Engineering geology soil layer name judgment system carried on static sounding instrument, method and equipment

Abstract

The invention belongs to the field of engineering geology, particularly relates to an engineering geology soil layer name judgment system, method and equipment carried on a static sounding instrument, and aims to solve the problem that real-time dynamic soil layer name presumption cannot be carried out in an existing geotechnical engineering investigation technology. The system comprises a spectrum acquisition module which comprises a spontaneous light source and a multispectral sensor and is integrated between static sounding probe rods; the data processing module is used for analyzing the spectral data through a neural network and processing end resistance / side resistance / friction resistance ratio data of static sounding through a regression algorithm; and the data output module compares the two types of judgment results: outputting a single soil layer name when the two types of judgment results are consistent, and outputting double results and converting the double results into standard results for display when the two types of judgment results are inconsistent. According to the method, the soil layer name is presumed by adopting a real-time dynamic method, the detected sample does not need to be remodeled, the recognition accuracy and efficiency are improved, the model is fed back and optimized according to a new test set, the prediction precision is kept, and the detection speed is improved.

Description

Technical Field

[0001] This invention belongs to the field of engineering geology, and specifically relates to an engineering geological soil layer name determination system, method and equipment mounted on a static cone penetrometer. Background Technology

[0002] Spectroscopic sensors are primarily based on the principles of spectroscopy, utilizing the absorption, emission, or scattering properties of light by substances to acquire information. Different substances have different electron movements within their atoms, resulting in different emitted or absorbed light waves, thus forming the characteristic spectra of those substances. This technology can be used to analyze the physical structure, chemical composition, and other indicators of objects. Spectroscopic sensors have a wide range of applications, including but not limited to chemical analysis, environmental monitoring, biomedicine, and industrial inspection. In chemical analysis, spectroscopic sensors can be used to identify the components of compounds; in environmental monitoring, they can be used to detect pollutants in the atmosphere or water; and in the medical field, spectroscopic imaging technology can aid in disease diagnosis.

[0003] In engineering geology, soil layers are named based on their physical, chemical, and mechanical properties, as well as their environment and engineering requirements. A soil layer is a layer of soil roughly parallel to the ground surface in a soil profile, possessing unique physical, chemical, and biological characteristics. Soil formation is influenced by various factors, including soil-forming processes, topography, climate, and human activities. Each soil layer differs in color, structure, and texture, reflecting variations in its characteristics and composition. In engineering geology, soil layers are named by taking samples from them for classification.

[0004] In engineering geology, soil is classified according to its particle size and distribution into categories such as silt, fine sand, medium sand, coarse sand, angular gravel, and rounded gravel. Based on different plasticity indices, soil can be classified into silt, silty clay, and clay. Different regions and industries may use different classifications for soil, but the plasticity index and particle size distribution are the primary indicators. Plasticity index (I...) P The liquid limit (ω) is the threshold water content at which soil transitions from a plastic to a semi-solid state. This index is determined by the liquid limit (ω). L ) and plastic limit (ω) P It is calculated from the difference between I and 2. The specific formula is: I P =ω L -ω P .

[0005] The reflectance spectral characteristics of soil can comprehensively reflect its physicochemical properties and internal structure. Different soils exhibit different spectral characteristics, and soils of the same type within the same region show remarkable similarity in their spectral characteristics. This makes it possible to classify soil layers using spectral data. Spectroscopic analysis is a non-destructive testing method that does not require complex pretreatment of the soil and can be performed without altering its original state. This method is not only low-cost but also fast, providing a novel method for soil classification and naming in engineering geology.

[0006] This invention proposes an engineering geological soil layer name determination system, method, and equipment mounted on a static cone penetration test instrument. Summary of the Invention

[0007] To address the aforementioned problems in the prior art, namely that existing methods for estimating soil layer names involve soil layer damage, complex soil pretreatment, high costs, and slow speed, this invention provides an engineering geological soil layer name determination system, method, and device mounted on a static cone penetrometer.

[0008] In a first aspect, the present invention provides an engineering geological soil layer name determination system mounted on a static cone penetrometer, comprising a spectral acquisition module, a data processing module, and a data output module connected in sequence. The spectral acquisition module includes a multispectral sensor and a self-luminous source. The self-luminous source is used to provide stable lighting conditions for the soil layer, and the multispectral sensor is used to receive light reflected or transmitted from the soil layer, convert the spectrum of the light at multiple wavelengths into electrical signals, and use them as spectral data to be judged. The spectral acquisition module is embedded in the spectral acquisition section. The outer diameter of the spectral acquisition section matches the static cone probe rod. The lower end is screwed to the static cone probe, and the upper end is screwed to the static cone probe rod. The data processing module, based on a neural network model, judges the spectral data to be judged and obtains the discrimination result based on the spectral data. The data processing module is also used to receive static penetration data collected by the static penetration probe, the static penetration data including at least end resistance, side resistance and friction ratio, and input the static penetration data into the regression prediction algorithm to obtain the discrimination result based on static penetration. The data output module is used to output a single discrimination result and convert it into a preset result display when the discrimination result based on spectral data is consistent with the discrimination result based on static penetration. When the discrimination results based on spectral data are inconsistent with those based on static cone penetration, two discrimination results are output and converted into preset results for display.

[0009] Furthermore, the spectral acquisition module is embedded in the spectral acquisition section. The outer diameter of the spectral acquisition section matches the static cone penetration probe rod. The lower end is screwed to the static cone penetration probe, and the upper end is screwed to the static cone penetration probe rod. The static cone penetration probe is used to acquire static cone penetration data, which includes at least end resistance, side resistance, and friction ratio. The data processing module is used to receive the static penetration data, input the static penetration data into the regression prediction algorithm, and obtain the discrimination result based on the static penetration test. The data output module is used to input a single discrimination result when the discrimination result based on spectral data is consistent with the discrimination result based on static probe, and convert it into a preset result display. When the discrimination result based on spectral data is inconsistent with the discrimination result based on static probe, it outputs the discrimination result and converts it into a preset result display respectively.

