency" occurs (800), resulting in 680 / 800 = 0.85. This value indicates that the conditional probability of the pavement exhibiting a "strong compaction" state under the condition that the equipment is in a "high frequency" state is 0.85. Similarly, calculate the conditional probabilities for all other combinations. For example, if "low frequency" occurs 200 times and "low frequency" and "strong compaction" occur 30 times, then the conditional probability of "strong compaction" under the condition of "low frequency" is 30 / 200 = 0.15. All the conditional probability combinations calculated at each time step (e.g., P(strong compaction|high frequency) = 0.85, P(weak compaction|high frequency) = 0.15, P(strong compaction|low frequency) = 0.15, P(weak compaction|low frequency) = 0.85) constitute a data point. Perform this calculation continuously throughout the entire construction process to ultimately generate the conditional probability distribution set for equipment and road surface.

[0035] The status determination submodule calls the equipment and road surface condition probability distribution set, calculates the probability fluctuation interval based on the amplitude of condition probability change over a continuous time period and divides it, compares the condition probability intervals corresponding to multiple time periods with the classification threshold, determines the construction status category of multiple intervals, and generates a construction status sequence for multiple time periods. The generated set of conditional probability distributions for equipment and road surface is retrieved; this is a time series data set. A key conditional probability, such as "P(compaction strength | high frequency)", is selected, with its sequence values ​​over 10 consecutive time steps: [0.85, 0.88, 0.92, 0.91, 0.82, 0.78, 0.79, 0.96, 0.97, 0.95]. First, the probability fluctuation interval is calculated and segmented based on the amplitude of the conditional probability change over consecutive time periods. A sliding window with a width of 5 time steps is set, and the difference between the maximum and minimum probability values ​​within the window is calculated, i.e., the amplitude of change. For the first 5 points of the sequence [0.85, 0.88, 0.92, 0.91, 0.82], the amplitude of change is 0.92 - 0.82 = 0.10. A threshold of 0.09 for the amplitude of change is set; when the calculated amplitude exceeds this threshold, a state transition is considered to have occurred, and the time period is segmented at this point. The threshold of 0.09 was set based on the analysis of 50 known construction data segments that transitioned from "undervoltage" to "appropriate voltage" states. Statistical results showed that the conditional probability change amplitude within the 5-step window corresponding to 95% of the state transition events exceeded 0.09. In the aforementioned example, the change amplitude of 0.10 in the first window was greater than 0.09, thus creating a segmentation point at the 5th time step. The conditional probability intervals corresponding to multiple time periods were compared with the classification threshold to determine the construction state category of the multiple intervals. For the first segmented time period [0.85, 0.88, 0.92, 0.91, 0.82], the arithmetic mean of all probability values ​​within it was calculated, resulting in (0.85+0.88+0.92+0.91+0.82) / 5=0.876. The conditional probability average intervals corresponding to different construction state categories are: undervoltage (below 0.80), appropriate voltage (0.80 to 0.95), and overvoltage (above 0.95). These thresholds were set based on statistical analysis of a large amount of marked road segment data (verified by core sampling). The average value of 0.876 calculated in the example was compared with the threshold; since it fell within the range [0.80, 0.95], the construction status for that time period was determined to be "appropriate pressure state". For the last 5 points of the sequence [0.78, 0.79, 0.96, 0.97, 0.95], the average value was 0.89, also determined to be "appropriate pressure state". By performing this segmentation, calculation, and comparison process on the entire conditional probability sequence, a multi-time period construction status sequence composed of categories such as "underpressure state", "appropriate pressure state", and "overpressure state" was finally generated.

