Device and method for detecting resolution of drill-borne laser radar

Abstract

The invention provides a device and method for detecting the resolution of a drill-borne laser radar, and relates to the field of laser radar performance detection, and the method comprises the steps: reproducing an extreme environment condition of a drilling site; the bionic target simulates a real rock stratum; laser radar signals are collected at high precision under extreme interference; the acquired signals are processed based on prior knowledge, and a resolution index is inverted; and based on the convolutional neural network, analyzing the resolution index to obtain a high-confidence resolution index, and carrying out intelligent grading evaluation on the detection precision of the laser radar. The problems that a traditional device cannot adapt to the extreme drilling environment and the detection result distortion rate is high are effectively solved, and accurate, reliable and intelligent detection of the laser radar resolution in a laboratory is achieved.

Description

Technical Field

[0001] This application relates to the field of lidar performance testing, and in particular to an apparatus and method for testing the resolution of a drilling lidar. Background Technology

[0002] In deep drilling projects, onboard lidar is a key device for accurately detecting geological structures, and its resolution directly determines the accuracy of identifying rock fractures and delineating ore body boundaries. However, the drilling environment is characterized by strong vibrations (10-200Hz), high dust levels (0.1-10mg / m³), and drastic temperature and humidity fluctuations (-20℃ to 60℃). Traditional lidar resolution detection devices are difficult to adapt to these extreme conditions, resulting in significant deviations between the detection results and actual working conditions.

[0003] Chinese patent CN114137558B discloses a scheme to improve the accuracy of lidar. It enhances range resolution and reduces circuit resource waste through multi-channel TDC circuit timing coordination control and data fusion, including timing synchronization, data acquisition and processing, and error optimization mechanisms. However, the disclosed multi-TDC frequency increase method only optimizes the internal timing of the lidar and does not consider vibration and dust interference in the drilling environment, resulting in a detection data distortion rate exceeding 30%. Chinese patent CN115704887A discloses a scheme to improve the resolution of lidar. By determining the interpolation channel to be generated, finding the associated channels and determining their weights, and then generating the interpolation channel waveform based on the waveform information and weights of the associated channels, and combining the actual data with the interpolation channel waveform to obtain point cloud data, the resolution can be improved without increasing the hardware load. However, it does not replicate the microstructure of real geological targets (such as cracks and joints) and cannot reflect the lidar's ability to detect complex rock masses.

[0004] Therefore, there is an urgent need for a detection system and device that can adapt to harsh environments, accurately simulate geological scenarios, and match the performance of high-resolution radar. Summary of the Invention

[0005] The purpose of this invention is to address the problem that existing drilling-mounted lidar detection methods cannot accurately evaluate lidar performance due to their inability to simulate real-world harsh working conditions, and to provide a device and method for detecting the resolution of drilling-mounted lidar.

[0006] The above-mentioned objective of this application is achieved through the following technical solution: The module includes a multi-dimensional physical environment simulation module, a biomimetic ground feature reconstruction module, an anti-interference high-fidelity signal acquisition module, a data processing module based on prior knowledge, and an artificial intelligence deep evaluation module. The multi-dimensional physical environment simulation module, the biomimetic ground feature reconstruction module, the anti-interference high-fidelity signal acquisition module, the data processing module based on prior knowledge, and the artificial intelligence deep evaluation module are connected in sequence. The multi-dimensional physical environment simulation module is used to reproduce the extreme environmental conditions of vibration, dust and temperature and humidity at the drilling site. The biomimetic ground feature reconstruction module is used to manufacture biomimetic targets that simulate the microstructure and optical properties of real rock strata. The anti-interference high-fidelity signal acquisition module is used to acquire the transmitted and echo signals of the lidar with high precision under extreme interference. The data processing module based on prior knowledge is used to integrate physical models and machine learning algorithms to invert resolution indicators from transmitted and echo signals; The artificial intelligence deep evaluation module is used to intelligently grade and evaluate the detection accuracy of LiDAR based on resolution indicators.

[0007] Optionally, the multidimensional physical environment simulation module includes: a dynamic vibration simulation submodule, a dynamic vibration simulation submodule, and a temperature and humidity control submodule; The dynamic vibration simulation submodule, the dynamic vibration simulation submodule, and the dynamic vibration simulation submodule are connected sequentially. The dynamic vibration simulation submodule is used to generate mechanical vibrations that simulate drilling vibrations using a piezoelectric vibrator; The dust disturbance simulation submodule is used to generate a dust environment with controllable concentration and particle size through a dust generator. The temperature and humidity control submodule is used to precisely control the temperature and humidity of the test environment.

[0008] Optionally, the dust disturbance simulation submodule adopts an intelligent control method that includes a physical model, wherein the physical model is:

[0009] in, This represents the attenuation coefficient of laser light in dust. This indicates the real-time dust concentration. This is the proportionality coefficient. The basic attenuation coefficient.

[0010] Optionally, the preparation steps of the biomimetic target are as follows: Mechanical structure and microstructure replication: High-precision machining technology is used to process the microstructure of cracks, joints and pores on the matrix, and the hardness of the matrix structure is adjusted by adding reinforcing phase; Quantitative design and implementation of optical properties: Functional coatings with specific reflectivity and roughness are prepared on structural substrates using magnetron sputtering, micro-spraying and laser micro-etching technologies; The effective reflectivity of the functional coating is given by the formula Quantitative design is carried out, in which, For effective reflectivity, The intrinsic reflectivity of the coating material, This is the attenuation coefficient related to roughness.