[0010] Furthermore, the pre-defined results display includes the soil name, type estimation results, or characteristic spectral curves.

[0011] Furthermore, the data processing module includes: The system includes a data preprocessing unit, a feature extraction unit, a neural network construction unit, a model training and optimization unit, a model evaluation and validation unit, and a data decision unit. The data preprocessing unit is configured to obtain parameter data of various soil layers based on geotechnical tests and perform cleaning and preprocessing. The feature extraction unit is configured to extract features corresponding to spectral characteristics from the cleaned and preprocessed parameter data as a dataset. Neural network building units are configured to pre-build neural network models; The model training and optimization unit is configured to train the neural network model using the dataset as ground truth labels to obtain prediction results, until the difference between the ground truth labels and the prediction results is within a preset range, at which point training stops and the trained neural network model is obtained; the trained neural network model is then optimized based on a pre-built test set to obtain an optimized neural network model. The data determination unit is configured to input the spectral data to be determined into the optimized neural network model to obtain a discrimination result based on the spectral data.

[0012] Furthermore, the data determination unit is also configured to compare the discrimination result based on spectral data with the discrimination result corresponding to the geotechnical test to obtain a data difference. When the data difference is greater than a preset threshold, it jumps to the model training and optimization unit, replaces the dataset with the dataset of the current region, and retrains the pre-built neural network model based on the dataset of the current region.

[0013] Furthermore, the data processing module also includes a static cone penetration data prediction unit; The static penetration test data prediction unit is configured to acquire static penetration test data and, in conjunction with a regression prediction algorithm, obtain a discrimination result based on the static penetration test. : ; in, For end resistance, For lateral resistance, Friction ratio, The regression coefficients are determined and dynamically updated by fitting regional experimental data through a neural network.

[0014] Furthermore, based on geotechnical tests, parameter data for various soil layers are obtained, including at least the soil name, void ratio, water content, organic matter content, liquid limit index, and plastic limit index.

[0015] In a second aspect, the present invention proposes a method for determining the names of engineering geological soil layers mounted on a static cone penetrometer. Based on a system for determining the names of engineering geological soil layers mounted on a static cone penetrometer, the method includes the following steps: Step S100: Embed the spectral acquisition module into the spectral acquisition section, so that the outer diameter of the spectral acquisition section matches the static cone penetration test rod, the lower end is screwed to the static cone penetration test probe, and the upper end is screwed to the static cone penetration test rod; provide stable lighting conditions for the soil layer through a self-luminous source; The light reflected or transmitted by the soil layer is received by a multispectral sensor, and the multi-wavelength spectrum of the light is converted into an electrical signal as spectral data to be judged. Static cone penetration data is collected simultaneously using a static cone penetration probe, and the static cone penetration data includes at least end resistance, side resistance, and friction ratio. Step S200: The spectral data to be judged is input into the optimized neural network model to obtain the discrimination result based on the spectral data; the static cone penetration data is input into the regression prediction algorithm to calculate the discrimination result based on the static cone penetration. In step S300, if the discrimination result based on spectral data is consistent with the discrimination result based on static cone penetration test, a single discrimination result is generated; if the discrimination result based on spectral data is inconsistent with the discrimination result based on static cone penetration test, spectral data discrimination result and static cone penetration test discrimination result are generated separately.

[0016] In a third aspect, the present invention provides an electronic device comprising: At least one processor; and A memory communicatively connected to at least one of the processors; wherein, The memory stores instructions that can be executed by the processor to implement a method for determining the name of engineering geological soil layers mounted on a static cone penetrometer.

[0017] In a fourth aspect, the present invention provides a computer-readable storage medium storing computer instructions for execution by a computer to implement a method for determining the name of engineering geological soil layers mounted on a static cone penetrometer.

[0018] The beneficial effects of this invention are: (1) The system utilizes a combination of a highly sensitive multispectral sensor and a self-luminous source, without damaging the soil structure, to ensure stable and high-quality spectral data under different environmental conditions. This stable illumination condition and wide-band spectral analysis capability significantly improve the accuracy and efficiency of soil layer identification, effectively distinguishing different soil types, and accurately capturing even minute changes in soil composition.

[0019] (2) Through a pre-built neural network model, the system can automatically learn and understand the relationship between complex and varied spectral features and soil types, thus achieving intelligent analysis. The training process is based on a large dataset, and the optimized model can quickly and accurately classify unknown soil samples, reducing the subjectivity and error of human judgment and improving the intelligence level of the inference system.

[0020] (3) The data processing module not only trains the model based on the initial dataset, but also continuously optimizes the model performance based on feedback from new test sets through a continuous learning mechanism. This means that the system has good adaptability and can improve itself as time goes by and new data is added, maintaining high prediction accuracy and improving detection speed and efficiency.

