Method and system for automatically measuring compactness

Abstract

The invention discloses an automatic compaction degree measuring method and system, and relates to the field of compaction quality monitoring, and the method comprises the steps: initializing and constructing parameter configuration, outputting configuration data, constructing a multi-mode excitation signal library and a self-adaptive scheduling strategy, and outputting an excitation mode code and a digital waveform sequence. Synchronously acquiring multi-source sensor data based on excitation mode coding and a digital waveform sequence, and outputting a multi-source sensor data frame; performing feature extraction on the multi-source sensor data frame to obtain a broad-spectrum dynamic fingerprint vector and a mesoscopic surface topography feature vector, and combining the broad-spectrum dynamic fingerprint vector and the mesoscopic surface topography feature vector into a feature triple data packet; and performing mode matching and similarity evaluation on the feature triple data packet and a pre-stored compaction state feature template library, and outputting a real-time compaction degree value with confidence rating through a searching and mapping mechanism. According to the method, high-robustness and high-adaptability compaction degree evaluation independent of an accurate physical model is realized, and meanwhile, a quantitative index of result reliability is provided.

Description

Technical Field

[0001] This invention relates to the field of compaction quality monitoring, and in particular to an automatic method and system for measuring compaction degree. Background Technology

[0002] In the field of compaction quality monitoring, continuous measurement technology based on onboard sensors of vibratory rollers has become the mainstream development direction to replace traditional point detection methods. By collecting the dynamic response signal of the vibratory roller, combined with a simplified physical model or empirical regression model, the compaction degree of the soil can be indirectly evaluated, and the real-time monitoring of the construction process has been initially realized. Existing technologies usually adopt a single excitation mode with fixed frequency and amplitude, and mainly rely on single-type sensor signals such as acceleration for analysis. Its system structure is relatively simple and easy to deploy in engineering.

[0003] In practical applications, the measurement accuracy and robustness are challenged. The limitation lies in the passive excitation method and the single perception dimension. Fixed excitation signals are difficult to fully stimulate the dynamic response characteristics of soils with different soil types, moisture contents and compaction stages, resulting in a limited amount of collected signal information. At the end of compaction, the signal changes are weak and the discrimination decreases. The analysis model based on a single-dimensional signal is easily affected by various field factors such as changes in rolling speed, mechanical state drift and heterogeneity of the underlying soil, which makes it difficult to calibrate model parameters and has insufficient adaptability. The stability and reliability of the measurement results are difficult to guarantee under complex and variable working conditions. Summary of the Invention

[0004] In view of the aforementioned existing problems, the present invention is proposed.

[0005] Therefore, the present invention provides an automatic method for measuring compaction degree to solve the problem of insufficient measurement accuracy under complex working conditions caused by the fixed excitation mode and single sensing information in the prior art.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution: In a first aspect, the present invention provides an automatic method for measuring compaction degree, which includes initialization and construction parameter configuration, outputting configuration data, constructing a multi-mode excitation signal library and an adaptive scheduling strategy, and outputting excitation mode encoding and digital waveform sequence. Multi-source sensor data is acquired synchronously based on excitation mode encoding and digital waveform sequence, and multi-source sensor data frames are output. Feature extraction is performed on multi-source sensor data frames to obtain a broad-spectrum dynamic fingerprint vector and a mesoscopic surface morphology feature vector, which are combined into a feature triplet data packet. The feature triplet data package is matched and similarity is evaluated with a pre-stored compaction state feature template library. The real-time compaction degree value with confidence rating is output through a lookup and mapping mechanism. Based on real-time compaction values ​​with confidence ratings and spatial location information, the compaction field and construction guidance information are output.

[0007] As a preferred embodiment of the automatic compaction degree measurement method of the present invention, the initialization and construction parameter configuration, and the output of configuration data, include the following steps: Select a soil type from the preset list, obtain the soil type code, input the value, obtain the compaction value, delineate the boundary on the map interface, and obtain the construction boundary coordinates; Based on the soil type code, the maximum dry density value, optimum moisture content range, plasticity index reference value, and typical compression modulus curve parameters of the standard compaction test are retrieved from the database. Create a file with a specified name, obtain a pointer to the data record file, establish a connection with the cloud server, and obtain the wireless communication link status; It integrates soil type code, compaction degree value, construction boundary coordinates, maximum dry density value of standard compaction test, optimum moisture content range, plasticity index reference value, typical compression modulus curve parameters, data log file pointer, and wireless communication link status, and outputs configuration data.

[0008] As a preferred embodiment of the automatic measurement method for compaction degree described in this invention, the method includes the following steps: constructing a multi-mode excitation signal library and an adaptive scheduling strategy, and outputting excitation mode codes and digital waveform sequences: Based on the soil type code, the multi-mode excitation signal library pre-stored for the soil type is called from the non-volatile memory; The multi-mode excitation signal library contains preset digital waveform sequences, linear sweep frequency signals, pseudo-random frequency conversion signals, multi-frequency composite signals, and single-pulse impulse signals. Real-time working parameters are obtained by acquiring instantaneous rolling speed, current number of rolling passes, compaction degree change gradient, and GNSS location information; The adaptive scheduling strategy analyzes real-time operating parameters based on fuzzy decision rules, selects the optimal excitation mode from a multi-mode excitation signal library, and generates an excitation mode code corresponding to the selected optimal excitation mode. Extract the digital waveform sequence corresponding to the excitation mode encoding from the multi-mode excitation signal library.

[0009] As a preferred embodiment of the automatic measurement method for compaction degree described in this invention, the method includes the following steps: synchronously acquiring multi-source sensor data based on excitation mode encoding and digital waveform sequence, and outputting multi-source sensor data frames: The system receives the excitation mode encoding and digital waveform sequence, and generates a unified microsecond-level precision timestamp based on the start time of the digital waveform sequence. The excitation mode is encoded with a unified timestamp and sent to the acquisition controller of the triaxial accelerometer, rotary encoder, pressure sensor array, near-infrared spectral moisture sensor and high-precision GNSS receiver. The triaxial accelerometer collects simulated acceleration signals in the X, Y, and Z directions; the pressure sensor array collects simulated pressure distribution signals on the contact surface between the steel wheel and the soil; and the near-infrared spectral moisture sensor collects spectral reflectance signals from the soil surface. The excitation mode encoding, unified timestamp, acceleration simulation signal, pressure distribution simulation signal, spectral reflection signal, three-dimensional coordinates, velocity, and heading are packaged into a multi-source sensor data frame.

[0010] As a preferred embodiment of the automatic measurement method for compaction degree described in this invention, the method includes the following steps: feature extraction of multi-source sensor data frames to obtain a broad-spectrum dynamic fingerprint vector and a mesoscopic surface morphology feature vector. Acceleration simulation signal, angular velocity sequence, pressure distribution simulation signal and spectral reflectance signal were extracted from multi-source sensor data frames; The vertical component of the acceleration simulation signal is zero-mean and five-point cubic smoothed to obtain a preprocessed vertical acceleration sequence. A 256-point fast Fourier transform is then performed to obtain the spectrum. Extract the frequency domain features of dominant frequency value, dominant band energy, frequency weighting center, spectral variance, and approximate entropy from the spectrum; Using angular velocity sequences and acceleration simulation signals, the wheel-soil contact force is estimated through a single-degree-of-freedom dynamic model, and the force-displacement hysteresis curve is plotted in combination with displacement. Geometric features such as hysteresis loop area, average slope, and shape fullness coefficient are extracted from the force-displacement hysteresis curve. The frequency domain features and geometric features are arranged in a fixed order to form a broad spectrum dynamic fingerprint vector. Threshold segmentation and connected component analysis are performed on the pressure distribution simulation signal to identify the effective contact area and calculate the morphological features of average grounding pressure, pressure non-uniformity coefficient, and pressure center offset. The morphological features are arranged in a fixed order to form a mesoscopic surface morphological feature vector.

[0011] As a preferred embodiment of the automatic measurement method for compaction degree described in this invention, the method comprises: combining data into a feature triplet data packet, including the following steps: The spectral reflectance signal is processed to retrieve the water content value. The broad-spectrum dynamic fingerprint vector, mesoscopic surface morphology feature vector, and water content value are aligned by timestamp and combined into a feature triplet data packet.