[0036] The status results submodule performs continuity detection on the status categories of multiple time periods based on the construction status sequence of multiple time periods, calculates the transition frequency value between adjacent status categories, reconstructs the status time mapping relationship based on the transition frequency distribution, and generates construction status classification results. Based on the generated multi-time-segment construction state sequence, for example, a long sequence of [appropriate compaction, appropriate compaction, excessive compaction, appropriate compaction, appropriate compaction, appropriate compaction, insufficient compaction, insufficient compaction, insufficient compaction, appropriate compaction]. First, a continuity test is performed on this sequence. A minimum continuous length threshold of 3 time segments is set. This threshold is based on the following experiment: short-term (lasting 1-2 time segments) fluctuations in equipment parameters are artificially created on the test section, and it is found that this produces isolated, discontinuous state "jumps" in the state sequence, while real changes in compaction state (such as from insufficient compaction to appropriate compaction) all show continuous states lasting more than 3 time segments. Therefore, the state sequence is checked, and if a certain state category appears less than 3 times consecutively, it is judged as a pseudo state. For example, in the sequence [appropriate compaction, appropriate compaction, excessive compaction, appropriate compaction, appropriate compaction], the "excessive compaction" state only appears once consecutively, less than 3 times, so it is judged as a pseudo state and corrected to the previous state "appropriate compaction", and the corrected sequence is [appropriate compaction, appropriate compaction, appropriate compaction, appropriate compaction, appropriate compaction]. In the sequence [undervoltage, undervoltage, undervoltage], "undervoltage" appears three times consecutively, meeting the condition and being identified as a valid state segment. Subsequently, on the sequence after continuity detection correction, the transition frequency between adjacent state categories is calculated. In a sample dataset containing 1000 state transitions, the number of transitions from "appropriate voltage state" to "undervoltage state" is counted as 30 times; the number of transitions from "appropriate voltage state" to "overvoltage state" is counted as 10 times. Based on this, the transition frequency from "appropriate voltage" to "undervoltage" is calculated as 30 / 1000 = 0.03. The state-time mapping relationship is reconstructed based on the transition frequency distribution. This process sets an extremely low transition frequency threshold of 0.005. This threshold is set based on the analysis of massive historical data, which reveals that direct transitions between certain states almost never occur under real-world operating conditions, with a statistical frequency below 0.005. If a transition frequency below 0.005 is detected in the current construction sequence, such as a direct transition from "overpressure" to "underpressure," the subsequent state ("underpressure") is marked as an anomaly requiring further manual verification. Finally, all valid state segments processed through the above steps and their corresponding time information are integrated to generate the final construction state classification result. For example, the result might be expressed as: time 0 to 50 seconds, state "appropriate pressure"; time 51 to 85 seconds, state "underpressure."

[0037] Specifically, such as Figure 2 , 6 As shown, the status detection module includes: The time segmentation submodule acquires and segments the time series data corresponding to the construction status classification results, extracts the proportion of status categories and the duration of status within multiple time periods, analyzes the continuity of categories between adjacent time periods, calculates the number of status changes and the duration within multiple segments, and generates segment distribution data. The generated construction status classification result is obtained, which is a status sequence with time information. For example, within a construction segment with a total duration of 200 seconds, a status sequence consisting of 40 consecutive time periods (5 seconds each) is obtained: [Appropriate pressure, Appropriate pressure, Appropriate pressure, Low pressure, Low pressure, Low pressure, Low pressure, Appropriate pressure, Appropriate pressure, Overpressure, ..., Appropriate pressure]. First, this time series data is segmented. Specifically, time periods with consecutive identical states in the sequence are merged into a "segment". For example, the [Appropriate pressure, Appropriate pressure, Appropriate pressure] at the beginning of the sequence is merged into the first segment, with a state of "Appropriate pressure" and a duration of 3*5=15 seconds. The following [Low pressure, Low pressure, Low pressure, Low pressure] is merged into the second segment, with a state of "Low pressure" and a duration of 4*5=20 seconds. The entire status sequence is processed according to this rule. Subsequently, the percentage of status categories and the duration of each status are extracted within multiple time periods. Within a total duration of 200 seconds, by summing the durations of each segment, it was calculated that the "appropriate pressure" state occurred for a total of 95 seconds, the "underpressure" state for a total of 60 seconds, and the "overpressure" state for a total of 45 seconds. Based on this, the percentage of each state category was calculated as follows: "appropriate pressure" accounted for 95 / 200 = 47.5%, "underpressure" for 60 / 200 = 30%, and "overpressure" for 45 / 200 = 22.5%. Next, the continuity of categories between adjacent time periods was analyzed; this process was completed in the segmentation step, defining the boundaries of segments by identifying the points of change in state. Finally, the number of state changes and their duration within multiple segments were calculated. The number of state changes is the total number of segments obtained after segmentation minus one. If the entire 200-second construction period was divided into 11 segments, the number of state changes would be 11 - 1 = 10. The duration of each segment was also recorded simultaneously, and the generated data structure is shown below.