[0011] Optionally, the anti-interference high-fidelity signal acquisition module includes: a multi-channel TDC phase-out acquisition unit, an anti-interference support structure, and a reference optical path closed-loop and timing self-calibration unit; The multi-channel TDC phase misalignment acquisition unit uses a multi-channel TDC chip; The anti-interference support structure provides a mechanical and thermal platform for the multi-channel TDC phase-shifting acquisition unit; The anti-interference support structure uses a silicon carbide composite material to manufacture the system bracket, and topology optimization is performed through finite element analysis; an elastic damping pad is set between the bracket and the main shell to absorb vibration energy and compensate for deformation caused by thermal expansion and contraction during temperature cycling; The reference optical path closed loop and timing self-calibration unit is used for timing calibration.

[0012] Optionally, the multi-channel TDC phase-shifting acquisition unit improves the equivalent time resolution through parallel phase-shifting acquisition technology. The principle behind this improvement is quantified by the following formula:

[0013] in, This represents the system's equivalent time resolution. This represents the resolution of a single TDC channel, and N represents the number of parallel channels.

[0014] Optionally, the data processing module based on prior knowledge processes the data according to the following process: Forward physical model calculation and theoretical prediction: The built-in optical transmission model is invoked, and environmental parameters are combined to calculate the theoretical echo signal strength, signal-to-noise ratio threshold, and theoretical resolution limit; Parallel processing and preliminary feature extraction of point cloud data: FPGA is used for real-time parallel processing of point cloud data, and preliminary quantization of angular resolution and distance resolution is performed, specifically including: The number of point clouds within a 1°×1° field of view is counted using the DBSCAN clustering algorithm to quantify the angular resolution of the radar. Calculate the standard deviation of the distance difference between adjacent points. ,by Range resolution of value-quantized radar; Resolution metrics include angular resolution and distance resolution; Based on convolutional neural networks, the resolution metrics are validated and confidence is output. Specifically, this includes calling a pre-trained convolutional neural network to perform deep analysis on point cloud data, removing invalid data, and outputting high-confidence resolution metrics.

[0015] Optionally, the artificial intelligence deep evaluation module includes: a multi-dimensional feature fusion and resolution intelligent grading unit and a time-series-based performance degradation modeling and lifetime prediction unit; The multi-dimensional feature fusion and resolution intelligent classification unit is used to intelligently classify radar resolution performance based on the XGBoost classifier model and generate the correlation mapping curve between resolution and environmental parameters. The time-series-based performance degradation modeling and lifetime prediction unit is used to learn the resolution performance time series using an LSTM network, predict the remaining time for resolution degradation to reach a critical value, and trigger an early warning.

[0016] A method for detecting the resolution of a drill-mounted lidar, the method comprising: Recreate the extreme environmental conditions at the drilling site; A biomimetic target that simulates real rock strata; High-precision acquisition of lidar signals under extreme interference; The resolution index is retrieved by processing the acquired signals based on prior knowledge. Based on convolutional neural networks, resolution indicators are analyzed to obtain high-confidence resolution indicators, enabling intelligent hierarchical evaluation of the detection accuracy of lidar.

[0017] The beneficial effects of the technical solution provided in this application are: This invention achieves closed-loop detection by constructing five major modules: a multi-dimensional physical environment simulation module reproduces vibration, dust, and temperature and humidity interference; a biomimetic ground feature reconstruction module creates targets with realistic rock strata microstructure and optical properties; an anti-interference high-fidelity signal acquisition module achieves high-precision time measurement; a data processing module based on prior knowledge integrates physical models and AI algorithms to invert resolution; and an artificial intelligence deep evaluation module grades radar performance and predicts its lifespan. This invention effectively solves the problems of traditional devices being unable to adapt to extreme drilling environments and having high distortion rates in detection results. It achieves accurate, reliable, and intelligent detection of lidar resolution in the laboratory, possessing significant engineering application value. Attached Figure Description

[0018] The present application will be further described below with reference to the accompanying drawings and embodiments. In the accompanying drawings: Figure 1 This is a schematic diagram of the five-dimensional closed-loop working principle in the embodiments of this application; Figure 2 This is a schematic diagram of the five-dimensional closed-loop working principle in the embodiments of this application; Figure 3 This is a flowchart of the high-precision signal acquisition and self-calibration process in an embodiment of this application; Figure 4 This is a flowchart of the data intelligent processing and evaluation process in the embodiments of this application. Detailed Implementation

[0019] To provide a clearer understanding of the technical features, objectives, and effects of this application, the specific embodiments of this application will now be described in detail with reference to the accompanying drawings.

[0020] An embodiment of this application provides an apparatus for detecting the resolution of a drill-mounted lidar.

[0021] Please refer to Figure 1 , Figure 1 This is a five-dimensional closed-loop working principle diagram of a device for detecting the resolution of a drill-mounted lidar in an embodiment of this application, including: a multi-dimensional physical environment simulation module, a biomimetic ground feature reconstruction module, an anti-interference high-fidelity signal acquisition module, a data processing module based on prior knowledge, and an artificial intelligence deep evaluation module; The multi-dimensional physical environment simulation module, the biomimetic ground feature reconstruction module, the anti-interference high-fidelity signal acquisition module, the data processing module based on prior knowledge, and the artificial intelligence deep evaluation module are connected in sequence. The multi-dimensional physical environment simulation module is used to reproduce the extreme environmental conditions of vibration, dust and temperature and humidity at the drilling site. The biomimetic ground feature reconstruction module is used to manufacture biomimetic targets that simulate the microstructure and optical properties of real rock strata. The anti-interference high-fidelity signal acquisition module is used to acquire the transmitted and echo signals of the lidar with high precision under extreme interference. The data processing module based on prior knowledge is used to integrate physical models and machine learning algorithms to invert resolution indicators from transmitted and echo signals; The artificial intelligence deep evaluation module is used to intelligently grade and evaluate the detection accuracy of LiDAR based on resolution indicators.