[0012] As a preferred embodiment of the automatic measurement method for compaction degree described in this invention, the method includes the following steps: performing pattern matching and similarity evaluation between the feature triplet data package and a pre-stored compaction state feature template library: Load the pre-stored compaction state feature template library from local storage, and perform an overall comparison between the feature triplet data package and each feature triplet template in the pre-stored compaction state feature template library; Based on the comparison results, a similarity score is generated for each feature triplet template in the pre-stored compaction state feature template library, reflecting the degree of similarity with the feature triplet data package. Based on the similarity scores, the templates are sorted from high to low to generate a template similarity ranking list. According to the preset selection quantity, several feature triplet templates with the highest similarity scores are selected from the top of the template similarity ranking list to form a set of highly similar templates.

[0013] As a preferred embodiment of the automatic compaction degree measurement method of the present invention, the method includes the following steps: outputting a real-time compaction degree value with a confidence rating through a lookup and mapping mechanism: From each feature triplet template in the highly similar template set, find the associated true values ​​of the measured compaction of the sand cone method obtained during calibration, and assign weight factors to the true values ​​of the measured compaction of the sand cone method. Based on the preset mapping rules, the true values ​​of the measured compaction degree of the sand cone method associated with the feature triplet template and the corresponding weighting factors are combined to generate the compaction degree estimate. Based on the overall characteristics of highly similar template sets, a reliability index is generated, which is converted into a confidence rating value. The compaction estimate is combined with the confidence rating, and the output is a real-time compaction value with a confidence rating.

[0014] As a preferred embodiment of the automatic compaction degree measurement method of the present invention, the method includes the following steps: Based on the real-time compaction degree value with confidence rating and spatial location information, outputting compaction degree field and construction guidance information. The real-time compaction values ​​with confidence ratings and associated spatial location information are updated into a gridded digital map in memory. The compaction estimates and confidence ratings are then used to interpolate and smooth each grid cell in the digital map. The compaction field data layer is rendered as a compaction cloud map represented by color depth and confidence level. The compaction field data layer is scanned to identify all continuous grid areas where the compaction estimate is lower than the preset target and marked as weak compaction areas. Based on the spatial distribution and size of the weak compaction areas, as well as the real-time spatial location and attitude of the roller, a recommended compaction path covering all weak compaction areas is planned, forming construction guidance information.

[0015] Secondly, the present invention provides an automatic measurement system for compaction degree, including an initialization module, initialization and construction parameter configuration, output configuration data, construction of a multi-mode excitation signal library and adaptive scheduling strategy, and output of excitation mode encoding and digital waveform sequence. The acquisition module synchronously acquires multi-source sensor data based on excitation mode encoding and digital waveform sequence, and outputs multi-source sensor data frames. The fusion module extracts features from multi-source sensor data frames to obtain a broad-spectrum dynamic fingerprint vector and a mesoscopic surface morphology feature vector, which are then combined into a feature triplet data packet. The evaluation module performs pattern matching and similarity assessment between the feature triplet data package and the pre-stored compaction state feature template library, and outputs real-time compaction values ​​with confidence ratings through a lookup and mapping mechanism. The decision output module outputs compaction field and construction guidance information based on real-time compaction values ​​with confidence ratings and spatial location information.

[0016] Thirdly, the present invention provides a computer device including a memory and a processor, wherein the memory stores a computer program, wherein when the computer program is executed by the processor, it implements any step of the automatic measurement method for compaction as described in the first aspect of the present invention.

[0017] Fourthly, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, it implements any step of the automatic measurement method for compaction as described in the first aspect of the present invention.

[0018] The beneficial effects of this invention are as follows: By transforming the road roller vibrator from a fixed vibration source into an active sensing tool that can dynamically output the optimal detection signal according to the working conditions, adaptive matching between the excitation mode and the soil state and construction stage is achieved, thereby enriching the excitation dimensions and enhancing the information content and relevance of the response signal. By performing pattern matching and similarity evaluation between the feature triplet data package and the pre-stored compaction state feature template library, and outputting real-time compaction values ​​with confidence ratings through a search and mapping mechanism, the invention adopts a pattern matching and direct mapping mechanism based on an instance library, replacing the traditional complex model inversion. By comparing the real-time extracted multi-dimensional feature vectors with the historical calibration template library, the compaction estimate is found from the most similar known state instances. A confidence rating is generated based on the consistency of the matching results, thereby bypassing the calibration difficulties of complex models and the sensitivity to interference factors. This achieves highly robust and highly adaptable compaction assessment that does not rely on an accurate physical model, and at the same time provides a quantitative indicator of the reliability of the results. Attached Figure Description

[0019] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the 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.

[0020] Figure 1 This is a flowchart of an automatic method for measuring compaction degree.

[0021] Figure 2 This is a schematic diagram of an automatic compaction measurement system. Detailed Implementation

[0022] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0023] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0024] Secondly, the term "one embodiment" or "example" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the invention. The appearance of an embodiment in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that mutually excludes other embodiments.

[0025] Reference Figures 1-2 As one embodiment of the present invention, this embodiment provides an automatic method for measuring compaction degree, comprising the following steps: S1. Initialize and configure construction parameters, and output configuration data.

[0026] S1.1 Select the soil type from the preset list, obtain the soil type code, input the value, obtain the compaction value, delineate the boundary on the map interface, and obtain the construction boundary coordinates.

[0027] Furthermore, the complex characteristics of the on-site soil are abstracted into a standardized soil type code that can be directly recognized and invoked by computer programs. Simultaneously, two key engineering objectives and spatial constraints—compaction degree values ​​and construction boundary coordinates—are integrated to construct a digital construction task framework. This allows all subsequent automated processing flows to operate within a clear and structured context. Now, soil type codes are determined through selection rather than input, avoiding potential input errors by operators unfamiliar with specialized geotechnical classifications, ensuring the authority and consistency of parameters. The abstract compaction degree target is bound to specific, visualized construction boundary coordinates on a digital map, creating a strong correlation between quality objectives and spatial scope. Specifically, for example, on the vehicle-mounted human-machine interface, the operator selects from the list of graphical icons for clay and sand, and the software automatically maps them to the corresponding soil type codes; on the same map view, after entering the target compaction value, the operator can directly outline the work area with their finger or stylus, and the software automatically converts the outlined polygon into a series of construction boundary coordinates.

[0028] S1.2. Based on the soil type code, retrieve from the database the maximum dry density value, optimum moisture content range, plasticity index reference value, and typical compression modulus curve parameters of the standard compaction test.

[0029] Furthermore, the intrinsic parameters characterizing the compaction and mechanical properties of this type of soil, obtained from indoor geotechnical tests, are automatically and dynamically linked to the on-site construction process. A structured parameter database is pre-set, in which each soil type code is associated with a specific set of verified geotechnical property parameters, including the maximum dry density value of the standard compaction test, the optimum moisture content range, the plasticity index reference value, and typical compression modulus curve parameters. This avoids the need to manually consult paper reports or re-enter these highly specialized parameters during each construction phase, improving efficiency and eliminating the risk of data inconsistency caused by human memory or transcription errors.

[0030] Specifically, for example, when the soil type code is determined to be a specific code representing silty clay, this step will automatically retrieve from the database the maximum dry density value of the standard compaction test corresponding to the peak value of the standard compaction curve of that type of soil, its corresponding optimum moisture content range, the reference value of the plasticity index describing its plasticity, and the typical compression modulus curve parameters reflecting its stress-strain relationship, ensuring the soil adaptability of the measurement method.

[0031] S1.3 Create a file with a specified format name, obtain a pointer to the data record file, establish a connection with the cloud server, and obtain the wireless communication link status.

[0032] Furthermore, by creating localized data storage anchors, data record file pointers, and establishing remote communication channels and wireless communication link status, the system achieves data lifecycle synchronization and cloud interconnection. By creating files with specific naming formats and returning a data record file pointer for subsequent data write operations, a unique and ordered storage space and access handle are pre-allocated to the massive amounts of time-series data, ensuring the continuity and immutability of data records. It also proactively establishes a connection with the cloud server and verifies the wireless communication link status, paving the way for real-time or near-real-time uploading, remote monitoring, centralized analysis, and archiving of measurement data. Specifically, for example, the software generates a unique filename such as project number year month day hour minute second.log based on the current time and creates the file. It obtains a programmable interface pointing to the file, i.e., a data log file pointer. It establishes a secure connection with the specified cloud service address through the built-in cellular network module and returns the wireless communication link status of successful connection or quality. This provides basic support for compaction measurement to move from single-machine operation to Internet of Things-based, traceable digital construction management.

[0033] S1.4 integrates soil type code, compaction degree value, construction boundary coordinates, maximum dry density value of standard compaction test, optimum moisture content range, plasticity index reference value, typical compression modulus curve parameters, data log file pointer, and wireless communication link status, and outputs configuration data.