[0038]

[0039] As shown in Table 3, this table divides the continuous construction process into different sections according to the status category and records the detailed time information of each section, which is the generated section distribution data.

[0040] The State Trend submodule, based on the segment distribution data, calls the state category sequence of adjacent time periods, judges the direction of state change between adjacent time periods, calculates the rate of change and duration of change corresponding to the direction of change, analyzes the fluctuation amplitude value sequence, and obtains the state fluctuation amplitude sequence. Based on the generated segment distribution data, the state category sequence of adjacent time periods is first called. According to Table 3, this sequence is [Appropriate Compaction, Under-Compaction, Appropriate Compaction, Over-Compaction, Appropriate Compaction, Under-Compaction, Appropriate Compaction]. To determine the direction of state change between adjacent time periods, the non-numerical state categories need to be quantified first. The quantification standard is set as follows: "Appropriate Compaction" is used as the baseline state and assigned a value of 0; "Under-Compaction" indicates below the baseline and is assigned a value of -1; "Over-Compaction" indicates above the baseline and is assigned a value of +1. This quantification standard is established by analyzing the deviation relationship between each state and the ideal compaction target. Based on this standard, the state sequence is converted into a numerical sequence [0, -1, 0, 1, 0, -1, 0]. The direction of state change is determined by calculating the difference between adjacent values. For example, the change direction value from segment 1 to segment 2 is -1 - 0 = -1, indicating a negative change. The change value from segment 2 to segment 3 is 0 - (-1) = 1, indicating a positive change. Thus, the change direction sequence [-1, 1, 1, -1, -1, 1] is obtained. Next, the rate of change and duration corresponding to the direction of change are calculated. In this embodiment, the rate of change is directly represented by the quantized value (-1 or 1) of the direction of change, indicating the unit step of the change, while the duration is defined as the duration of the segment before the change occurs. For example, before the first change (direction -1), the segment is in segment 1 (appropriate pressure), and its duration is 15 seconds. Before the second change (direction 1), the segment is in segment 2 (underpressure), and its duration is 20 seconds. Finally, the fluctuation amplitude value sequence is analyzed. A sliding window with a width of 3 segments is set, and the range of the quantized values ​​of the state within the window (maximum value minus minimum value) is calculated as the fluctuation amplitude of the window. For the first window, segments 1, 2, and 3 are covered, and the state value is [0, -1, 0], with a fluctuation amplitude of 0 - (-1) = 1. For the second window, segments 2, 3, and 4 are covered, and the state value is [-1, 0, 1], with a fluctuation amplitude of 1 - (-1) = 2. By sliding the window in this way, a fluctuation amplitude value sequence is calculated. For the data in Table 3, the complete state fluctuation amplitude sequence is [1, 2, 2, 2, 1], which is the state fluctuation amplitude sequence.

[0041] The stability determination submodule compares the difference between the fluctuation amplitude and the fluctuation threshold over multiple time periods based on the state fluctuation amplitude sequence, determines the stability interval distribution of state changes within multiple segments, calculates the ratio of the proportion of stable segments to the proportion of fluctuating segments, and obtains the construction status detection results. Based on the generated state fluctuation amplitude sequence, [1, 2, 2, 2, 1], the differences between the fluctuation amplitudes over multiple time periods and the fluctuation threshold are first compared. The fluctuation threshold is set to 1.5. This threshold is based on an analysis experiment of historical state data from 100 completed road sections (50 rated as high quality and 50 as medium quality). The experiment involved calculating the state fluctuation amplitude sequence for all road sections and statistically analyzing its value distribution. The statistical results showed that 97% of the fluctuation amplitude values ​​in high-quality road sections were no greater than 1, while over 60% of the fluctuation amplitude values ​​in medium-quality road sections reached 2. Therefore, 1.5 was set as the critical value for distinguishing between stability and fluctuation. Subsequently, each value in the sequence was judged: if the fluctuation amplitude was greater than 1.5, the section was judged as a "fluctuating section"; if it was less than or equal to 1.5, it was a "stable section". For the sequence [1, 2, 2, 2, 1], the differences between it and the threshold 1.5 are -0.5, 0.5, 0.5, 0.5, and -0.5, respectively. Based on the sign of the difference, the sequence of stability interval distribution is determined as [stable segment, fluctuating segment, fluctuating segment, fluctuating segment, stable segment]. Next, the ratio of the proportion of stable segments to the proportion of fluctuating segments is calculated. In this sequence, "stable segment" appears twice, and "fluctuating segment" appears three times. The proportion of stable segments is 2 / 5 = 40%. The proportion of fluctuating segments is 3 / 5 = 60%. The ratio is calculated as 40% / 60%, resulting in 0.67. Finally, this ratio is compared with a stability benchmark value of 1.0 to obtain the construction status detection result. This benchmark value of 1.0 is also verified through the aforementioned experiment using historical data from hundreds of road sections. The experiment found that the ratio for high-quality road sections is greater than 1.0, while that for medium-quality road sections is less than 1.0. The advantage of this method is that by calculating the ratio of stable to fluctuating sections, the dynamic stability of the entire construction process is quantified into a single, clear indicator, surpassing the evaluation of isolated state points. Comparing the example result of 0.67 with the benchmark value of 1.0, since 0.67 < 1.0, the generated construction state detection result is "insufficient stability of the construction process." This result indicates that during the entire monitored construction process, periods of drastic state fluctuations dominated, and the overall control stability of the compaction process needs improvement.