[0021] (4) The data output module transforms complex analysis results into an easy-to-understand form, such as directly giving the name and type of soil layer, or displaying characteristic spectral curves, so that researchers and geologists can intuitively grasp the soil characteristics and provide direct and effective information support for subsequent soil research, geological exploration, environmental assessment and other work. Attached Figure Description

[0022] Other features, objects, and advantages of this application will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings: Figure 1 This is a schematic diagram of the processing flow of an engineering geological soil layer name determination system mounted on a static cone penetration test instrument according to the present invention. Figure 2This is a schematic diagram of the spectral acquisition module in an engineering geological soil layer name determination system mounted on a static cone penetrometer, according to the present invention. Figure 3 This is a schematic diagram of the spectral acquisition section structure and mounting method of an engineering geological soil layer name determination system mounted on a static cone penetrometer, according to the present invention. Detailed Implementation

[0023] The present application will now be described in further detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are for illustrative purposes only and are not intended to limit the invention. Furthermore, it should be noted that, for ease of description, only the parts relevant to the invention are shown in the accompanying drawings.

[0024] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. This application will now be described in detail with reference to the accompanying drawings and embodiments.

[0025] like Figures 1-3 As shown, the present invention provides an engineering geological soil layer name determination system mounted on a static cone penetrometer, comprising a spectral acquisition module, a data processing module and a data output module connected in sequence; The spectral acquisition module includes a self-emitting light source 1 and a multispectral sensor 2; The self-luminous source 1 is used to provide stable lighting conditions for the soil layer; The multispectral sensor 2 is used to receive light reflected or transmitted from the soil layer, convert the spectrum of the light at multiple wavelengths into electrical signals, and use them as spectral data to be judged. The data processing module is used to pre-build a neural network model, train the neural network model using the acquired dataset, optimize the trained neural network model based on the test set, and input the spectral data to be judged into the optimized neural network model to obtain the discrimination result. The data output module is used to convert the discrimination result into a preset result and display it; wherein, the preset result includes the soil name and type estimation result or characteristic spectral curve.

[0026] To more clearly explain the engineering geological soil layer naming system mounted on a static cone penetrometer according to the present invention, the following description is in conjunction with... Figure 1 The following is a detailed description of each module in the embodiments of the present invention: The data output module includes a user interface for displaying results according to user needs and providing auxiliary information for dividing soil strata.

[0027] The self-luminous source 1 of this invention emits light that is reflected off the soil being tested, and the reflected spectrum is received by the multispectral sensor 2. The self-luminous source 1 ensures high-quality spectral data acquisition even in the absence of light. The self-luminous source reduces the influence of external light; it is activated during spectral testing, coordinating with the sampling time of the multispectral sensor 2. This system can be mounted on geotechnical drilling and testing equipment, enabling soil identification and classification in the absence of natural light.

[0028] The self-luminescent source 1 of the present invention is arranged on both sides of the multispectral sensor 2.

[0029] The multispectral sensor 2 in this invention is the core component of the spectral acquisition module. It is responsible for receiving light reflected or transmitted from the surface of the soil being measured, decomposing it into spectra of different wavelengths, and then converting them into electrical signals. After processing, these electrical signals can provide information about the composition and properties of the soil being measured.

[0030] The spectral acquisition module, data processing module, and data output module are connected by a power supply and communication line 11.

[0031] The main function of the data output module of this invention is to provide the processed data to external systems or users in the format and manner required by the user. The aforementioned preset results include estimating and displaying the names and types of soil layers according to user-defined naming rules; and, during continuous stratigraphic section testing, outputting characteristic spectral curves as needed by the user, using the differences in characteristic spectra to help the user classify soil layer sequences.

[0032] In this embodiment, the self-emitting light source 1 integrates ultraviolet, visible, and near-infrared light, and is used for detection based on a multi-band light source control method, specifically: Step A1: Obtain initial spectral data of the soil layer by irradiating it with a multi-band light source; Step A2: Based on the band response characteristics in the initial spectral data, dynamically select and optimize the illumination mode: Step A21: When the first type of spectral feature is detected, activate the combined ultraviolet and near-infrared irradiation mode and adjust the emission parameters; Step A22: When the second type of spectral feature is detected, the high-energy pulse mode in the visible light band is activated; Step A3: Obtain enhanced spectral data based on the optimized illumination pattern for use by the soil layer type identification model.

[0033] The first type of spectral features includes the correlation between ultraviolet absorption features and near-infrared reflectance abrupt changes. Specifically, it refers to the situation where soil organic matter exhibits strong absorption characteristics in a specific ultraviolet band (e.g., 250-400 nm) and simultaneously shows discontinuous abrupt changes in reflectance in the near-infrared band (e.g., 1300-2500 nm), forming a statistically significant coupled response pattern.

[0034] The second type of spectral features includes the continuity of the reflectance curve shape in the visible light band. Specifically, it refers to the smooth, gradual change of the soil reflectance curve in the visible light range (400-700 nm).

[0035] The emission parameter adjustments include increasing the intensity of the ultraviolet band and / or setting the near-infrared pulse duty cycle; the spectral data from the combined irradiation mode are used for quantitative analysis of organic matter content.

[0036] The specific calculation method for increasing the intensity of the ultraviolet band is as follows: Step A211: Determine the initial reference intensity of the ultraviolet light source based on the signal-to-noise ratio requirements of the multispectral sensor and the optical properties of the soil layer; The initial reference intensity is calibrated through pre-experiments, specifically by iterative testing in samples with different organic matter contents until a stable spectral response is obtained. Establish a baseline strength adjustment curve that is negatively correlated with soil moisture content.

[0037] Step A212: Real-time acquisition of characteristic wavelength absorbance and reflectance data of the target soil layer; Step A213: Generate an ultraviolet intensity correction coefficient based on the degree of deviation of the absorbance and reflectance data from the preset threshold. The calculation of the correction coefficient includes a combined operation of the positive absorbance compensation term and the negative reflectance compensation term; When an oxide layer is detected on the soil surface, the nonlinear enhancement factor is activated.