[0034] Furthermore, through structured data encapsulation operations, a unified and self-consistent configuration data object is integrated. A fixed data integration structure and output specification are defined, and nine types of information are forcibly packaged, including soil type code, compaction degree value, construction boundary coordinates, maximum dry density value of standard compaction test, optimum moisture content range, plasticity index reference value, typical compression modulus curve parameters, data record file pointer, and wireless communication link status. Encapsulation ensures the integrity and consistency of configuration information. Any subsequent steps that need to reference construction parameters, soil characteristics, or data services only need to extract them from this configuration data object as needed, avoiding version confusion or loss that may occur when parameters are transferred between multiple functional units.

[0035] Specifically, for example, within the software, a configuration data structure or object conforming to a predetermined format is created, and the above nine fields are assigned values ​​one by one. Among them, the soil type code comes from user selection, the compaction degree value comes from user input, and the maximum dry density value of the standard compaction test comes from the database query results. The data record file pointer and the wireless communication link status come from the successful return of the initialization operation. The configuration data package contains comprehensive information such as target, attributes, space, benchmark, and service, which serves as the complete input conditions for starting the entire automatic compaction degree measurement process.

[0036] S2. Construct a multi-mode excitation signal library and an adaptive scheduling strategy, and output excitation mode codes and digital waveform sequences.

[0037] S2.1. Based on the soil type code, call the multi-mode excitation signal library pre-stored for the soil type from the non-volatile memory.

[0038] Furthermore, by encoding and associating customized multi-mode excitation signal libraries for different soil types, a preliminary match between excitation and soil type is achieved. Non-volatile memory is used as a knowledge carrier to pre-store signal libraries that have been analyzed and optimized offline and are specifically configured for the dynamic response characteristics of different soil types.

[0039] Specifically, for cohesive soils, the pre-stored signal library may contain more low-frequency energy to excite their viscous response; for granular soils, more mid-to-high-frequency components may be configured to excite inter-particle collisions and rearrangements, so that the excitation from the source is targeted rather than blindly applied, improving the pertinence and effectiveness of the initial excitation, and providing a high-quality set of alternative signals that conforms to the soil background for subsequent adaptive scheduling. This step completes the accurate call from general soil classification to the dedicated excitation knowledge base.

[0040] S2.2 The multi-mode excitation signal library contains preset digital waveform sequences, linear sweep frequency signals, pseudo-random frequency conversion signals, multi-frequency composite signals, and single-pulse impulse signals.

[0041] Furthermore, linear sweep frequency signals, pseudo-random frequency conversion signals, multi-frequency composite signals, and single-pulse impact signals each play different diagnostic roles, providing a complete excitation toolbox to address the detection needs at different stages of the compaction process. The signal types are differentiated as follows: linear sweep frequency signals are used for wideband, continuous stiffness spectrum measurements; pseudo-random frequency conversion signals have excitation characteristics similar to white noise and can quickly excite multiple modes of the system; multi-frequency composite signals can simultaneously detect the response of multiple characteristic frequency points with high efficiency; and single-pulse impact signals can provide rich instantaneous frequency response information, making them suitable for rapid diagnosis.

[0042] Specifically, pseudo-random frequency conversion signals can be used for rapid scanning during compaction quality surveys, while linear frequency sweep signals are used for fine monitoring of stiffness changes in specific frequency bands. This diversity provides a wealth of optional actions for adaptive scheduling strategies, freeing the excitation from a single, fixed mode and transforming it into a detection method that can be flexibly combined according to the target. This expands the information dimension and detection capability of the excitation, laying the physical foundation for a comprehensive and accurate perception of soil conditions. This step defines and provides the material basis for the execution of the excitation strategy.

[0043] S2.3 Obtain real-time working parameters by acquiring instantaneous compaction speed, current number of compaction passes, compaction degree change gradient, and GNSS location information.

[0044] Furthermore, four key parameters—instantaneous compaction speed, current number of compaction passes, compaction gradient, and GNSS location information—are defined and integrated to form a set of real-time working condition parameters describing the real-time construction status. This provides multi-dimensional and dynamically updated environmental awareness inputs for adaptive scheduling strategies, covering time, space, process history, and state change trends: instantaneous compaction speed affects load application time and dynamic response frequency; the current number of compaction passes identifies the process stage; the compaction gradient reflects the real-time evolution rate of compaction effect; and GNSS location information can determine whether the location is in a special spatial position such as a curve, joint, or already compacted area.

[0045] Specifically, for example, by acquiring speed through encoders, obtaining the number of passes through trajectory point statistics, obtaining the change gradient through short-term historical compaction data, and obtaining location through positioning receivers, this information can be aggregated to construct a comprehensive digital snapshot of when, where, what stage of the process is being carried out, and how the effects are changing. This makes subsequent decisions no longer blind or fixed, but based on a real-time understanding of the construction process and environmental context, providing the necessary perception conditions for intelligent incentive scheduling, and realizing multi-dimensional perception and parameterized description of the real-time construction status.

[0046] S2.4 The adaptive scheduling strategy analyzes real-time operating parameters based on fuzzy decision rules, selects the optimal excitation mode from the multi-mode excitation signal library, and generates an excitation mode code corresponding to the selected optimal excitation mode.

[0047] Furthermore, analyzing real-time operating parameters and selecting the optimal excitation mode from a multi-mode excitation signal library aims to simulate the decision-making logic of experienced operators. This allows the selection of excitation modes to intelligently adapt to complex and ever-changing on-site conditions. Fuzzy decision rules are used to handle the numerous unclear state judgments that exist during the compaction process. The following rules are defined: if the current number of compaction passes is small and the compaction degree change gradient is large, then a pseudo-random frequency conversion signal is selected first for rapid survey; if the compaction degree change gradient becomes very small, then a multi-frequency composite signal is switched to enhance the sensitivity to slight stiffness changes; if GNSS location information indicates that the area has entered a curve region, then a single-pulse impact signal is temporarily called for rapid state assessment.

[0048] Specifically, by combining the fuzzy processing of real-time operating parameters with such expert experience rules, the adaptive scheduling strategy can output a clear optimal excitation mode selection and generate a unique corresponding excitation mode code. This overcomes the shortcomings of fixed excitation modes, which cannot adapt to changes in process stages and special operating conditions. It makes the excitation action itself an intelligent and adaptive component of the construction process, achieving the best match between the excitation mode and the construction state. Thus, at any time, the excitation signal most likely to reveal the key state information of the current soil can be applied, completing intelligent excitation decision-making based on real-time operating conditions.

[0049] S2.5 Extract the digital waveform sequence corresponding to the excitation mode encoding from the multi-mode excitation signal library.

[0050] Furthermore, based on the excitation mode code, the corresponding digital waveform sequence is extracted from the multi-mode excitation signal library. This precisely maps the abstract mode command output from the previous intelligent decision into a specific, executable sequence of physical commands to control the exciter's actions. The excitation mode code acts as the unique key connecting decision-making and execution, ensuring accurate scheduling. In software implementation, the multi-mode excitation signal library can be a lookup table or dictionary structure with the excitation mode code as the key and the corresponding array of digital waveform sequences as the value. Once the adaptive scheduling strategy generates a specific excitation mode code, this code is used as an index to quickly retrieve and read the complete, predefined digital waveform sequence from the library. The digital waveform sequence contains detailed information on the time-varying amplitude, frequency, and phase of all parameters required to drive the exciter.

[0051] Specifically, for example, code 01 corresponds to a complete linearly increasing voltage digital sequence, while code 02 corresponds to a sequence with randomly changing frequencies. By outputting this digital waveform sequence, the digital controller of the road roller vibrator provides a direct basis for generating the target vibration, enabling lossless, accurate, and rapid execution of intelligent scheduling decisions. This ensures that the intent of the adaptive scheduling strategy can be accurately translated into actual physical excitation of the soil, outputting specific excitation action commands to drive the subsequent data acquisition process.

[0052] S3. Based on excitation mode encoding and digital waveform sequence synchronous acquisition of multi-source sensor data, output multi-source sensor data frames.

[0053] S3.1 Receive the excitation mode encoding and digital waveform sequence, and generate a unified microsecond-level precision timestamp based on the start time of the digital waveform sequence.

[0054] Furthermore, by using the moment of excitation generation, rather than an arbitrary clock signal, as the synchronization origin, all subsequently acquired sensor signals have a common reference point on the time axis with clear physical meaning. That is, the moment the excitation begins to act on the soil. At the instant when the digital signal processor begins to output the digital waveform sequence to the power amplifier, a dedicated timer or clock module is triggered to generate a global timestamp with microsecond resolution. The unified timestamp not only marks the absolute time but also marks the moment when the excitation event occurred.