[0022] In one embodiment, the five modules are connected in sequence to form a closed-loop system for detecting the resolution of the drill-mounted lidar.

[0023] As one embodiment, this patent relates to a dedicated device and method for detecting the resolution of a drill-mounted lidar, primarily serving the fields of geological exploration and engineering. This type of lidar is applied inside wells or boreholes to perform critical tasks such as borehole wall morphology scanning, geological structure identification, and underground void detection; its resolution directly determines the accuracy of geological identification.

[0024] To address the problem of high distortion rates in traditional detection methods caused by extreme environments such as strong vibrations, high dust levels, and drastic temperature and humidity fluctuations at drilling sites, this invention provides a high-precision detection solution for laboratory environments. The device replicates real-world working conditions through a multi-dimensional physical environment simulation module and utilizes a biomimetic ground feature reconstruction module to create optical targets with rock strata characteristics, constructing controllable and quantifiable testing conditions. This allows for a comprehensive performance evaluation of the resolution, reliability, and lifespan of drill-mounted lidar under simulated harsh environments.

[0025] This invention fills the gap in the field of performance verification of drilling-borne lidar under extreme working conditions, and provides a reliable means for its performance calibration, quality acceptance and condition monitoring, which has important engineering application value.

[0026] The multidimensional physical environment simulation module includes: a dynamic vibration simulation submodule, a dynamic vibration simulation submodule, and a temperature and humidity control submodule; The dynamic vibration simulation submodule, the dynamic vibration simulation submodule, and the dynamic vibration simulation submodule are connected sequentially. The dynamic vibration simulation submodule is used to generate mechanical vibrations that simulate drilling vibrations using a piezoelectric vibrator; The dust disturbance simulation submodule is used to generate a dust environment with controllable concentration and particle size through a dust generator. The temperature and humidity control submodule is used to precisely control the temperature and humidity of the test environment.

[0027] As one embodiment, the dynamic vibration simulation submodule, the dust interference simulation submodule, and the temperature and humidity control submodule work together to reproduce comprehensive environmental interference.

[0028] As one embodiment, the dynamic vibration simulation submodule integrates three sets of miniature piezoelectric vibrators (PZT-5H), arranged in an equilateral triangle on the back of the target. The frequency (10-500Hz) and amplitude (0.1-10μm) of each vibrator are precisely controlled by independent drive circuits. Its control logic is not fixed; instead, it uses an IMU sensor to collect real-time vibration feedback from the tested lidar unit and compares it with a preset drilling condition database (e.g., 300Hz high frequency for granite layers, 50Hz low frequency for shale layers), dynamically adjusting the vibration parameters to ensure the realism and representativeness of the simulated vibration.

[0029] As one embodiment, the temperature and humidity control submodule provides stable environmental boundary conditions for the entire test. Its temperature and humidity control ranges comprehensively cover drilling environments from the surface to deep underground. This module uses a combination of a PTC heater and a semiconductor cooler to achieve rapid and precise temperature control (accuracy ±0.5℃), and controls humidity through a combination of an ultrasonic humidifier and a dehumidifying membrane. This step is crucial: extreme temperature and humidity conditions directly affect the performance of the vibrator, the physical state of dust (e.g., preventing agglomeration), and the optical components of the lidar itself. A stable temperature and humidity environment is fundamental to ensuring that vibration and dust simulations function as expected, guaranteeing the repeatability and reliability of the entire environmental simulation experiment.

[0030] As one example, the multi-dimensional physical environment simulation module simulates the three-step process of "vibration-dust-temperature and humidity" and uses the dust attenuation model formula to achieve a quantitative mapping from environmental parameters to laser transmission effects, thus constructing a highly realistic, controllable and quantifiable extreme working condition test environment for subsequent lidar performance testing.

[0031] In this step, based on the original dust attenuation model, the coupling effect of vibration and temperature / humidity on laser transmission is introduced, and a comprehensive environmental interference index is proposed to quantify the coupling interference effect of multi-physical field extreme environments such as vibration, dust, temperature, and humidity on lidar signals, rather than a simple superposition:

[0032] in, The dust attenuation coefficient (calculated from the original formula); The current vibration amplitude (μm); Reference vibration amplitude (μm); The current temperature (°C); Reference temperature (°C); Current humidity (%); Reference humidity (%) The weighting coefficients for each environmental factor are determined experimentally.

[0033] It should be noted that the comprehensive environmental interference index is used to quantify the comprehensive interference intensity of multi-physics coupling on lidar signals, and is input as prior knowledge into the data processing module to correct theoretical model predictions and resolution inversion algorithms.

[0034] The dust disturbance simulation submodule adopts an intelligent control method that includes a physical model, which is:

[0035] in, This represents the attenuation coefficient of laser light in dust. This indicates the real-time dust concentration. This is the proportionality coefficient. The basic attenuation coefficient.