[0038] Step A214: Limit the corrected ultraviolet intensity within a preset ratio range and perform dynamic attenuation compensation by associating it with the probe temperature parameters; The upper limit of the preset ratio range is dynamically adjusted according to the heat dissipation capacity of the light source module; The dynamic attenuation compensation includes establishing an exponential relationship model between the temperature and intensity reduction coefficients.

[0039] Step A215: Adjust the calculation weight of the correction coefficient in reverse based on historical recognition accuracy data.

[0040] Analyze the changing trend of soil layer identification accuracy under specific strength parameters using machine learning models; When the accuracy improvement rate is lower than the set threshold after three consecutive adjustments, the benchmark strength recalibration process is triggered.

[0041] Near-infrared light sources (such as LEDs or lasers) generate significant heat when operating at continuous high power, leading to wavelength drift, reduced efficiency, and even hardware damage.

[0042] Therefore, this invention reduces average power consumption and heat accumulation by periodically switching the light source through pulse duty cycle control, specifically including: Step B1: Determine the initial pulse duty cycle of the near-infrared light source based on the soil type and the performance requirements of the spectral sensor. The initial duty cycle was calibrated through pre-experimentation, and baseline values ​​for the duty cycle were established for sandy soil, cohesive soil, and organic soil respectively. The baseline value is negatively correlated with the light transmittance characteristics and particle size distribution of the soil layer.

[0043] Step B2: Monitor the spectral signal-to-noise ratio and light source module temperature in the near-infrared band in real time to determine whether the preset adjustment conditions are met. The preset adjustment conditions include a continuous decrease rate of spectral signal-to-noise ratio exceeding a set gradient value; The safe threshold for the temperature of the light source module is dynamically compensated based on the ambient temperature.

[0044] Step B3, duty cycle optimization adjustment: Step B31: When the spectral signal-to-noise ratio is detected to be lower than the preset threshold, the pulse duty cycle is increased according to the first adjustment rule; The first adjustment rule includes a nonlinear incremental algorithm based on the signal-to-noise ratio change rate: Suppose that when the current signal-to-noise ratio decreases at a rate of v, the duty cycle adjustment ΔD satisfies: ;in, The preset calibration coefficient is v, which is greater than the rate threshold that triggers the adjustment.

[0045] Each time the duty cycle is increased, a spectral data quality verification is triggered. If the verification fails, the data reverts to the state before adjustment.

[0046] Step B32: When the temperature of the light source module is detected to exceed the safety threshold, the pulse duty cycle is reduced according to the second adjustment rule. Step B4: Dynamically adjust the duty cycle based on the real-time identified soil physical characteristics. When identified as a high-density cohesive soil layer, the step-by-step increase strategy of the duty cycle is activated: Let the temperature of the light source module be T, and the decrease in duty cycle be... ; ,in, pThis is the preset attenuation coefficient.

[0047] Step B5: Limit the duty cycle within a preset range and optimize the pulse timing in conjunction with the light source lifetime parameters.

[0048] When identified as a high-density cohesive soil layer, the step-by-step increase strategy of the duty cycle is activated: The initial duty cycle is D0, preferably 30% in this embodiment, with a single-step increase of 10% and a duration of 2 seconds per step, until the signal-to-noise ratio reaches the threshold. This avoids overheating of the light source due to instantaneous high power and allows the sensor to adapt to the gradually increasing signal.

[0049] When the soil layer is identified as loose sandy soil, the duty cycle smooth transition mode is enabled to prevent signal oscillation. When it is necessary to change the duty cycle of the light source (such as from 40% to 60%), instead of jumping directly to the target value, it is broken down into multiple small adjustments (such as increasing by only 1-3% each time).

[0050] After each fine-tuning, wait for a short period of stabilization (usually 0.1-0.5 seconds) until the optical signal is stable before proceeding to the next adjustment.

[0051] For loose sandy soil (with large interparticle gaps and easy scattering of light), a slower adjustment rate is used: Low-density sandy soil: the increase per step ≤ 1.5%, and the interval between steps ≥ 0.5 seconds; High-density sand: increase per step ≤3%, interval between steps ≥0.3 seconds.

[0052] After each duty cycle adjustment, the spectral acquisition is paused for 1.0 second to avoid transient scattering noise from sand particles caused by sudden changes in illumination.

[0053] When resuming data acquisition, apply real-time filtering to the first batch of data (e.g., take the average of 5 consecutive samples) to suppress residual fluctuations.

[0054] For further explanation of the present invention, see [link to relevant documentation]. Figure 2 The spectral acquisition module also includes a light-transmitting protective cover 3 and a circuit board 4; The light-transmitting protective cover 3 houses and protects the self-emitting light source 1, the multispectral sensor 2, and the circuit board 4. The circuit board 4 is connected to the multispectral sensor 2 and the terminal, respectively, and is used to realize the transmission and conversion of current and signals between the multispectral sensor 2 and the terminal.

[0055] The circuit board 4 in the spectral acquisition module is connected to the power supply and communication line 11. The circuit board 4 realizes the transmission and conversion of current and signals between the multispectral sensor 2 and the outside world. It provides mechanical protection and support for the multispectral sensor 2 and the self-emitting light source 1, and constitutes the supporting carrier of the electronic circuit.

[0056] This system trains a neural network model by collecting a large amount of spectral data from different soil layers, and then deploys the neural network model on the system's GPU. When a new user uses this system, they place the surface of the soil to be tested against the light-transmitting protective cover 3 to begin spectral data acquisition. During spectral data acquisition, light emitted from the self-emitting light source 1 passes through the light-transmitting protective cover 3, is reflected back after passing through the soil sample surface, and the multispectral sensor 2 collects the reflected spectral information. The collected spectral information is then transmitted to the data judgment unit of the data processing module via the power supply and communication line 11. After analysis, the data judgment unit transmits the results to the data output module for display.