[0055] Specifically, it fundamentally ensures the comparability and correlation of four types of information—dynamic response, contact state, material properties, and spatial location—in the time dimension, provides a strict time alignment basis for analyzing the causal relationship between excitation and response, as well as the intrinsic connection between multi-source signals, and establishes a unified and accurate time coordinate system for the fusion of multi-source asynchronous sensor data.

[0056] S3.2. Encode the excitation mode with a unified timestamp and send it to the acquisition controller of the triaxial accelerometer, rotary encoder, pressure sensor array, near-infrared spectral moisture sensor and high-precision GNSS receiver.

[0057] Furthermore, each sensor data stream to be acquired is pre-labeled with a context tag to clearly identify which specific stimulus event the data is responding to and when the response begins. Broadcast or multicast communication methods are used to ensure that the acquisition controllers of the three-dimensional accelerometer, rotary encoder, pressure sensor array, near-infrared spectral moisture sensor, and high-precision GNSS receiver can learn the identity and start time of this acquisition task almost simultaneously within a very short time difference.

[0058] Specifically, for example, data packets or electrical pulse signals containing excitation mode encoding and a unified timestamp are distributed to each acquisition controller via a high-speed bus or trigger line. This overcomes the shortcomings of traditional methods where data from each sensor operates independently and is difficult to accurately match with specific excitation events later. From the source of the acquisition command, all data are forcibly given unified context information, ensuring that all data segments in the subsequent multi-source sensor data frames logically belong to the same detection event, avoiding data association errors, and completing the unified context identification and synchronous triggering of multi-sensor acquisition tasks.

[0059] S3.3 A three-dimensional accelerometer collects simulated acceleration signals in the X, Y, and Z directions; a pressure sensor array collects simulated pressure distribution signals on the contact surface between the steel wheel and the soil; and a near-infrared spectral moisture sensor collects spectral reflection signals from the soil surface.

[0060] Furthermore, a triaxial accelerometer collects simulated acceleration signals in the X, Y, and Z directions; a pressure sensor array collects simulated pressure distribution signals on the contact surface between the steel wheel and the soil; and a near-infrared spectral moisture sensor collects spectral reflectance signals from the soil surface. Within one excitation cycle, the instantaneous responses of the soil state in three orthogonal dimensions—dynamics, contact mechanics, and material composition—are simultaneously captured. Using synchronously triggered sensor controllers, each sensor unit is driven to acquire data according to its optimal sampling frequency and operating mode.

[0061] Specifically, the triaxial accelerometer captures the inertial motion of the vibrating wheel with its high sampling rate; the pressure sensor array captures the spatial distribution image of the contact surface pressure at a slightly lower frame rate; and the near-infrared spectral moisture sensor acquires the reflectance spectrum at a rate suitable for spectral analysis. These simulated signals are the source for subsequent quantitative analysis, enabling multi-dimensional, instantaneous, and synchronous snapshot-like perception of the soil compaction state. This provides rich, time-correlated raw information, overcoming the limitations of single-type sensors with limited sensing dimensions or information fragmentation caused by time-division acquisition. It offers unprecedentedly rich raw data material for a comprehensive and integrated assessment of compaction, and simultaneously acquires raw simulated signals reflecting the multi-physics response of the soil. S3.4 Pack the excitation mode encoding, unified timestamp, acceleration simulation signal, pressure distribution simulation signal, spectral reflection signal, three-dimensional coordinates, velocity, and heading into a multi-source sensor data frame.

[0062] Furthermore, the eight heterogeneous data types—excitation mode encoding, unified timestamp, acceleration simulation signal, pressure distribution simulation signal, spectral reflection signal, three-dimensional coordinates, velocity, and heading—are encapsulated into a logical whole. The scattered, multi-stream parallel sensor raw data are reorganized into a self-describing, atomic data unit based on their inherent spatiotemporal and logical correlations. The data frame fields cover four major categories of information: event identity, absolute time, multiphysics response, and spatial state.

[0063] Specifically, after receiving digitized acceleration simulation signals, pressure distribution simulation signals, spectral reflectance signals, and three-dimensional coordinates, velocity, and heading from the high-precision GNSS receiver, the firmware packages these data along with the initially received excitation mode encoding and a unified timestamp, according to a predefined format and order, to generate a complete multi-source sensor data frame. Within this data frame, all data points to the same excitation event, the same start time, and the same spatial location, creating a highly cohesive and loosely coupled data object. This allows subsequent processing steps to easily extract and analyze all the associated information obtained under a single excitation, just like processing a single data stream. This simplifies the complexity of the data processing flow and ensures the integrity and correctness of data association, serving as a key data structure supporting subsequent feature fusion and intelligent evaluation. The output is a standardized data packet containing complete context and multi-dimensional responses. S4. Extract features from the multi-source sensor data frames to obtain the broad-spectrum dynamic fingerprint vector and the mesoscopic surface morphology feature vector.

[0064] S4.1 Extract the acceleration simulation signal, angular velocity sequence, pressure distribution simulation signal, and spectral reflection signal from the multi-source sensor data frame.

[0065] Furthermore, the structured multi-source sensor data frames are precisely deconstructed to separate the heterogeneous raw data streams required for subsequent analyses, namely, acceleration simulation signals, angular velocity sequences, pressure distribution simulation signals, and spectral reflectance signals. This provides a clean and independent input data source for subsequent parallel and targeted feature extraction processes. Based on the predefined field formats and arrangement order within the multi-source sensor data frames, data parsing programs or hardware interfaces accurately restore the packaged mixed data streams into their respective independent physical quantity data.

[0066] Specifically, based on the frame header identifier and length information, the parsing program sequentially reads and separates the following arrays: an acceleration simulation signal array representing three-dimensional vibration, an angular velocity sequence array representing rotational motion, a pressure distribution simulation signal matrix representing a two-dimensional pressure image, and a spectral reflectance signal array representing spectral intensity. This ensures that multi-dimensional data collected from the same timestamp and the same excitation event can be extracted independently and without error. It achieves a smooth transition from fused transmission to independent analysis of multi-source heterogeneous data, guarantees the integrity and accuracy of the data link, completes the precise unpacking of multi-source sensor data frames, and outputs four independent raw signal data streams.

[0067] S4.2. Perform zero-mean and five-point cubic smoothing on the vertical component of the acceleration simulation signal to obtain the preprocessed vertical acceleration sequence, and perform a 256-point fast Fourier transform to obtain the spectrum.

[0068] Furthermore, in order to extract frequency components that clearly reflect the soil's response to vertical excitation from the noisy raw vibration signal, zero-meaning is used to eliminate the DC offset of the signal, so that the subsequent spectrum analysis focuses on the AC dynamic components. Five-point cubic smoothing is used to filter out high-frequency random noise and improve the smoothness of the signal. Then, a fast Fourier transform is performed on the preprocessed vertical acceleration sequence to transform it from the time domain to the frequency domain, and the spectrum is obtained.

[0069] Specifically, zero-mean normalization is achieved by subtracting the signal mean, five-point cubic smoothing is achieved using a convolution kernel with fixed coefficients, and the fast Fourier transform is implemented using an efficient butterfly algorithm. These processes work together to transform the original, chaotic time-domain vibration waveform into a spectrum graph with frequency on the horizontal axis and amplitude on the vertical axis. The graph clearly shows the energy distribution at different frequencies. The standardized preprocessing enhances the usability of the signal, and the frequency structure information hidden within the time-domain waveform is revealed through frequency domain transformation.

[0070] S4.3 Extract the frequency domain features of the dominant frequency value, dominant band energy, frequency weighting center, spectral variance, and approximate entropy from the spectrum.

[0071] Furthermore, in order to characterize the spectral characteristics of soil vibration response from multiple angles and in a quantitative manner, a comprehensive set of frequency domain descriptors is formed. Each feature captures a different aspect of the spectrum: the dominant frequency value identifies the frequency point with the strongest response, reflecting the resonance characteristics of the soil; the dominant frequency band energy quantifies the magnitude of the main oscillation energy; the frequency weighting center describes the overall distribution of spectral energy; the spectral variance measures the degree of energy dispersion on the frequency axis; and the approximate entropy measures the complexity and regularity of the signal sequence.