[0036] As one embodiment, the dust disturbance simulation submodule employs an airflow-driven dust generator with a built-in replaceable dust chamber containing particles of different rock types (0.5-10μm in diameter). A precision flow meter controls the dust concentration within the range of 0.1-10 mg / m³. Based on the real-time collected concentration value C, the attenuation coefficient μ of the current environment is dynamically calculated. This parameter is directly transmitted to the subsequent data processing module for predicting the echo signal intensity and theoretical resolution limit of the lidar in the forward physics model. Simultaneously, this submodule intelligently links with the vibration module: when the dust concentration exceeds 5 mg / m³, the system automatically reduces the vibration frequency to simulate typical scenarios such as "high dust - low vibration" karst caves, preventing excessive dust dispersion and ensuring the accuracy of environmental reproduction.

[0037] The preparation steps of the biomimetic target are as follows: Mechanical structure and microstructure replication: High-precision machining technology is used to process the microstructure of cracks, joints and pores on the matrix, and the hardness of the matrix structure is adjusted by adding reinforcing phase; Quantitative design and implementation of optical properties: Functional coatings with specific reflectivity and roughness are prepared on structural substrates using magnetron sputtering, micro-spraying and laser micro-etching technologies; The effective reflectivity of the functional coating is given by the formula Quantitative design is carried out, in which, For effective reflectivity, The intrinsic reflectivity of the coating material, This is the attenuation coefficient related to roughness.

[0038] As one example, the core function of the biomimetic feature reconstruction module is to create a biomimetic target that highly simulates the microstructure and optical properties of real rock strata, replacing the traditional simple target plate and providing a reliable target for evaluating the ability of lidar to identify geological details. Its workflow can be simplified into two core steps: The replication of mechanical structure and microstructure aims to address the issue of target "morphological similarity." First, a high-precision five-axis CNC machine tool is used to precisely machine microstructures such as microcracks, joints, and pores—characteristics unique to real rock masses—on an aluminum alloy substrate. Their size and distribution are strictly based on statistical data from a large number of rock core samples. Then, by incorporating a specific proportion of silicon carbide particles into the substrate material, its hardness is controlled to the HRC30-50 range, thereby simulating the mechanical properties from soft to hard rock. This ensures that the target possesses a mechanical response and surface geometry similar to real rock masses, preventing measurement errors introduced during laser scanning due to target deformation or feature distortion.

[0039] The quantitative design and implementation of optical properties aims to solve the problem of achieving a "realistic" target, that is, to accurately simulate the optical response of rock strata. By depositing multilayer functional coatings (such as Fe2O3, CaCO3, and C powder) on a pre-fabricated structural substrate using magnetron sputtering and micro-spraying processes, the intrinsic reflectivity of different lithologies (such as hematite, limestone, and coal seams) is simulated. Subsequently, laser micro-etching technology is used to roughen the coating surface to control its diffuse reflection characteristics.

[0040] Through the and With precise design and control, surfaces with specific optical scattering properties can be prepared proactively and predictably, rather than through blind trial and error. Ultimately, each batch of targets must be calibrated using equipment such as integrating spheres to form an optical database, ensuring that the error between its reflectivity, roughness, and other parameters and those of the actual rock core is ≤3%, thus becoming a traceable standard optical target.

[0041] Based on the original effective reflectivity model, the incident angle is introduced. and polarization angle To address the impact of [the pandemic], a generalized effective reflectance model is proposed:

[0042] Where Reff(θ,φ) is the generalized effective reflectivity (in %) at a specific incident angle θ and polarization angle φ. Intrinsic reflectance of the coating material (unit: %) δ is the attenuation coefficient related to surface roughness (dimensionless); θ is the incident angle between the laser beam and the target normal direction (unit: degrees or radians); n is the incident angle attenuation index related to the optical properties of the coating material (dimensionless); φ is the polarization angle between the laser polarization direction and the incident plane (unit: degrees or radians); δ is the polarization sensitivity coefficient of the material (dimensionless). It should be noted that this model enables the biomimetic target to simulate the scenario of radar beams scanning the surface of real rock strata at different angles, providing a highly realistic test target for the "anti-interference high-fidelity signal acquisition module". This ensures that the acquired laser echo signal contains various optical effects that would be encountered in real detection, making the subsequent resolution evaluation results more realistic and reliable.

[0043] The anti-interference high-fidelity signal acquisition module includes: a multi-channel TDC phase-out acquisition unit, an anti-interference support structure, and a reference optical path closed-loop and timing self-calibration unit; The multi-channel TDC phase misalignment acquisition unit uses a multi-channel TDC chip; As one embodiment, the multi-channel TDC phase-out acquisition unit is used to improve the equivalent time resolution by having multiple TDC chips operate in phase-out; the anti-interference support structure is used to isolate the external environment from interference to the precision acquisition circuit from a mechanical structure level; the reference optical path closed-loop and timing self-calibration unit is used to achieve timing self-calibration by establishing an internal timing reference and PID control algorithm to eliminate timing errors.

[0044] As one embodiment, multi-channel TDC phase-shifting acquisition and equivalent resolution enhancement aim to improve the system's fundamental time measurement accuracy from the hardware architecture level. The system integrates four high-precision time-to-digital converter (TDC-GPX2) chips, each with an inherent time resolution of 5ps. It is synchronously driven by a 500MHz oven-controlled crystal oscillator (OCXO) with a frequency stability ≤1ppm, but the start-up clock phase of each channel operates out of phase at preset intervals (t / n=0.5ns, n=4).