[0057] For further explanation of the present invention, see [link to relevant documentation]. Figure 1 The data processing module includes a data preprocessing unit, a feature extraction unit, a neural network construction unit, a model training and optimization unit, a model evaluation and verification unit, and a data determination unit. The data preprocessing unit is configured to obtain parameter data of various soil layers based on geotechnical tests and perform cleaning and preprocessing. The feature extraction unit is configured to extract features corresponding to spectral characteristics from the cleaned and preprocessed parameter data as a dataset. Neural network building units are configured to pre-build neural network models; The model training and optimization unit is configured to train the neural network model using the dataset as ground truth labels to obtain prediction results, until the difference between the ground truth labels and the prediction results is within a preset range, at which point training stops and the trained neural network model is obtained; the trained neural network model is then optimized based on a pre-built test set to obtain an optimized neural network model. The data determination unit is configured to input the spectral data to be determined into the optimized neural network model to obtain the determination result.

[0058] This involves extracting features corresponding to the spectral characteristics from the cleaned and preprocessed parameter data. These features reflect the spectral properties of the tested soil and are used to obtain the optimal weights and thresholds for various soil types. Commonly used feature extraction methods include principal component analysis and linear discriminant analysis.

[0059] This involves training the neural network model using a known dataset. This process includes operations such as weight updates and bias adjustments to minimize the difference between the predicted and actual results. Through iterative training, the model's performance is gradually improved.

[0060] This invention evaluates the model's performance on an independent test set to examine its generalization ability. Based on the evaluation results, the model is adjusted and optimized to improve its predictive accuracy on new data.

[0061] In this invention, the data determination unit provides a way to strengthen and retrain the neural network model for regional characteristics of soil. The specific method includes: comparing the discrimination result corresponding to the spectral data to be judged with the discrimination result corresponding to the geotechnical test corresponding to the spectral data to be judged to obtain the data difference; when the data difference is greater than a preset threshold, jumping to the model training and optimization unit, replacing the dataset with the dataset of the current region, training the pre-constructed neural network model based on the dataset of the current region, re-extracting the optimal weights and thresholds of each type of soil in the region, and solidifying the training results of the neural network model in the data determination unit to improve the regional applicability of this invention.

[0062] When retraining a pre-built neural network model based on the dataset of the current region, the neural network model needs to be retrained based on the geotechnical test results of the soil samples in the region and the spectral data of the soil samples in order to improve the accuracy of the regional test.

[0063] The reason for the above-mentioned content of this invention is that the formation of soil mainly involves natural processes such as rock weathering, transportation, and deposition. Primary minerals in soil, such as quartz, feldspar, and muscovite, as well as secondary minerals such as hematite and pyrite, all affect the spectral characteristics of soil. The spectral characteristics of soil are affected by factors such as soil-forming minerals, water content, organic matter, and depositional environment. In addition, the absorption spectra of organic matter in soil are mainly generated in the visible-near infrared and mid-infrared bands. These spectral characteristics make it difficult for the fixed model in this system to be universally applicable.

[0064] Therefore, this invention provides a method for reinforcing and retraining neural network models to address the regional characteristics of soil. Users in different regions can use this method and system to verify and test the identification and spectral characteristics of samples within their region. The test results of this method and system can be compared with those from geotechnical laboratories. When there are significant differences in the test results of soil samples within a region, regional data can be used to reinforce and train the existing model.

[0065] This invention applies the spectral characteristics of soil to the field of engineering geology for soil naming and classification, which is an important innovation in engineering geology. Previously, there was little accumulated data and the data standards were not uniform. The data collection and preprocessing method of this system is as follows: (1) Spectral testing is performed on the test sample in the engineering geology geotechnical laboratory, and the sample number is recorded; (2) After the geotechnical laboratory completes the routine geotechnical test of the sample, the experimental results of the sample are collected. After the experimental results are analyzed by data fitting of the neural network model, the parameter data of various soil layers are obtained. The parameter data is cleaned and preprocessed. The parameter data includes at least soil naming, void ratio, water content, organic matter content, liquid limit index and plastic limit index.

[0066] As a further explanation of the present invention, the method for cleaning and preprocessing the parameter data includes at least noise removal, background interference elimination, and normalization processing to ensure the quality and accuracy of the data.

[0067] As a further explanation of the present invention, the method for extracting features corresponding to spectral characteristics from cleaned and preprocessed parameter data includes principal component analysis or linear discriminant analysis.

[0068] For further explanation of the present invention, see [link to relevant documentation]. Figure 2 The light-transmitting protective cover 3 is composed of multiple detachable and fixedly connected light-transmitting plates.

[0069] The light-transmitting protective cover 3 is made of high-transmittance, high-hardness glass (light-transmitting plate), and its main function is to protect the internal self-luminous source 1 and spectral sensor 2 from damage caused by direct contact with external factors such as dust and liquids. The light-transmitting plates on both sides of the light-transmitting protective cover 3 are fixed with screws, and can be disassembled and replaced from the outside when the light-transmitting protective cover 3 is severely scratched and affects the light transmission effect.

[0070] See Figure 3 In this embodiment, the spectral acquisition module is embedded in the spectral acquisition section 13. The outer diameter of the spectral acquisition section 13 matches the static cone penetration probe 12. The lower end is screwed to the static cone penetration probe 14, and the upper end is screwed to the static cone penetration probe 12. The static cone penetration probe 14 is used to acquire static cone penetration data. The static cone penetration data includes at least end resistance, side resistance, and friction ratio. The data processing module is used to receive the static penetration data, input the static penetration data into the regression prediction algorithm, and obtain the discrimination result based on the static penetration test. The data output module is used to input a single discrimination result when the discrimination result based on spectral data is consistent with the discrimination result based on static probe, and convert it into a preset result display. When the discrimination result based on spectral data is inconsistent with the discrimination result based on static probe, it outputs the discrimination result and converts it into a preset result display respectively.