[0072] Specifically, the dominant frequency is obtained by finding the point with the maximum amplitude in the spectrum. The dominant frequency band energy is obtained by integrating the sum of squares of amplitudes within a certain bandwidth near the dominant frequency. The frequency weighting center is obtained by averaging all frequency points with their amplitude weights. The spectral variance is obtained by the dispersion of the weighted frequencies. The approximate entropy of the time series is calculated using a specific algorithm. These five features together constitute the frequency domain part of a broad-spectrum dynamic fingerprint that describes the spectrum from five dimensions: central tendency, energy intensity, distribution location, dispersion, and complexity. It provides rich spectral information far exceeding that of a single frequency or amplitude index, and can more sensitively capture the complex mapping of soil stiffness, damping, and other characteristics in the frequency domain. Multi-dimensional frequency domain features are quantified and extracted from the spectrum.

[0073] S4.4. Using angular velocity sequences and acceleration simulation signals, estimate the wheel-soil contact force through a single-degree-of-freedom dynamic model, and plot the force-displacement hysteresis curve in combination with displacement.

[0074] Furthermore, by utilizing the measurable motion parameters (angular velocity and acceleration) of the vibratory roller's wheel, combined with known or estimable inertial parameters such as its moment of inertia, and based on Newton's second law or the law of angular momentum, the wheel-soil contact force, which is difficult to measure directly, can be calculated in reverse. The estimated wheel-soil contact force is then combined with the vertical displacement of the steel wheel obtained by integrating the angular velocity, and a curve is plotted in the force-displacement coordinate system, i.e., the force-displacement hysteresis curve.

[0075] Specifically, the vibrating wheel is treated as a concentrated mass. Inertial force is estimated using its vertical acceleration and known mass, and then contact force is calculated by combining this with other known forces. Displacement is obtained by integrating the angular velocity signal and converting it using geometric relationships. The area enclosed by this curve represents energy dissipation, while changes in its shape and slope reflect the plastic deformation, elastic recovery, and damping characteristics of the soil. This approach bypasses the technical difficulties of directly measuring contact force, obtaining a key illustration reflecting the nonlinear mechanical behavior of the soil in an indirect but physically meaningful way. This provides a foundation for extracting features from a geometric perspective and constructs a key illustration characterizing the nonlinear dynamics of wheel-soil interaction.

[0076] S4.5 Extract the geometric features of hysteresis loop area, average slope and shape fullness coefficient from the force-displacement hysteresis curve.

[0077] Furthermore, from the force-displacement hysteresis curve, three geometric features with clear physical meanings are extracted: hysteresis loop area, average slope, and shape fullness coefficient. The visual shape of the curve is transformed into quantifiable numerical indicators to describe the soil's energy dissipation capacity, equivalent stiffness, and the degree of nonlinearity in hysteresis behavior. These three geometric features were chosen because they correspond to important properties of soil dynamics: the hysteresis loop area corresponds to the energy dissipated due to internal friction and plastic deformation in a loading-unloading cycle; the average slope is usually approximated by the secant stiffness of the curve, reflecting the soil's average resistance to deformation under cyclic loading; and the shape fullness coefficient (usually defined as the ratio of the loop area to the area of ​​its circumscribed rectangle) describes the fullness of the hysteresis loop and is related to the soil's damping characteristics and the degree of nonlinearity in the stress-strain relationship.

[0078] Specifically, the hysteresis loop area is calculated through numerical integration, the average slope is calculated through linear fitting or connecting specific points, and the shape fullness coefficient is obtained by dividing the loop area by the area of ​​the rectangle formed by the maximum span of its projection on the force axis and displacement axis. The rich mechanical information contained in the complex and nonlinear force-displacement hysteresis curve is condensed into a few simple, stable and physically clear numerical features, which facilitates subsequent mathematical processing and pattern recognition. Quantitative temporal geometric features are extracted from the hysteresis curve.

[0079] S4.6 Arrange the frequency domain features and geometric features in a fixed order to form a broad spectrum dynamic fingerprint vector. Perform threshold segmentation and connected component analysis on the pressure distribution simulation signal to identify the effective contact area and calculate the morphological features of average grounding pressure, pressure non-uniformity coefficient, and pressure center offset.

[0080] Furthermore, the extracted frequency domain features and geometric features are combined in a preset fixed order to form a broad-spectrum dynamic fingerprint vector; secondly, the pressure distribution simulation signal is processed by image processing and morphological features are calculated. The objectives are to construct a multi-dimensional feature vector that comprehensively describes the dynamic response, and to extract spatial statistical features of the contact state from the contact surface pressure image. The construction of the broad-spectrum dynamic fingerprint vector integrates two different but complementary analytical perspectives: the frequency domain and the time domain force-displacement relationship, providing a more comprehensive digital description of the soil's dynamic characteristics. As for the processing of the pressure distribution simulation signal, threshold segmentation, a common technique in image processing, is used to distinguish the effective contact area from the background. The contact area is confirmed through connected component analysis, and then the average ground pressure is calculated to reflect the average load intensity, the pressure non-uniformity coefficient is calculated to reflect the concentration of the load distribution, and the pressure center offset is calculated to reflect the spatial symmetry of the load distribution.

[0081] Specifically, a pressure threshold is applied to the pressure image matrix, and pixels above the threshold are marked as valid contact points. Cluster analysis is performed on these points to identify connected regions. Finally, three features are calculated according to the formula: a broad-spectrum dynamic fingerprint vector that integrates the spectrum and constitutive information of the vibration response; and morphological features that provide a spatial quantitative description of the support state of the soil surface after compaction from the perspective of static contact. Together, they characterize the compaction effect from both dynamic and static levels, completing the transformation from dynamic signals and pressure images to standardized feature vectors.

[0082] The expression for average ground pressure is: ; in, For average grounding pressure, The total number of effective contact points. This represents the row number of the pressure sensor array. This represents the number of columns in the pressure sensor array. For column index variables, For the position located at the Line 1 The pressure value measured by the pressure sensor of the column, This is a row index variable.

[0083] The expression for the pressure non-uniformity coefficient is: ; in, This is the pressure non-uniformity coefficient. For the position located at the Line 1 The maximum pressure value measured by the pressure sensor in the column.

[0084] The expression for the pressure center offset is: ; in, This is the offset of the pressure center. For the first Column sensor Direction coordinates For the first Line sensor Direction coordinates The theoretical center of the rolling mill coordinate, The theoretical center of the rolling mill coordinate.

[0085] S4.7 Arrange the morphological features in a fixed order to form a mesoscopic surface morphological feature vector. Furthermore, the three morphological features—average grounding pressure, pressure non-uniformity coefficient, and pressure center offset—are arranged in a predefined, fixed order to form a mesoscopic surface morphology feature vector. Spatial statistical indicators describing the pressure distribution at the contact surface in different dimensions are integrated into a structured, one-dimensional mathematical vector. This facilitates unified mathematical operations and pattern matching with other features, such as the broad-spectrum dynamic fingerprint vector, to determine an invariant arrangement order.

[0086] Specifically, in the software code, an array of length 4 can be defined, and the four feature values ​​are stored in sequence according to the above order. This realizes the standardization and vectorization of mesoscopic contact morphology information, enabling it to easily participate in machine learning or pattern recognition algorithms such as distance calculation and similarity assessment based on vector space. This prepares the data format for the compaction assessment of multi-feature fusion and completes the vectorization encapsulation of the pressure distribution features of the contact surface.

[0087] S5, combined into a feature triplet data packet.

[0088] S5.1 Process the spectral reflectance signal to retrieve the water content value, and align the broad-spectrum dynamic fingerprint vector, mesoscopic surface morphology feature vector, and water content value according to the timestamp to form a feature triplet data packet.

[0089] Furthermore, features extracted from three independent physical dimensions—dynamic response, contact surface morphology, and material composition—namely, the broad-spectrum dynamic fingerprint vector, the mesoscopic surface morphology feature vector, and the water content value, are precisely aligned based on their shared timestamps from the original multi-source sensor data frames. These three are then combined into a logically unified, structured feature triplet data package to construct a standardized digital representation that comprehensively and collaboratively describes the overall state of soil under specific excitation at a given moment. First, the spectral reflectance signal is processed to invert the water content value. The water content value utilizes the known physical correlation between near-infrared spectroscopy and soil moisture content. Through a pre-calibrated inversion algorithm (such as partial least squares regression), the spectral data is converted into the key physical property parameter, water content value. The broad-spectrum dynamic fingerprint vector, the mesoscopic surface morphology feature vector, and this water content value are aligned by timestamp. This means that the three sets of data parsed from the multi-source sensor data frames, which have the same or correlated timestamps, are identified as belonging to the same spatiotemporal sampling point, thus ensuring that they describe the state of the same soil sample at the same instant. The three are combined into a feature triplet data package.