[0042] Specifically, such as Figure 2 , 7 As shown, the optimization and correction module includes: The data correction submodule acquires the fluctuation interval monitoring data in the construction status detection results, calculates the time change rate for vibration frequency, hydraulic pressure and construction pavement compaction degree, performs linear translation and amplitude redistribution within the interval on the data points, and generates a set of correction parameters. The generated construction status monitoring results are obtained, and the original monitoring data identified as "fluctuating sections" are extracted from them. Taking a data point lasting 20 seconds (from the 55th to the 75th second of the total task time) identified as a "fluctuating section" as an example, the original vibration frequency sequence corresponding to this section is [35.1, 35.0, 32.5, 32.6] Hz, the hydraulic pressure sequence is [14.1, 14.0, 14.8, 14.9] MPa, and the pavement compaction sequence is [94.1, 94.0, 92.5, 92.6]%. First, the time change rate, i.e., the difference between adjacent data points, is calculated for these three sets of data. For the vibration frequency sequence, its time change rate sequence is [35.0-35.1, 32.5-35.0, 32.6-32.5], i.e., [-0.1, -2.5, 0.1]. Next, linear translation and amplitude redistribution within the interval are performed on the data points. The goal of this process is to correct abrupt changes caused by abnormal operating conditions within the fluctuating segment. During execution, the translation reference for the fluctuating segment is first determined. This reference is calculated from the average values ​​of the two immediately preceding and following stable segments. Referring to Table 4, the preceding segment (segment 3) of the fluctuating segment (segment 4) is in a "suitable pressure" state, with an average internal vibration frequency of 35.5 Hz; the following segment (segment 5) is also in a "suitable pressure" state, with an average internal vibration frequency of 35.2 Hz. The target average value for linear translation is set to the arithmetic mean of the average values ​​of these two stable segments, i.e., (35.5 + 35.2) / 2 = 35.35 Hz. The average frequency within the current fluctuating segment is (35.1 + 35.0 + 32.5 + 32.6) / 4 = 33.8 Hz. The linear translation amount is the difference between the target average value and the current average value, i.e., 35.35 - 33.8 = 1.55 Hz. This shift is added to each data point in the fluctuating range, resulting in the shifted sequence: [36.65, 36.55, 34.05, 34.15]. Amplitude redistribution is then performed, a step that uses a redistribution coefficient to compress the deviation of the data points from the new mean. The redistribution coefficient is set experimentally: 100 data segments containing known equipment anomalies (such as engine stall precursors) were analyzed, testing redistribution coefficients ranging from 0.1 to 0.9. It was found that when the coefficient was 0.4, the corrected data distribution most closely resembled the data distribution under subsequent stable operating conditions. Therefore, the redistribution coefficient was set to 0.4. The amplitude redistribution is calculated as follows: New data point = New mean + (Original data point - Original mean) * 0.4. For example, for the first point 35.1, the formula is 35.35 + (35.1 - 33.8) * 0.4 = 35.35 + 0.52 = 35.87. Perform this operation on the entire sequence to obtain the final corrected vibration frequency sequence. Use the exact same steps to calculate the hydraulic pressure and pavement compaction sequences, ultimately generating a set of correction parameters composed of the three corrected sequences.