[0071] The data determination unit is further configured to compare the discrimination result based on spectral data with the discrimination result corresponding to the geotechnical test to obtain a data difference. When the data difference is greater than a preset threshold, it jumps to the model training and optimization unit, replaces the dataset with the dataset of the current region, and retrains the pre-built neural network model based on the dataset of the current region.

[0072] The static cone penetration probe 12 is powered by a static cone penetration device or other in-situ engineering geological testing device to carry the spectral acquisition section 13 into the soil layer.

[0073] In this embodiment, the static penetration test data prediction unit in the data processing module calculates the discrimination result of the static penetration test, and the method is as follows: The static penetration test data prediction unit is configured to acquire static penetration test data and, in conjunction with a regression prediction algorithm, obtain a discrimination result based on the static penetration test. : ; in, For end resistance, For lateral resistance, Friction ratio, The regression coefficients are determined and dynamically updated by fitting regional experimental data through a neural network.

[0074] In the first embodiment of the present invention, when the discrimination result based on spectral data is inconsistent with the discrimination result based on static probe, the discrimination result is output and converted into a preset result display respectively.

[0075] This invention integrates spectral analysis technology with static cone penetration data (end resistance, lateral resistance, and friction ratio) for dual detection, enabling complementary verification of multi-dimensional geological information. When the two judgment results are consistent, a single conclusion is output, significantly improving the accuracy of stratigraphic classification; when they are inconsistent, the different results are displayed separately, providing more comprehensive data support for engineering decisions and reducing the risk of misjudgment.

[0076] This invention uses a neural network model driven by regional experimental data to fit regression coefficients. Furthermore, a threshold triggering mechanism is introduced: when the discriminant result deviates significantly from the geotechnical test data, the system automatically switches to the current regional dataset for model retraining. This design greatly improves the system's adaptability to different geological conditions, ensuring continuous optimization of prediction accuracy during long-term monitoring.

[0077] The spectral acquisition section and static cone penetration test probe of this invention adopt a standardized screw-in structure, with an outer diameter matching existing probe systems, achieving seamless integration of the spectral detection module. This integrated design eliminates the need for additional drilling equipment, is compatible with conventional static cone penetration test devices, significantly simplifies on-site operation procedures, and reduces the cost of multi-device collaboration.

[0078] This invention automatically converts spectral data and static cone penetration test results into engineering charts or geological profiles using preset result display templates, supporting visualized result comparison. The dual-output mode of the discrepancy results provides cross-validation evidence for complex stratigraphic analysis, assisting technicians in quickly locating anomalous sections and improving the efficiency of exploration decision-making.

[0079] This invention, through regression prediction algorithm and real-time data closed-loop verification mechanism, can reduce the frequency of laboratory soil sample analysis and shorten the exploration cycle while ensuring accuracy. It is especially suitable for large-area regional surveys or engineering scenarios with high timeliness requirements.

[0080] Based on this, the present invention provides another discrimination method when the discrimination result based on spectral data is inconsistent with the discrimination result based on static probe, as follows: Based on the preset confidence level calculation rules, the first confidence level of the spectral data is calculated respectively. Second confidence level of static cone penetration data ; First confidence level The calculation method is as follows: ;in, The predicted value corresponding to the spectral data output by the data determination unit. The environmental disturbance factor is calculated by weighting the light intensity fluctuation rate, humidity deviation coefficient and dust obstruction rate. Second confidence level The calculation method is as follows: ;in, The variances of end resistance and side resistance. w This represents the probe wear coefficient.

[0081] when >When setting a preset threshold, select the judgment result with high confidence as the final output; when When the threshold is less than or equal to a preset threshold, the spectral feature data and the static cone penetration data are input into a joint classifier for multimodal fusion to generate a fusion probability distribution; wherein, the spectral feature data includes feature spectral curves; the joint classifier is preferably a random forest or a deep cross network; If the maximum value of the fusion probability distribution exceeds a preset probability threshold, the corresponding category is output as the final result; Otherwise, a regionalized historical geological database is invoked for geographic matching, and the output is corrected based on the matching results: Access the historical geological database for the current location and match similar static cone penetration data. The corresponding typical soil layer name.

[0082] If the historical matching rate is greater than 80%, historical results will be used first.

[0083] If the system still cannot determine the cause, it will prompt the user to collect a small amount of soil sample for laboratory testing.

[0084] The laboratory results are fed back to the model to dynamically update the dataset and classification rules for the current region.

[0085] Light intensity fluctuation rate in this embodiment The calibrated light intensity is a preset value of the equipment under ideal laboratory conditions. Humidity deviation coefficient The ambient humidity is obtained based on a humidity sensor embedded in a spectral acquisition section.

[0086] The actual transmittance is measured in real time by the multispectral sensor self-test module.

[0087] ;in, The weighting coefficients are obtained through experimental calibration and are used in this embodiment. The values ​​are 0.5, 0.3, and 0.2, respectively.

[0088] The probe wear coefficient W The calculation method is as follows: W=max (Wear due to usage frequency and calibration error); Wherein, frequency of wear = number of times used / design life number of times; Calibration error wear = Current calibration error / Maximum permissible error; In this embodiment, if the spectral data and the static cone penetration data are inconsistent multiple times consecutively, a self-test program is initiated: Check the stability of the light source (such as voltage fluctuations) and the cleanliness of the light-transmitting protective cover.