[0045] The anti-interference support structure provides a mechanical and thermal platform for the multi-channel TDC phase-shifting acquisition unit; The anti-interference support structure uses a silicon carbide composite material to manufacture the system bracket, and topology optimization is performed through finite element analysis; an elastic damping pad is set between the bracket and the main shell to absorb vibration energy and compensate for deformation caused by thermal expansion and contraction during temperature cycling; As one embodiment, the anti-interference support structure and thermo-mechanical co-design aim to isolate the precision acquisition circuit from interference from harsh external environments at the mechanical structural level. Silicon carbide composite material (CTE≤5×10⁻⁶) is used. -6 The system support was manufactured at / ℃, and topology optimization was performed using finite element analysis to ensure that its natural frequency (250-300Hz) was far from the main frequency band of drilling rig vibration (10-200Hz), thereby effectively suppressing the vibration transmission rate to ≤5%. Simultaneously, an elastic damping pad (Shore A 40) was installed between the support and the main housing to absorb vibration energy and compensate for deformation caused by thermal expansion and contraction during temperature cycling (-20℃~60℃).

[0046] As one embodiment, the anti-interference support structure provides a mechanical and thermal platform for the high-precision TDC unit in the first step, ensuring that the relative pose deviation between the TDC chip and the laser optical path is strictly controlled within 10μm, thus physically avoiding measurement reference inaccuracies caused by structural deformation or vibration.

[0047] The reference optical path closed-loop and timing self-calibration unit is used to perform timing calibration, eliminating all residual timing errors.

[0048] As one implementation, the closed-loop reference optical path and timing self-calibration aim to ultimately eliminate all residual timing errors at the system and algorithm levels, achieving picosecond-level absolute measurement accuracy. The system couples 0.1% of the optical power from the laser's transmitter through an optical fiber beam splitter, which is then input as a reference signal to a high-speed InGaAs detector (response time <1ns) at a fixed distance (1m), thus establishing an extremely stable "transmit-reference-receive" internal timing reference. Subsequently, the system compares the timestamp difference between the measured echo signal and the reference signal recorded in the TDC in real time every second.

[0049] It should be noted that the fluctuation of this time difference (typically ≤5ps) directly reflects the timing drift caused by temperature and voltage fluctuations. The system uses a PID (Proportional-Integral-Derivative) control algorithm to dynamically and inversely adjust the clock phase of the drive TXC based on this error. This forms a high-precision negative feedback closed-loop control system, the purpose of which is to continuously bring the timing error to zero, thereby ultimately and stably controlling the time measurement error at the entire system level to within 10ps, ensuring the absolute accuracy and reliability of the raw time data under any harsh operating conditions.

[0050] Based on the original equivalent time resolution formula, timing jitter caused by environmental disturbances is introduced. A system-level time error synthesis formula is proposed:

[0051] in, The total time measurement error of the system under extreme conditions (unit: ps); The inherent time resolution of a single-channel TDC chip (unit: ps); The number of TDC channels operating in parallel (dimensionless). Timing jitter caused by environmental factors such as temperature drift, mechanical vibration, and power supply noise (unit: ps); Residual calibration error (unit: ps) that cannot be completely eliminated by the timing self-calibration system.

[0052] It should be noted that this formula comprehensively quantifies the performance of the three major sub-units in the anti-interference high-fidelity signal acquisition module: This item reflects the theoretical performance improvement of the multi-channel TDC phase-shifting acquisition unit; This item demonstrates the effect of the anti-interference support structure in suppressing environmental interference; This item reflects the calibration capability of the reference optical path closed loop and timing self-calibration unit.

[0053] It should be noted that this model is a mathematical guarantee ensuring the accuracy of the original data. It proves and ensures the high-precision timestamp data output by the system. The extremely low level of reliability provides a solid data foundation for accurate resolution inversion by data processing modules based on prior knowledge. Without this accuracy guarantee, all subsequent algorithmic processing would be meaningless.

[0054] The multi-channel TDC phase-shifting acquisition unit improves the equivalent time resolution through parallel phase-shifting acquisition technology. The principle behind this improvement is quantified by the following formula:

[0055] in, This represents the system's equivalent time resolution. This represents the resolution of a single TDC channel, and N represents the number of parallel channels.

[0056] As one example, this parallel phase-shifting acquisition technology calculates the signal-to-noise ratio (SNR) improvement and equivalent resolution improvement brought about by multi-channel parallel acquisition, increasing the system's basic time measurement capability from 5ps for a single channel to the equivalent 0.5ns (500ps) level, laying a solid hardware foundation for subsequent high-precision measurements.

[0057] The data processing module based on prior knowledge processes the data according to the following process: Forward physical model calculation and theoretical prediction: The built-in optical transmission model is invoked, and environmental parameters are combined to calculate the theoretical echo signal strength, signal-to-noise ratio threshold, and theoretical resolution limit; As one embodiment, the forward physics model calculation and theoretical prediction system receives sensor data from the environmental simulation module in real time, including temperature T, pressure P, and dust concentration C. It then calls upon a built-in optical transmission model for calculation, which includes both TOF (Time-of-Flight) and FMCW (Frequency Modulated Continuous Wave) modes. For the TOF model, the system calculates the attenuation of the laser in the dusty environment based on Lambert's law, and its attenuation coefficient... From the formula Confirmed, among which This is the proportionality coefficient, with a value of 0.05. The base attenuation coefficient is set to 0.05m. -1 , This represents the real-time dust concentration. The model outputs the theoretical echo signal strength and signal-to-noise ratio threshold under the current environment. For the FMCW model, the system predicts the theoretical resolution limit by mapping the beat frequency signal frequency to distance and compensating for frequency shifts caused by vibration.