[0043] The weighted smoothing submodule assigns weights to the multi-parameter sequence based on the calibration parameter set. It analyzes the weight coefficients according to the ratio of vibration frequency change rate, hydraulic pressure difference value and compaction deviation, and calls the average weight of adjacent time series to perform sliding smoothing on the parameters to generate a smoothed weighted sequence set. Based on the generated set of correction parameters, weights are assigned to the multi-parameter sequences. Taking the corrected vibration frequency sequence as an example, a sequence with 5 time steps is [35.87, 35.82, 34.90, 34.94, 35.41] Hz. The weighting coefficients are assigned based on three indicators: the rate of change of vibration frequency, the difference in hydraulic pressure, and the proportion of compaction deviation. First, the values ​​of each indicator are calculated. The rate of change of vibration frequency is the absolute value of the difference between the current point and the previous point; for example, the rate of change for the second point is |35.82-35.87|=0.05. The difference in hydraulic pressure is the absolute value of the difference between the hydraulic pressure value at the current point and the set target pressure value. The target pressure value is set to 14.0 MPa, which is derived from the optimal working pressure calibrated at the equipment's factory and statistically based on a large amount of excellent working condition data. If the pressure value corresponding to the current point is 14.2 MPa, then the difference value is |14.2-14.0|=0.2. The compaction deviation ratio is calculated by dividing the absolute value of the difference between the target compaction degree and the current compaction degree by the target compaction degree. The target compaction degree is set to 95% according to the construction specifications. If the current compaction degree is 94%, the deviation ratio is |95-94| / 95 = 0.0105. Subsequently, weighting coefficients are assigned to these three indicators, which are 0.2, 0.5, and 0.3, respectively. These coefficient values ​​are determined based on the following experiment: 30 construction data segments containing known mixed interference (sensor noise, operational fluctuations) are selected, and different combinations of weighting coefficients (step size 0.1) are tested using a grid search method. The goal is to determine the coefficient values ​​based on the highest correlation between the smoothed data and the actual core sample test results. For each time point, the comprehensive weight value is calculated as (0.2 * vibration frequency change rate) + (0.5 * hydraulic pressure difference value) + (0.3 * compaction deviation ratio). For example, if the three indicator values ​​for a certain point are 0.05, 0.2, and 0.0105, then its comprehensive weight value is (0.2*0.05)+(0.5*0.2)+(0.3*0.0105)=0.01+0.1+0.00315=0.11315. Finally, the parameters are smoothed by using the weighted average of adjacent time series. This process uses a weighted moving average with a window width of 3. For example, to smooth the third data point, the calculation method is (point 2*weight 2 + point 3*weight 3 + point 4*weight 4) / (weight 2 + weight 3 + weight 4). The calculated comprehensive weight values ​​and corresponding parameter values ​​for all time points are substituted into this smoothing calculation. This operation is performed for each sequence in the calibration parameter set to generate a smoothed weighted sequence set.

[0044] The anomaly correction submodule calculates the difference rate between multiple sequence points and the mean of the same interval based on the smooth weighted sequence set, compares the difference rate with the anomaly threshold, filters out anomaly points and replaces them with the mean within the interval, and generates the construction status detection optimization result set through difference rate correction and reconstruction of the parameter sequence. The abnormal threshold is set by statistically analyzing the distribution characteristics of the difference rate of multiple parameters in the construction monitoring data; Based on the generated smoothed weighted sequence set, residual single-point mutations in the sequence are corrected. Taking a smoothed compaction sequence as an example, the values ​​in a certain segment are [94.5, 94.6, 95.8, 94.7, 94.5]%. First, the difference rate between multiple sequence points and the mean of the same interval is calculated. The mean of this interval is (94.5+94.6+95.8+94.7+94.5) / 5=94.82%. For each point in the sequence, its difference rate with the mean is calculated as |data point - interval mean| / interval mean. For the third data point 95.8, the difference rate is calculated as |95.8-94.82| / 94.82=0.98 / 94.82, resulting in 0.0103, or 1.03%. Next, the calculated difference rate is compared with the anomaly threshold. The anomaly threshold is set to 1%. The threshold setting is based on an experiment with 500 samples, all from a stable construction phase, with artificially injected single-point large-amplitude noise. After calculating the difference rate of these sample data, it was found that 99.5% of the normal data points had a difference rate below 1%, while the difference rate of all injected noise points was above 1%. Therefore, 1% was chosen as the boundary distinguishing normal fluctuations from outliers. In the aforementioned example, the difference rate of the third data point, 1.03%, is greater than the outlier threshold of 1%, so this data point, 95.8, was selected as an outlier. The selected outliers will be replaced with the mean of that interval. Therefore, data point 95.8 was replaced with 94.82. The corrected sequence becomes [94.5, 94.6, 94.82, 94.7, 94.5]. Finally, through difference rate correction and reconstruction of the parameter sequence, an optimized result set for construction status detection is generated. This step performs the complete process of "calculating the difference rate, comparing thresholds, filtering and replacing" on the three smoothed weighted sequence sets of vibration frequency, hydraulic pressure, and pavement compaction degree, and integrates all data sequences after outlier correction. The integrated dataset is the construction condition monitoring optimization result set, which contains three high-quality construction parameter time series that have undergone multiple corrections and adjustments.