[0089] Verify the calibration parameters (such as regression coefficients) of the static cone penetrometer 14. β 0− β 3).

[0090] This invention constructs a dynamic confidence calculation model by introducing environmental interference factors and probe wear coefficients. This study quantifies the real-time impact of light, humidity, and dust on spectral data, as well as the mechanical wear state of the probe. It is based on a dual-confidence difference threshold. The intelligent arbitration strategy prioritizes the judgment results with high confidence, effectively reducing the probability of misjudgment caused by environmental noise or equipment aging, and improving the system robustness under extreme working conditions.

[0091] When the difference between the two confidence levels is small, this invention employs a joint classifier of random forest / deep cross-network to analyze the spectral curve and probe parameters. Feature-level fusion is performed to generate a fusion probability distribution. High-confidence output is triggered by a preset probability threshold (e.g., >85%), which breaks through the classification bottleneck of single data sources in heterogeneous soil layers (such as gravelly clay and sandy layers) and significantly improves the recognition accuracy of complex geological interfaces.

[0092] When the fusion determination fails, a regionalized historical geological database is invoked for similarity matching (e.g., Euclidean distance matching). (Sequence), prioritizing typical soil layer results with a historical matching degree >80%. This mechanism fully utilizes regional geological regularities, reduces repetitive manual interpretation, and is particularly suitable for sedimentary plains with strong stratigraphic continuity, shortening the decision-making response time for anomalous strata.

[0093] When the system cannot make a judgment autonomously, gold standard data is introduced by minimizing soil sampling (a small number of laboratory tests), and the results are fed back to the model for dynamic training. This closed-loop design realizes continuous optimization of "field perception - laboratory verification - model iteration", gradually building a region-specific classification rule base and significantly reducing the application threshold for cross-regional projects.

[0094] This invention addresses continuous inconsistent alarms by automatically triggering light source stability detection (voltage fluctuation analysis), light-transmitting cover cleanliness assessment, and probe calibration parameter verification. Through predictive maintenance (such as prompting for cleaning the light-transmitting cover or replacing worn probes), it avoids data drift issues caused by hardware failure, extending the effective operating time of the equipment in harsh environments (high dust, high humidity).

[0095] This invention precisely quantifies the impact of external conditions on spectral acquisition by decomposing light intensity fluctuation rate, humidity deviation coefficient, and dust obstruction rate, and then weighting and superimposing them to generate an environmental interference factor. Combined with real-time data quality correction through transmittance self-checking, the spectral discrimination results maintain high reliability even in open-air engineering scenarios (such as coastal mudflats and deserts).

[0096] The probe wear coefficient of this invention employs a dual wear assessment (usage frequency wear + calibration error wear). When the calibration error approaches a threshold, it automatically prompts for recalibration, avoiding the risks of premature scrapping or exceeding usage limits caused by traditional single-lifespan counting. Combined with dynamic updates of regression coefficients, it achieves collaborative optimization of "hardware wear and algorithm compensation," reducing long-term monitoring costs.

[0097] A method for determining the name of engineering geological soil layers mounted on a static cone penetrometer, according to a second embodiment of the present invention, is based on a system for determining the name of engineering geological soil layers mounted on a static cone penetrometer. The method includes the following steps: Step S100: Receive light reflected or transmitted from the soil layer, convert the spectrum of the light at multiple wavelengths into an electrical signal, and use it as spectral data to be judged; Step S200: Pre-build a neural network model, train the neural network model using the acquired dataset, optimize the trained neural network model based on the test set, and input the spectral data to be judged into the optimized neural network model to obtain the discrimination result; Step S300: Convert the discrimination result into a preset result and display it; wherein, the preset result includes the soil name and type estimation result or characteristic spectral curve.

[0098] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working process and related explanations of the methods described above can be found in the corresponding processes in the foregoing system embodiments, and will not be repeated here.

[0099] Although the steps in the above embodiments are described in the above order, those skilled in the art will understand that in order to achieve the effect of this embodiment, different steps do not need to be executed in such an order. They can be executed simultaneously (in parallel) or in a reverse order. These simple variations are all within the protection scope of this invention.

[0100] An electronic device according to a third embodiment of the present invention includes: At least one processor; and A memory communicatively connected to at least one of the processors; wherein, The memory stores instructions that can be executed by the processor to implement the above-described method for determining the name of engineering geological soil layers mounted on a static cone penetrometer.

[0101] A computer-readable storage medium according to a fourth embodiment of the present invention stores computer instructions, which are executed by the computer to implement the above-described method for determining the name of engineering geological soil layers mounted on a static cone penetrometer.

[0102] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working process and related descriptions of the storage device and processing device described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0103] Those skilled in the art will recognize that the modules and method steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. The programs corresponding to the software modules and method steps can be placed in random access memory (RAM), main memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disks, removable disks, CD-ROMs, or any other form of storage medium known in the art. To clearly illustrate the interchangeability of electronic hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in electronic hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of the invention.

[0104] The terms “first”, “second”, etc., are used to distinguish similar objects, not to describe or indicate a specific order or sequence.

[0105] The term "comprising" or any other similar term is intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus / device that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent in such process, method, article, or apparatus / device.

[0106] The technical solution of the present invention has been described above with reference to the preferred embodiments shown in the accompanying drawings. However, it will be readily understood by those skilled in the art that the scope of protection of the present invention is obviously not limited to these specific embodiments. Without departing from the principles of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will all fall within the scope of protection of the present invention.