[0090] Specifically, it is encapsulated into a data structure containing three fields, overcoming the shortcomings of traditional methods where different physical quantity feature information is isolated and lacks effective correlation. It binds together the three key factors affecting compaction degree: dynamic characteristics, surface contact state, and material inherent properties (moisture content), forming a state snapshot with complete information and clear internal correlation. This provides a high-dimensional, complementary, and physically interpretable input feature set for subsequent pattern matching and evaluation, greatly enhancing the information dimension and discrimination ability of compaction state identification. It completes the spatiotemporal alignment and structured integration of multi-source heterogeneous feature information and outputs a standardized data package characterizing the instantaneous comprehensive state of the soil.

[0091] S6. Perform pattern matching and similarity evaluation between the feature triplet data package and the pre-stored compaction state feature template library.

[0092] S6.1 Load the pre-stored compaction state feature template library from local storage, and perform an overall comparison between the feature triplet data package and each feature triplet template in the pre-stored compaction state feature template library.

[0093] Furthermore, a pre-stored compaction state feature template library is used as the benchmark knowledge base for evaluation, and a core operation is defined to perform an overall comparison between the feature triplet data package and each feature triplet template in the pre-stored compaction state feature template library. To achieve an instance-based, non-parametric compaction state assessment model, a pre-stored compaction state feature template library is constructed. This library is formed by synchronously collecting a large number of feature triplet data packets and corresponding true compaction values ​​accurately measured by standard methods such as sand cone method during the early stages of construction or test sections. These feature triplets and true values ​​are then paired and stored. Each such pairing record constitutes a feature triplet template, which contains a complete multi-dimensional feature pattern observed under a specific historical compaction state. This pre-stored compaction state feature template library is completely loaded into memory from local storage media such as hard disk or solid-state storage. The currently generated feature triplet data packet to be evaluated is sequentially compared with each feature triplet template in the library. The overall comparison means that the broad-spectrum dynamic fingerprint vector, mesoscopic surface morphology feature vector, and moisture content value of the current data packet are treated as a complete feature combination and compared with the corresponding vector group of each template in the template library for overall similarity, rather than comparing a single feature in isolation.

[0094] Specifically, for example, the comparison process can be initiated by calculating a certain distance or similarity metric between two feature triples in the space spanned by all feature dimensions. Based on historical measured data, there is no need to build and solve complex physical or statistical inversion models, thus avoiding the problems of model mismatch and parameter calibration. The evaluation process directly relies on the inherent pattern of the data itself to complete the operation from loading the knowledge base to initiating the comparison with historical templates one by one.

[0095] S6.2. Based on the comparison results, generate a similarity score for each feature triplet template in the pre-stored compaction state feature template library, reflecting the degree of similarity with the feature triplet data package.

[0096] Furthermore, a quantitative similarity score is assigned to the result of each overall comparison. This score accurately reflects the degree of similarity between the current feature triplet data package and each feature triplet template in the pre-stored compaction state feature template library. It transforms the fuzzy, qualitative concept of similarity into a precise, sortable value, providing an objective basis for subsequent selection and decision-making. The similarity score generation mechanism is usually based on the distance or similarity metric between the feature triplet data package and each feature triplet template in the multidimensional feature space, such as Mahalanobis distance or cosine similarity.

[0097] Specifically, the feature triplet data package and all feature values ​​(including dynamic fingerprint vector, morphological feature vector, and water content value) of a certain feature triplet template are concatenated into a long feature vector. Then, the Euclidean distance between these two long vectors is calculated. The reciprocal of the distance or the value after negative exponential transformation can be used as the similarity score. The smaller the distance, the higher the score. Each feature triplet template obtains a specific similarity score for the current data package to be evaluated. Through quantitative scoring, the closeness of different historical templates to the current state becomes comparable, thereby accurately selecting the most similar historical instances. This overcomes the uncertainty of matching based on subjective experience or a single indicator, and realizes the objectification and digitization of the evaluation process. A quantitative similarity measure for the current state is generated for each historical template in the template library.

[0098] S6.3 Sort the templates from high to low according to the similarity scores to generate a template similarity ranking list. Based on the preset selection quantity, select several feature triplet templates with the highest similarity scores from the top of the template similarity ranking list to form a set of highly similar templates.

[0099] Furthermore, a highly similar template set is selected based on the ranking results. Utilizing the idea of ​​swarm intelligence or ensemble learning, instead of relying on a single best-matching template, a group of highly similar templates is aggregated, and their collective experience is used for the final evaluation to improve the robustness and noise resistance of the evaluation. The operation consists of two steps: First, the templates are sorted from high to low according to their similarity scores, generating a template similarity ranking list. This is a process of arranging all feature triplet templates according to their similarity scores with the current data packet from high to low, forming an ordered list. Then, based on a preset selection quantity, several feature triplet templates with the highest similarity scores are selected from the top of the template similarity ranking list to form a highly similar template set. Here, the preset selection quantity is an adjustable parameter, such as selecting the top ten.

[0100] Specifically, after obtaining the scores of all templates, a quicksort algorithm is used to sort all template-score pairs in descending order of score. Then, the top few templates in the list are simply extracted, forming a highly similar template set. This method effectively filters out irrelevant or significantly different historical samples, focusing on the most relevant empirical subset. Because the set contains multiple similar samples, it can mitigate the measurement errors or randomness that may exist in a single sample, making subsequent inferences based on this set more stable and reliable. This overcomes the risk of significant deviations in the overall evaluation results due to random errors (such as outliers) of a single best-matching template, completing the screening of historical templates and obtaining a set of reference templates most relevant to the current state.

[0101] S7. Output real-time compaction values ​​with confidence ratings through a lookup and mapping mechanism.

[0102] S7.1 From each feature triplet template in the high similarity template set, find the associated true value of the measured compaction degree obtained during calibration using the sand cone method, and assign a weight factor to the true value of the measured compaction degree using the sand cone method.

[0103] Furthermore, from each feature triplet template in the highly similar template set, the associated true values ​​of compaction obtained during calibration using the sand cone method are identified, and weighting factors are assigned to these true values. From the most similar historical instances selected, their corresponding authoritative compaction standard values ​​are directly extracted, and different importance is assigned according to their degree of similarity. Each feature triplet template in the highly similar template set has established an inseparable association with a true compaction value accurately measured at the same location in the field using standard test methods such as the sand cone method when constructing the pre-stored compaction state feature template library.

[0104] Specifically, based on the unique identifier or storage structure of each template, a lookup operation is performed to retrieve the true value of the measured compaction degree of the sand cone method associated with it. Simultaneously, a weighting factor is assigned to each true value. This weighting factor is typically positively correlated with the similarity score obtained by the template in previous steps; that is, the more similar the template is to the current state, the greater the weight of the corresponding true value in subsequent comprehensive calculations. For example, the similarity score can be normalized and directly used as a weighting factor, or the score can be mapped to a weight using a monotonically increasing function.

[0105] S7.2. Based on the preset mapping rules, the true value of the measured compaction degree of the sand cone method associated with the feature triplet template and the corresponding weighting factor are combined to generate the compaction degree estimate.

[0106] Furthermore, by using mapping rules, multiple measured compaction values ​​from the sand cone method and their weighting factors are mathematically synthesized to generate a compaction estimate. This aggregates a set of dispersed, weighted historical values ​​into a single numerical estimate representing the most likely compaction degree of the current state. The most common and effective mapping rule is the weighted average. The compaction estimate is equal to the sum of the measured compaction values ​​from the sand cone method associated with all feature triplet templates in the highly similar template set, multiplied by their corresponding weighting factors, and then divided by the sum of all weighting factors. If the highly similar template set contains three templates with true values ​​a, b, and c, and weighting factors w1, w2, and w3, then the physical meaning of this mapping rule is that it assumes the compaction degree of the current state is most likely located at the weighted center position formed by the compaction degree true values ​​of the most similar historical states.

[0107] Specifically, this method interpolates or infers the value of the current unknown point from discrete historical sample points in a smooth, continuous, and mathematically robust manner. This avoids the random errors that may arise from directly using the true value of a single most similar template, improving the stability and accuracy of the estimation and overcoming the drawback of the simple nearest neighbor method being sensitive to noise. Through a weighted averaging mapping rule, multiple similar historical true values ​​are fused into a single compaction estimate.

[0108] S7.3 Based on the overall characteristics of the highly similar template set, a reliability index is generated, which is converted into a confidence rating value. The compaction estimate is combined with the confidence rating, and the output is a real-time compaction value with a confidence rating.