[0045] The above are merely specific embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A road construction status monitoring system based on dynamic data monitoring, characterized in that, The system includes: The information fusion module collects vibration frequency, hydraulic pressure, construction speed and road surface compaction of multiple construction equipment, performs time-series alignment and weighted average analysis on all data to generate construction status monitoring dataset and transmits it to the feature extraction module. The feature extraction module analyzes equipment and road surface features based on the construction status monitoring dataset, calculates the covariance relationship between the vibration frequency of the construction equipment, hydraulic pressure, construction speed and compaction degree of the construction road surface, extracts key dynamic features and normalizes them, generates a construction dynamic feature set and transmits it to the status classification module. The status classification module classifies the construction status based on the construction dynamic feature set, calculates the conditional probability distribution between equipment and road surface features, determines the construction status category for multiple time periods, generates the construction status classification result, and transmits it to the status detection module. The status detection module analyzes the status changes in the time series based on the construction status classification results, calculates the status change trend of adjacent time periods and extracts the fluctuation amplitude and segment, judges the stability of the construction status based on the fluctuation amplitude and fluctuation threshold, and forms the construction status detection results.

2. The road construction status monitoring system based on dynamic data monitoring according to claim 1, characterized in that: The construction status monitoring dataset includes time-aligned vibration frequency data, hydraulic pressure data, and construction pavement compaction data. The construction dynamic feature set includes the covariance of equipment vibration frequency and hydraulic pressure, key dynamic feature vectors, and normalized feature scales. The construction status classification results include a multi-time period construction status category sequence, conditional probability distribution, and time period classification results. The construction status detection results include status change trends, fluctuation amplitudes, and stability judgment indicators.

3. The road construction status monitoring system based on dynamic data monitoring according to claim 1, characterized in that, The information fusion module includes: The data analysis submodule acquires vibration frequency, hydraulic pressure, construction speed and road surface compaction data of multiple construction equipment, extracts the main frequency component from the vibration signal through frequency domain decomposition, extracts transient pressure from the hydraulic pressure, and synchronizes it with the compaction data in time to generate a multi-source construction raw dataset. The time-series alignment submodule calculates the time difference between sampling timestamps of multiple types of sensor data based on the multi-source construction original dataset, performs interpolation and resampling operations on multiple data sequences, reconstructs continuous time series according to a unified time step, calculates the mean square of synchronization error, and generates a multi-source synchronized time series dataset. The weighted analysis submodule calculates the response amplitudes of vibration frequency, hydraulic pressure, construction speed, and road surface compaction at multiple time steps based on the multi-source synchronous time series dataset, analyzes the weight ratios, combines multiple weighted terms into construction response quantities, and smooths them by moving average to generate a construction status monitoring dataset.

4. The road construction status monitoring system based on dynamic data monitoring according to claim 1, characterized in that, The feature extraction module includes: The covariance calculation submodule, based on the construction status monitoring dataset, extracts vibration frequency, hydraulic pressure, construction speed and road surface compaction for time synchronization, and calculates the covariance relationship between multiple time steps and integrates them with weights to generate a set of construction feature covariance values. The feature extraction submodule calls the construction feature covariance value group, performs difference analysis on the multi-covariance value sequence, extracts the rate of change and fluctuation amplitude within the time period, filters the main response components based on the amplitude change interval, and generates the key dynamic feature sequence of construction. The feature normalization submodule performs interval linear transformation on each component based on the key construction dynamic feature sequence, mapping the values ​​to a unified scale range and generating a construction dynamic feature set.