[0058] Parallel processing and preliminary feature extraction of point cloud data: FPGA is used for real-time parallel processing of point cloud data, and preliminary quantization of angular resolution and distance resolution is performed, specifically including: The number of point clouds within a 1°×1° field of view is counted using the DBSCAN clustering algorithm to quantify the angular resolution of the radar. Calculate the standard deviation of the distance difference between adjacent points. ,by Range resolution of value-quantized radar; Resolution metrics include angular resolution and distance resolution; Based on convolutional neural networks, the resolution metrics are validated and confidence is output. Specifically, this includes calling a pre-trained convolutional neural network to perform deep analysis on point cloud data, removing invalid data, and outputting high-confidence resolution metrics.

[0059] As one implementation, intelligent verification and confidence output based on convolutional neural networks (CNNs) utilize a pre-trained lightweight CNN to perform deep analysis on point cloud data. This network, trained on over 100,000 sets of drilling scenario data, analyzes the edge gradient features and noise distribution characteristics of the point cloud data. The network automatically identifies and removes invalid data with a signal-to-noise ratio below 10dB or edge blurring exceeding 2 pixels. For valid data, the network outputs a confidence score, which must be no less than 95%. This step utilizes artificial intelligence technology to verify and correct the prediction results of the physical model, ultimately outputting a high-confidence resolution index that is both theoretically predicted and data-verified.

[0060] Based on this, the output of the first step "forward physics model calculation and theoretical prediction" in this module is deeply integrated. This is used as a constraint to guide the learning process of the neural network, thus proposing a physically guided neural network loss function for training CNNs for resolution verification:

[0061] in, The total loss function used to train the neural network; This is a traditional loss function (such as mean squared error) based on predicted values ​​and labeled data. These are the distance resolution predictions and angular resolution predictions output by the neural network. This is the theoretical resolution limit calculated based on the "forward physics model"; , which is a dimensionless hyperparameter used to balance data-driven loss and physical constraint loss.

[0062] It should be noted that the CNN trained using this loss function outputs resolution metrics with extremely high reliability and physical consistency. These high-quality metrics serve as a direct and reliable input source for the "Artificial Intelligence Deep Evaluation Module" in its intelligent grading and lifetime prediction, ensuring the authority of the final evaluation results.

[0063] As one embodiment, the data processing module based on prior knowledge integrates physical models and machine learning algorithms to accurately retrieve the resolution parameters of the lidar from raw data affected by environmental interference. The module achieves intelligent and highly reliable extraction of resolution parameters by establishing theoretical benchmarks and cross-validating them with measured data.

[0064] The artificial intelligence deep evaluation module includes: a multi-dimensional feature fusion and resolution intelligent grading unit and a time series-based performance degradation modeling and lifetime prediction unit; The multi-dimensional feature fusion and resolution intelligent classification unit is used to intelligently classify radar resolution performance based on the XGBoost classifier model and generate the correlation mapping curve between resolution and environmental parameters. The time-series-based performance degradation modeling and lifetime prediction unit is used to learn the resolution performance time series using an LSTM network, predict the remaining time for resolution degradation to reach a critical value, and trigger an early warning.

[0065] As one embodiment, the system features multi-dimensional feature fusion and intelligent resolution grading. It receives high-confidence indicators and environmental parameters from the data processing module in real time, constructing a 128-dimensional feature vector that includes signal features, point cloud features, and environmental features. This feature vector is input into a classifier model trained using the XGBoost algorithm. The classifier has been pre-trained using over 100,000 sets of measured data covering 20 mainstream drilling-mounted lidars, enabling it to classify the radar's resolution performance into five distinct levels: Level 1 (≤3 cm), Level 2 (3-5 cm), Level 3 (5-8 cm), Level 4 (8-10 cm), and Level 5 (>10 cm).

[0066] It's important to note that this step not only outputs grading results, but more importantly, generates a correlation mapping curve between resolution and environmental parameters, quantitatively revealing the specific pattern, such as the 0.5-1 cm decrease in resolution caused by a 1 mg / m³ increase in dust concentration. Based on time-series performance degradation modeling and lifespan prediction, the system utilizes a Long Short-Term Memory (LSTM) network to perform deep learning on the continuously monitored resolution performance time series. This network learns the decay pattern and trend of resolution with increasing operating time and extrapolates predictions based on the average lifespan of the drilling equipment (typically 5000 hours). Its output is the remaining time for radar resolution to decay to a user-defined critical value (e.g., 10 cm), with a prediction error of no more than 10%. When the system predicts that the remaining lifespan is below the 500-hour threshold, it automatically triggers an early warning signal and generates concrete maintenance recommendations. This achieves a leap from condition monitoring to predictive maintenance, providing users with forward-looking decision support for equipment health management.

[0067] Based on LSTM lifetime prediction, a formula for quantifying resolution degradation rate is proposed for real-time monitoring of performance degradation trends:

[0068] in, This represents the average absolute performance degradation rate up to time t (unit: e.g., cm / hour). Its physical meaning is the average amount of degradation in resolution metrics per unit time. For timestamp The detected lidar resolution value (unit: e.g., cm); Cumulative running time (unit: e.g., hour); Let be the instantaneous rate of change of the resolution at time τ (i.e., the derivative).