[0109] Furthermore, to provide an endogenous measure of the reliability of the evaluation results, enabling users to understand the uncertainty or credibility of the estimate, the confidence rating is generated as follows: it calculates a reliability index based on the overall characteristics of the highly similar template set, and then converts this index into a standardized value. These overall characteristics can include two aspects: first, the concentration of similarity scores among the feature triplet templates in the highly similar template set; and second, the dispersion of the true values ​​of the measured compaction degree associated with these templates using the sand cone method. A common approach is to calculate the standard deviation of these true values ​​or the mean of the weighted similarity scores, and then convert these indices into a confidence rating value between zero and one using a predefined mapping function. The compaction estimate and this confidence rating value are then encapsulated together, for example, as a structure containing two fields, and output together as a real-time compaction degree value with a confidence rating.

[0110] Specifically, it imbues automated measurement results with crucial meta-information—reliability—allowing subsequent quality assessments, construction decisions, and data screening to be based on a more informed foundation. For example, a low confidence rating can prompt the operator that the point needs verification or that the area requires special attention. This overcomes the limitations of traditional methods that only provide measurement values ​​without any indication of their reliability, significantly enhancing the practicality and reference value of the measurement results. It also completes the comprehensive calculation and encapsulation of the evaluation results and their reliability, resulting in a comprehensive output.

[0111] S8. Based on the real-time compaction value with confidence rating and spatial location information, output compaction field and construction guidance information.

[0112] S8.1 Update the real-time compaction value with confidence rating and associated spatial location information to the gridded digital map in memory, and use the compaction estimate and confidence rating to perform interpolation and smoothing on each grid cell in the digital map.

[0113] Furthermore, discrete, real-time compaction values ​​with spatial coordinates and confidence ratings are dynamically integrated and processed into a continuous, gridded spatial distribution model, i.e., a digital map. The map is then intelligently interpolated and smoothed using compaction estimates and confidence ratings. Point-like measurement data is transformed into a surface-like, visualized compaction field that reflects spatial continuity and data reliability. A digital map covering the construction area, composed of regular grid cells, is built in computer memory. When a new real-time compaction value with a confidence rating and its spatial location information arrives, its corresponding grid cell is located based on the location information, and the cell's value is updated with the new compaction estimate. Since the measurement points are discrete, many grid cells may not have direct measurement values, thus requiring interpolation and smoothing processing.

[0114] Specifically, for example, in inverse distance weighted interpolation or Kriging interpolation, the weight of each known point is not only inversely proportional to its distance to the point to be interpolated, but also directly proportional to its own confidence rating. This means that a high-confidence measurement point contributes more to the surrounding area than a low-confidence point. Simultaneously, smoothing (such as using a Gaussian filter) can eliminate drastic local fluctuations in the data, making the field more continuous. This method no longer generates isolated measurement point data, but a smooth and continuous spatial distribution model of compaction (i.e., a compaction field) that integrates spatial location relationships and the reliability of the data itself, providing a foundation for macroscopic quality assessment and regional analysis. This overcomes the limitation of traditional point measurements in intuitively displaying the overall field distribution and trends by fusing discrete point data to generate a continuous, smooth spatial compaction field that considers data reliability.

[0115] S8.2 Render the compaction field data layer into a compaction cloud map represented by color depth and confidence level indicators. Scan the compaction field data layer to identify all continuous grid areas where the compaction estimate is lower than the preset target and mark them as weak compaction areas.

[0116] Furthermore, the abstract compaction field data layer is transformed into an intuitive compaction cloud map containing dual information encoding. Secondly, based on this data layer, defect area identification is automatically performed. The goal is to achieve a visual representation of the measurement results and automatic detection of areas with unqualified compaction quality. The rendering process maps the compaction estimate to color depth and represents the confidence rating through another visual channel.

[0117] Specifically, for example, high-confidence areas are brightly colored or highly saturated, while low-confidence areas are dark or semi-transparent. A compaction cloud map conveys two layers of information: the degree of compaction and the reliability of the data. By scanning the compaction field data layer, the program traverses each grid cell, marks the cells whose compaction estimates are lower than the preset target value, and uses region growing or connected component analysis algorithms in image processing to merge these spatially adjacent, unqualified cells into continuous weak compaction areas.

[0118] It should be noted that visualization makes complex data fields clear at a glance, allowing operators to quickly grasp the distribution and reliability of compaction quality across the entire field; the automatic identification function can objectively and without omission locate all non-conforming areas, avoiding the subjectivity and oversight of manual interpretation, overcoming the inefficiency and error-proneness of traditional methods that rely on manual identification of problems from data tables or scatter plots, generating a visual cloud map containing dual information and automatically identifying all weak compaction areas.

[0119] S8.3 Based on the spatial distribution and size of weak compaction areas, as well as the real-time spatial location and attitude of the roller, a recommended compaction path covering all weak compaction areas is planned to form construction guidance information. Furthermore, by combining the quality inspection results (weak compaction areas) with the real-time status (spatial position and attitude) of the construction machinery, an operable recommended compaction path is proactively generated, forming construction guidance information. This closes the measurement-evaluation process, extending from problem diagnosis to guidance and solution, achieving truly intelligent construction guidance. The path planning logic is as follows: the input is the spatial distribution and size of the weak compaction areas, as well as the real-time spatial position and attitude of the roller. The planning algorithm needs to solve an optimization problem: given the roller's current position and attitude, how to plan a travel path that can efficiently cover all weak compaction areas, while considering constraints such as the roller's turning radius and travel efficiency.

[0120] Specifically, for example, each weakly compacted area can be considered as a target point or region to be visited. The problem is transformed into a path planning problem with region coverage constraints. This can be solved using variations of the Traveling Salesman Problem, region coverage algorithms (such as the plowing method), or graph search-based algorithms (such as the A* algorithm) to generate a coherent and feasible recommended compaction path. This path information, along with the compaction degree contour map and the identification of weakly compacted areas, constitutes construction guidance information. This information is presented to the operator in real time on a display screen, upgrading compaction quality monitoring from passive result display to proactive work guidance. It directly tells the operator where the problem lies and how to proceed to correct it, greatly improving the efficiency and targeting of handling weak areas, reducing blind compaction and resource waste, and achieving a closed loop between measurement and construction decision-making.

[0121] This embodiment also provides an automatic measurement system for compaction degree, including: an initialization module, initialization and construction parameter configuration, output configuration data, construction of a multi-mode excitation signal library and adaptive scheduling strategy, and output of excitation mode encoding and digital waveform sequence; The acquisition module synchronously acquires multi-source sensor data based on excitation mode encoding and digital waveform sequence, and outputs multi-source sensor data frames. The fusion module extracts features from multi-source sensor data frames to obtain a broad-spectrum dynamic fingerprint vector and a mesoscopic surface morphology feature vector, which are then combined into a feature triplet data packet. The evaluation module performs pattern matching and similarity assessment between the feature triplet data package and the pre-stored compaction state feature template library, and outputs real-time compaction values ​​with confidence ratings through a lookup and mapping mechanism. The decision output module outputs compaction field and construction guidance information based on real-time compaction values ​​with confidence ratings and spatial location information.

[0122] This embodiment also provides a computer device suitable for an automatic compaction measurement method, comprising: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to implement the automatic compaction measurement method proposed in the above embodiment.

[0123] The computer device can be a terminal, comprising a processor, memory, communication interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, carrier networks, NFC (Near Field Communication), or other technologies. The display screen can be an LCD screen or an e-ink screen. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad on the computer device's casing, or an external keyboard, touchpad, or mouse.

[0124] This embodiment also provides a storage medium storing a computer program that, when executed by a processor, implements the automatic measurement method for compaction as proposed in the above embodiments. The storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read Only Memory (EPROM), Programmable Red-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0125] In summary, this invention transforms the road roller vibrator from a fixed vibration source into an active sensing tool that can dynamically output optimal detection signals according to working conditions. This achieves adaptive matching between the excitation mode and the soil state and construction stage, thereby enriching the excitation dimensions and enhancing the information content and relevance of the response signal. By performing pattern matching and similarity evaluation between the feature triplet data package and a pre-stored compaction state feature template library, and outputting real-time compaction values ​​with confidence ratings through a search and mapping mechanism, this invention replaces the traditional complex model inversion with a pattern matching and direct mapping mechanism based on an instance library. By comparing the real-time extracted multi-dimensional feature vectors with the historical calibration template library, the compaction estimate is found from the most similar known state instances. A confidence rating is generated based on the consistency of the matching results, thus bypassing the calibration difficulties of complex models and their sensitivity to interference factors. This achieves highly robust and adaptable compaction assessment without relying on a precise physical model, while also providing quantitative indicators of result reliability.