Claims

1. A system for determining the name of engineering geological soil layers mounted on a static cone penetrometer, characterized in that, It includes a spectral acquisition module, a data processing module, and a data output module connected in sequence; The spectral acquisition module includes a multispectral sensor and a self-luminous source. The self-luminous source is used to provide stable lighting conditions for the soil layer, and the multispectral sensor is used to receive light reflected or transmitted from the soil layer, convert the spectrum of the light at multiple wavelengths into electrical signals, and use them as spectral data to be judged. The spectral acquisition module is embedded in the spectral acquisition section. The outer diameter of the spectral acquisition section matches the static cone probe rod. The lower end is screwed to the static cone probe, and the upper end is screwed to the static cone probe rod. The data processing module, based on a neural network model, judges the spectral data to be judged and obtains the discrimination result based on the spectral data. The data processing module is also used to receive static penetration data collected by the static penetration probe, the static penetration data including at least end resistance, side resistance and friction ratio, and input the static penetration data into the regression prediction algorithm to obtain the discrimination result based on static penetration. The data output module is used to output a single discrimination result and convert it into a preset result display when the discrimination result based on spectral data is consistent with the discrimination result based on static penetration. When the discrimination results based on spectral data are inconsistent with those based on static cone penetration, two discrimination results are output and converted into preset results for display.

2. The engineering geological soil layer naming system mounted on a static cone penetration test instrument according to claim 1, characterized in that, The preset results display includes the soil name, type estimation results, or characteristic spectral curves.

3. The engineering geological soil layer naming system mounted on a static cone penetration test instrument according to claim 1, characterized in that, The spectral acquisition module also includes a light-transmitting protective cover and a circuit board; The light-transmitting protective cover houses and protects the self-emitting light source, the multispectral sensor, and the circuit board. The circuit board is connected to the multispectral sensor and the terminal, respectively, and is used to realize the transmission and conversion of current and signals between the multispectral sensor and the terminal.

4. The engineering geological soil layer naming system mounted on a static cone penetration test instrument according to claim 1, characterized in that, The data processing module includes: The system includes a data preprocessing unit, a feature extraction unit, a neural network construction unit, a model training and optimization unit, a model evaluation and validation unit, and a data decision unit. The data preprocessing unit is configured to obtain parameter data of various soil layers based on geotechnical tests and perform cleaning and preprocessing. The feature extraction unit is configured to extract features corresponding to spectral characteristics from the cleaned and preprocessed parameter data as a dataset. Neural network building units are configured to pre-build neural network models; The model training and optimization unit is configured to train the neural network model using the dataset as ground truth labels to obtain prediction results, until the difference between the ground truth labels and the prediction results is within a preset range, at which point training stops and the trained neural network model is obtained; the trained neural network model is then optimized based on a pre-built test set to obtain an optimized neural network model. The data determination unit is configured to input the spectral data to be determined into the optimized neural network model to obtain a discrimination result based on the spectral data.

5. The engineering geological soil layer naming system mounted on a static cone penetration test instrument according to claim 4, characterized in that, The data determination unit is further configured to compare the discrimination result based on spectral data with the discrimination result corresponding to the geotechnical test to obtain a data difference. When the data difference is greater than a preset threshold, it jumps to the model training and optimization unit, replaces the dataset with the dataset of the current region, and retrains the pre-built neural network model based on the dataset of the current region.

6. The engineering geological soil layer naming system mounted on a static cone penetration test instrument according to claim 2, characterized in that, The data processing module also includes a static cone penetration data prediction unit; The static penetration test data prediction unit is configured to acquire static penetration test data and, in conjunction with a regression prediction algorithm, obtain a discrimination result based on the static penetration test. : ； in, For end resistance, For lateral resistance, The friction ratio is... The regression coefficients are determined and dynamically updated by fitting regional experimental data through a neural network.

7. The engineering geological soil layer naming system mounted on a static cone penetration test instrument according to claim 4, characterized in that, Based on geotechnical tests, parameter data of various soil layers are obtained. The parameter data includes at least the soil name, void ratio, water content, organic matter content, liquid limit index, and plastic limit index.

8. A method for determining the name of engineering geological soil layers mounted on a static cone penetrometer, based on the engineering geological soil layer name determination system mounted on a static cone penetrometer as described in any one of claims 1-7, characterized in that, The method includes the following steps: Step S100: Embed the spectral acquisition module into the spectral acquisition section, so that the outer diameter of the spectral acquisition section matches the static cone penetration test rod, the lower end is screwed to the static cone penetration test probe, and the upper end is screwed to the static cone penetration test rod; provide stable lighting conditions for the soil layer through a self-luminous source; The light reflected or transmitted by the soil layer is received by a multispectral sensor, and the multi-wavelength spectrum of the light is converted into an electrical signal as spectral data to be judged. Static cone penetration data is collected simultaneously using a static cone penetration probe, and the static cone penetration data includes at least end resistance, side resistance, and friction ratio. Step S200: The spectral data to be judged is input into the optimized neural network model to obtain a discrimination result based on the spectral data; The static cone penetration data is input into a regression prediction algorithm to calculate the discrimination result based on the static cone penetration test. Step S300: If the discrimination result based on spectral data is consistent with the discrimination result based on static cone penetration, a single discrimination result is generated; If the discrimination results based on spectral data are inconsistent with the discrimination results based on static cone penetration, spectral data discrimination results and static cone penetration discrimination results are generated separately.

9. An electronic device, characterized in that, include: At least one processor; as well as A memory communicatively connected to at least one of the processors; wherein, The memory stores instructions that can be executed by the processor to implement the method for determining the name of engineering geological soil layers mounted on a static cone penetrometer as described in claim 8.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions that are executed by the computer to implement the method for determining the name of engineering geological soil layers mounted on a static cone penetrometer as described in claim 8.

Images