[0069] It should be noted that the calculated recession rate As a keen indicator of health status, it can trigger earlier and more accurate warnings. It complements and enhances the functionality of the LSTM lifetime prediction model, together forming a complete technical loop from "condition monitoring" to "predictive maintenance," providing forward-looking and refined decision support for user equipment health management, and reflecting the ultimate sublimation of the entire device's value.

[0070] This invention provides a system and device for detecting the resolution of drill-mounted lidar, particularly suitable for harsh working conditions such as deep geological exploration and tunnel construction. The system employs a five-dimensional closed-loop detection system encompassing "environment-target-signal-algorithm-evaluation," and introduces a series of original mathematical models (including a comprehensive environmental interference index, generalized effective reflectivity, system-level time error synthesis, physical information neural network loss function, and performance degradation rate quantification formula). This enables end-to-end quantitative design and precise control for extreme working conditions, anisotropic targets, high-fidelity signals, intelligent inversion, and lifetime prediction. This invention not only reproduces vibration, dust, and temperature / humidity interference but also quantifies their coupling effects; it not only fabricates a biomimetic target but also accurately simulates its optical response under different detection geometries; it not only achieves high-precision time measurement but also theoretically ensures and optimizes the system's ultimate accuracy; it not only integrates physics and AI algorithms but also ensures the physical reliability of output indicators through a physical constraint loss function; and it not only performs hierarchical evaluation but also achieves refined lifetime management based on instantaneous degradation rate. In summary, through dual innovation in theory and technology, this invention effectively solves the problems of traditional devices being unable to adapt to extreme drilling environments and having high distortion rates in detection results. It achieves accurate, reliable, traceable, and intelligent detection and forward-looking health management of drilling-borne lidar resolution in the laboratory, and has significant engineering application value.

[0071] A method for detecting the resolution of a drilling-mounted lidar, implemented based on an apparatus for detecting the resolution of a drilling-mounted lidar as described in any one of claims 1-8, characterized in that the method comprises: Recreate the extreme environmental conditions at the drilling site; A biomimetic target that simulates real rock strata; High-precision acquisition of lidar signals under extreme interference; The resolution index is retrieved by processing the acquired signals based on prior knowledge. Based on convolutional neural networks, resolution indicators are analyzed to obtain high-confidence resolution indicators, enabling intelligent hierarchical evaluation of the detection accuracy of lidar.

[0072] This application provides an embodiment as follows: the data processing module adopts a processing method that integrates physical models and machine learning algorithms, specifically including: firstly, providing a theoretical prediction benchmark through a forward physical model; then, using artificial intelligence algorithms to intelligently verify and correct the measured data; and finally outputting a high-confidence resolution index with dual guarantees. This application provides an embodiment as follows: Figure 1This is a five-dimensional closed-loop working principle diagram of the system of this invention. It shows the closed-loop system of this invention, which consists of five core modules. The main control system drives the environment simulation module to generate test conditions; after the biomimetic target is irradiated by laser, the signal acquisition module acquires data; the data processing module analyzes the data; and the intelligent evaluation module generates results and provides feedback, forming a closed loop. Figure 2 This is a flowchart illustrating the preparation and calibration of the optical properties of biomimetic targets. It demonstrates the quantitative preparation process of the target coating. Formulas guide coating design and processing, and calibration ensures that the error between its optical properties and those of real rock cores is ≤3%. Figure 3 This is a flowchart of high-precision signal acquisition and self-calibration. It demonstrates how an internal benchmark is established through a reference optical path, and how the echo time is calibrated in a closed loop using a PID algorithm to obtain high-precision measurement results even in harsh environments. Figure 4 This is a flowchart of data intelligent processing and evaluation. It illustrates the progressive process of data processing. The physical model provides the theoretical benchmark, FPGA and CNN work together to process and validate the data, and finally, the AI ​​model is used for performance grading and lifetime prediction.

[0073] The above are merely exemplary embodiments of this disclosure and should not be construed as limiting the scope of this disclosure. Any equivalent changes and modifications made in accordance with the teachings of this disclosure shall still fall within the scope of this disclosure.

[0074] This application is intended to cover any variations, uses, or adaptations of this disclosure that follow the general principles of this disclosure and include common knowledge or customary techniques in the art not described in this disclosure. The specification and embodiments are to be considered exemplary only, and the scope and spirit of this disclosure are defined by the claims.

Claims

1. A device for detecting the resolution of a drilling-mounted lidar, characterized in that, The device includes: a multi-dimensional physical environment simulation module, a biomimetic ground feature reconstruction module, an anti-interference high-fidelity signal acquisition module, a data processing module based on prior knowledge, and an artificial intelligence deep evaluation module. The multi-dimensional physical environment simulation module, the biomimetic ground feature reconstruction module, the anti-interference high-fidelity signal acquisition module, the data processing module based on prior knowledge, and the artificial intelligence deep evaluation module are connected in sequence. The multi-dimensional physical environment simulation module is used to reproduce the extreme environmental conditions of vibration, dust and temperature and humidity at the drilling site. The biomimetic ground feature reconstruction module is used to manufacture biomimetic targets that simulate the microstructure and optical properties of real rock strata. The anti-interference high-fidelity signal acquisition module is used to acquire the transmitted and echo signals of the lidar with high precision under extreme interference. The data processing module based on prior knowledge is used to integrate physical models and machine learning algorithms to retrieve resolution metrics from transmitted and echo signals. The artificial intelligence deep evaluation module is used to intelligently grade and evaluate the detection accuracy of lidar based on resolution indicators.