[0126] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for automatic measurement of compaction degree, characterized in that: Comprising, Initialization and construction parameter configuration, output configuration data, build multi-mode excitation signal library and adaptive scheduling strategy, output excitation mode code and digital waveform sequence; Based on the excitation mode code and digital waveform sequence, the multi-source sensor data is synchronously collected, and the multi-source sensor data frame is output; Feature extraction is performed on the multi-source sensor data frame to obtain a broad-spectrum dynamic fingerprint vector and a mesoscopic surface morphology feature vector, which are combined into a feature triple data packet; The feature triple data packet is matched with the pre-stored compaction state feature template library, and the similarity is evaluated, and the real-time compaction degree value with confidence rating is output through the lookup and mapping mechanism; According to the real-time compaction degree value with confidence rating and spatial position information, the compaction degree field and construction guidance information are output.

2. The method of automatically measuring the degree of compaction according to claim 1, characterized in that: Initialization and construction parameter configuration, output configuration data, including the following steps: Select the soil type from the preset list, obtain the soil type code, input the value, obtain the compaction degree value, and draw the boundary on the map interface to obtain the construction boundary coordinates; According to the soil type code, the standard compaction test maximum dry density value, the optimum water content range, the plasticity index reference value and the typical compression modulus curve parameters are obtained from the database; Create a format-named file, obtain a data record file pointer, establish a connection with a cloud server, and obtain a wireless communication link state; Integrate the soil type code, compaction degree value, construction boundary coordinates, standard compaction test maximum dry density value, optimum water content range, plasticity index reference value, typical compression modulus curve parameters, data record file pointer, and wireless communication link state, and output configuration data.

3. The method of automatically measuring the degree of compaction according to claim 2, wherein: Build multi-mode excitation signal library and adaptive scheduling strategy, output excitation mode code and digital waveform sequence, including the following steps: According to the soil type code, the multi-mode excitation signal library pre-stored for the soil type is called from the non-volatile memory; The multi-mode excitation signal library contains preset digital waveform sequences, linear sweep signals, pseudo-random frequency signals, multi-frequency composite signals and single pulse impact signals; Obtain the real-time working condition parameters of the instantaneous rolling speed, the current rolling number, the compaction degree change gradient and the GNSS position information; The adaptive scheduling strategy analyzes the real-time working condition parameters based on fuzzy decision rules, selects the optimal excitation mode from the multi-mode excitation signal library, and generates the excitation mode code corresponding to the selected optimal excitation mode; Extract the digital waveform sequence corresponding to the excitation mode code from the multi-mode excitation signal library.

4. The method of automatically measuring the degree of compaction according to claim 3, wherein: Based on the excitation mode code and digital waveform sequence, the multi-source sensor data is synchronously collected, and the multi-source sensor data frame is output, including the following steps: Receive the excitation mode code and digital waveform sequence, and generate a unified microsecond-level precision timestamp based on the starting time of the digital waveform sequence; Send the excitation mode code and the unified timestamp to the acquisition controllers of the three-axis acceleration sensor, the rotary encoder, the pressure sensor array, the near-infrared spectral moisture content sensor and the high-precision GNSS receiver; The three-way acceleration sensor collects X, Y, Z three direction acceleration analog signals, the pressure sensor array collects the pressure distribution analog signals on the contact surface of the steel wheel and the soil body, and the near-infrared spectrum moisture content sensor collects the spectrum reflection signals of the soil surface; The excitation mode code, unified time stamp, acceleration analog signal, pressure distribution analog signal, spectrum reflection signal, three-dimensional coordinate, speed, heading are packaged into a multi-source sensor data frame.

5. The method of automatically measuring the degree of compaction according to claim 4, characterized in that: Feature extraction is performed on the multi-source sensor data frame to obtain a wide spectrum of kinetic fingerprint vectors and mesoscopic surface topography feature vectors, including the following steps: The acceleration analog signal, angular velocity sequence, pressure distribution analog signal and spectrum reflection signal are parsed from the multi-source sensor data frame; The vertical direction component in the acceleration analog signal is zero-mean and five-point cubic smoothing processed to obtain a preprocessed vertical acceleration sequence, and a 256-point fast Fourier transform is performed to obtain a frequency spectrum diagram; The dominant frequency value, dominant band energy, frequency weighted center, spectrum variance and approximate entropy frequency domain features are extracted from the frequency spectrum diagram; The wheel-soil contact force is estimated by a single degree of freedom dynamic model using the angular velocity sequence and acceleration analog signal, and the force-displacement hysteresis curve is drawn in combination with the displacement; The hysteresis loop area, average slope and shape fullness coefficient geometric features are extracted from the force-displacement hysteresis curve; The frequency domain features and geometric features are arranged in a fixed order to form a wide spectrum of kinetic fingerprint vectors, the pressure distribution analog signal is threshold segmented and connected domain analyzed to identify the effective contact area, and the average ground pressure, pressure non-uniformity coefficient and pressure center offset are calculated. The topographic features are arranged in a fixed order to form a mesoscopic surface topography feature vector.

6. The method of automatically measuring the degree of compaction according to claim 5, wherein: The feature triple data packet is combined, including the following steps: The spectrum reflection signal is processed to obtain the moisture content value, and the wide spectrum of kinetic fingerprint vectors, mesoscopic surface topography feature vectors and moisture content values are aligned by time stamp to form a feature triple data packet.

7. The method of automatically measuring the degree of compaction according to claim 6, wherein: The feature triple data packet is matched with the pre-stored compaction state feature template library and the similarity is evaluated, including the following steps: The pre-stored compaction state feature template library is loaded from local storage, and the feature triple data packet is compared with each feature triple template in the pre-stored compaction state feature template library; According to the comparison result, a similarity score reflecting the similarity degree of each feature triple template in the pre-stored compaction state feature template library with the feature triple data packet is generated; According to the similarity score, a template similarity sorting list is generated from high to low, and a number of feature triple templates with the highest similarity score are selected from the top of the template similarity sorting list according to the preset selection number to form a high similarity template set.

8. The method of automatically measuring the degree of compaction according to claim 7, wherein: The real-time compaction degree value with confidence rating is output through the lookup and mapping mechanism, including the following steps: From each feature triple template in the high similarity template set, the associated sand replacement method measured compaction degree true value obtained during calibration is found out, and a weight factor is assigned to the sand replacement method measured compaction degree true value. According to the preset mapping rule, the sand filling method measured compaction degree true value associated with the feature triple template and the corresponding weight factor are integrated to generate a compaction degree estimation value; Based on the overall characteristics of the high-similarity template set, a reliability index is generated, which is converted into a confidence level rating value. The compaction degree estimation value and the confidence level rating are combined to output a real-time compaction degree value with a confidence level rating.

9. The method of automatically measuring the degree of compaction according to claim 8, wherein: According to the real-time compaction degree value with a confidence level rating and the spatial position information, the compaction degree field and the construction guidance information are output, including the following steps: The real-time compaction degree value with a confidence level rating and the associated spatial position information are updated to the grid-divided digital map in the memory, and the compaction degree estimation value and the confidence level rating are used to interpolate and smooth each grid cell in the digital map; The compaction degree field data layer is rendered into a compaction degree cloud chart represented by color depth and confidence level identification. All continuous grid regions with compaction degree estimation values below the preset target are identified as weak compaction regions by scanning the compaction degree field data layer. Based on the spatial distribution, size of the weak compaction region, and real-time spatial position information and attitude of the road roller, a recommended compaction path covering all weak compaction regions is planned to form the construction guidance information.

10. An automatic measurement system of compactness based on the automatic measurement method of compactness according to any one of claims 1 to 9, characterized in that: The initialization module initializes and configures the construction parameters, outputs the configuration data, builds a multi-mode excitation signal library and an adaptive scheduling strategy, and outputs the excitation mode code and digital waveform sequence. The acquisition module synchronously acquires multi-source sensor data based on the excitation mode code and digital waveform sequence, and outputs the multi-source sensor data frame. The fusion module extracts features from the multi-source sensor data frame to obtain a broad-spectrum dynamic fingerprint vector and a mesoscopic surface topography feature vector, which are combined into a feature triple data packet. The evaluation module performs pattern matching and similarity evaluation on the feature triple data packet and the pre-stored compaction state feature template library, and outputs a real-time compaction degree value with a confidence level rating through a lookup and mapping mechanism. The decision output module outputs the compaction degree field and the construction guidance information according to the real-time compaction degree value with a confidence level rating and the spatial position information.

Images