5. The road construction status monitoring system based on dynamic data monitoring according to claim 1, characterized in that, The state classification module includes: The conditional distribution submodule, based on the construction dynamic feature set, extracts equipment operation feature components and road surface response feature components and performs time step pairing, calculates the joint occurrence frequency of the two types of components at the same time step, analyzes the conditional probability by combining the occurrence frequency of single components, and generates a conditional probability distribution set of equipment and road surface. The state determination submodule calls the device and road surface condition probability distribution set, calculates and divides the probability fluctuation interval based on the amplitude of condition probability change over a continuous time period, compares the condition probability intervals corresponding to multiple time periods with the classification threshold, determines the construction state category of multiple intervals, and generates a multi-time period construction state sequence. The status result submodule performs continuity detection on the status categories of the multi-time period construction status sequence, calculates the transition frequency value between adjacent status categories, reconstructs the status time mapping relationship based on the transition frequency distribution, and generates construction status classification results.

6. The road construction status monitoring system based on dynamic data monitoring according to claim 5, characterized in that, The classification threshold is calculated by using statistical equipment and the range of conditional probability values ​​corresponding to multiple construction state categories in the road surface conditional probability distribution set. The mean and variance of the conditional probabilities under multiple state categories are calculated, and the degree of difference in the mean is analyzed.

7. The road construction status monitoring system based on dynamic data monitoring according to claim 1, characterized in that, The status detection module includes: The time segmentation submodule acquires and segments the time series data corresponding to the construction status classification results, extracts the proportion of status categories and the duration of status within multiple time periods, analyzes the continuity of categories in adjacent time periods, calculates the number of status changes and the duration within multiple segments, and generates segment distribution data. The state trend submodule, based on the segment distribution data, calls the state category sequence of adjacent time periods, judges the direction of state change between adjacent time periods, calculates the rate of change and duration of change corresponding to the direction of change, analyzes the fluctuation amplitude value sequence, and obtains the state fluctuation amplitude sequence. The stability determination submodule compares the difference between the fluctuation amplitude and the fluctuation threshold over multiple time periods based on the state fluctuation amplitude sequence, determines the stability interval distribution of state changes within multiple segments, calculates the ratio of the proportion of stable segments to the proportion of fluctuating segments, and obtains the construction state detection results.

8. The road construction status monitoring system based on dynamic data monitoring according to claim 7, characterized in that, The fluctuation threshold is obtained by extracting the fluctuation amplitude distribution of multiple state categories in the time series, calculating the mean and standard deviation of the fluctuation amplitude, and then analyzing the difference between the mean and standard deviation.

9. The road construction status monitoring system based on dynamic data monitoring according to claim 1, characterized in that, The system also includes: The optimization and correction module, based on the construction status detection results, recalibrates the vibration frequency, hydraulic pressure, and road surface compaction degree for the fluctuation range, and performs weighted smoothing analysis and anomaly correction to generate an optimized set of construction status detection results. The optimized result set for construction status detection includes a set of correction parameters, a set of smooth weighted sequences, and anomaly correction records.

10. The road construction status monitoring system based on dynamic data monitoring according to claim 9, characterized in that, The optimization and correction module includes: The data correction submodule acquires the fluctuation interval monitoring data in the construction status detection results, calculates the time change rate for vibration frequency, hydraulic pressure and construction pavement compaction degree, performs linear translation and amplitude redistribution within the interval on the data points, and generates a set of correction parameters. The weighted smoothing submodule, based on the correction parameter set, assigns weights to the multi-parameter sequence, analyzes the weight coefficients according to the ratio of vibration frequency change rate, hydraulic pressure difference value and compaction deviation, calls the average weight of adjacent time series to perform sliding smoothing on the parameters, and generates a smoothed weighted sequence set; The anomaly correction submodule calculates the difference rate between multiple sequence points and the mean of the same interval based on the smooth weighted sequence set, compares the difference rate with the anomaly threshold, filters out anomaly points and replaces them with the mean within the interval, and generates a construction status detection optimization result set through difference rate correction and reconstructed parameter sequence recombination. The abnormal threshold is set by statistically analyzing the distribution characteristics of the difference rate of multiple parameters in the construction monitoring data.

Images