2. The device for detecting the resolution of a drilling-mounted lidar as described in claim 1, characterized in that, The multidimensional physical environment simulation module includes: a dynamic vibration simulation submodule, a dynamic vibration simulation submodule, and a temperature and humidity control submodule; The dynamic vibration simulation submodule, the dynamic vibration simulation submodule, and the dynamic vibration simulation submodule are connected sequentially. The dynamic vibration simulation submodule is used to generate mechanical vibrations that simulate drilling vibrations using a piezoelectric vibrator; The dust disturbance simulation submodule is used to generate a dust environment with controllable concentration and particle size through a dust generator. The temperature and humidity control submodule is used to precisely control the temperature and humidity of the test environment.

3. The device for detecting the resolution of a drilling-mounted lidar as described in claim 2, characterized in that, The dust disturbance simulation submodule adopts an intelligent control method that includes a physical model, which is: in, This represents the attenuation coefficient of laser light in dust. This indicates the real-time dust concentration. This is the proportionality coefficient. The basic attenuation coefficient.

4. The device for detecting the resolution of a drilling-mounted lidar as described in claim 1, characterized in that, The preparation steps of the biomimetic target are as follows: Mechanical structure and microstructure replication: High-precision machining technology is used to process the microstructure of cracks, joints and pores on the matrix, and the hardness of the matrix structure is adjusted by adding reinforcing phase; Quantitative design and implementation of optical properties: Functional coatings with specific reflectivity and roughness are prepared on structural substrates using magnetron sputtering, micro-spraying and laser micro-etching technologies; The effective reflectivity of the functional coating is given by the formula Quantitative design is carried out, in which, For effective reflectivity, The intrinsic reflectivity of the coating material, This is the attenuation coefficient related to roughness.

5. The device for detecting the resolution of a drilling-mounted lidar as described in claim 1, characterized in that, The anti-interference high-fidelity signal acquisition module includes: a multi-channel TDC phase-out acquisition unit, an anti-interference support structure, and a reference optical path closed-loop and timing self-calibration unit; The multi-channel TDC phase misalignment acquisition unit uses a multi-channel TDC chip; The anti-interference support structure provides a mechanical and thermal platform for the multi-channel TDC phase-shifting acquisition unit; The anti-interference support structure uses a silicon carbide composite material to manufacture the system bracket, and topology optimization is performed through finite element analysis; an elastic damping pad is set between the bracket and the main shell to absorb vibration energy and compensate for deformation caused by thermal expansion and contraction during temperature cycling; The reference optical path closed loop and timing self-calibration unit is used for timing calibration.

6. The device for detecting the resolution of a drilling-mounted lidar as described in claim 5, characterized in that, The multi-channel TDC phase-shifting acquisition unit improves the equivalent time resolution through parallel phase-shifting acquisition technology. The principle behind this improvement is quantified by the following formula: in, This represents the system's equivalent time resolution. This represents the resolution of a single TDC channel, and N represents the number of parallel channels.

7. The device for detecting the resolution of a drilling-mounted lidar as described in claim 1, characterized in that, The data processing module based on prior knowledge processes the data according to the following process: Forward physical model calculation and theoretical prediction: The built-in optical transmission model is invoked, and environmental parameters are combined to calculate the theoretical echo signal strength, signal-to-noise ratio threshold, and theoretical resolution limit; Parallel processing and preliminary feature extraction of point cloud data: FPGA is used for real-time parallel processing of point cloud data, and preliminary quantization of angular resolution and distance resolution is performed, specifically including: The angular resolution of the radar is quantified by counting the number of point clouds within a 1°×1° field of view using the DBSCAN clustering algorithm. Calculate the standard deviation of the distance difference between adjacent points. ,by Range resolution of value-quantized radar; Resolution metrics include angular resolution and distance resolution; Based on convolutional neural networks, the resolution metrics are validated and confidence is output. Specifically, this includes calling a pre-trained convolutional neural network to perform deep analysis on point cloud data, removing invalid data, and outputting high-confidence resolution metrics.

8. The device for detecting the resolution of a drilling-mounted lidar as described in claim 1, characterized in that, The artificial intelligence deep evaluation module includes: a multi-dimensional feature fusion and resolution intelligent grading unit and a time series-based performance degradation modeling and lifetime prediction unit; The multi-dimensional feature fusion and resolution intelligent classification unit is used to intelligently classify radar resolution performance based on the XGBoost classifier model and generate the correlation mapping curve between resolution and environmental parameters. The time-series-based performance degradation modeling and lifetime prediction unit is used to learn the resolution performance time series using an LSTM network, predict the remaining time for resolution degradation to reach a critical value, and trigger an early warning.

9. A method for detecting the resolution of a drilling-mounted lidar, implemented based on an apparatus for detecting the resolution of a drilling-mounted lidar as described in any one of claims 1-8, characterized in that, The method includes: Recreate the extreme environmental conditions at the drilling site; A biomimetic target that simulates real rock strata; High-precision acquisition of lidar signals under extreme interference; The resolution index is retrieved by processing the acquired signals based on prior knowledge. Based on convolutional neural networks, resolution indicators are analyzed to obtain high-confidence resolution indicators, enabling intelligent hierarchical evaluation of the detection accuracy of lidar.

Images