A rock-breaking evaluation device and optimization method for carbon dioxide jet combined with particle impact
By designing a rock-breaking evaluation device combining carbon dioxide jet and particle impact with machine learning optimization methods, the problem of low traditional rock-breaking efficiency in deep hard rock formations was solved. This enabled the systematic quantification and optimization of rock-breaking effects, improving drilling efficiency and reducing costs.
Patent Information
- Application Number
- CN202611054230.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-07-15
- Publication Date
- 2026-11-17
AI Technical Summary
Existing technologies cannot systematically quantify the rock-breaking effect of carbon dioxide-carrying particle jets, have failed to establish a predictive model for the rock-breaking effect of mixed particle systems, and traditional mechanical rock-breaking technologies are inefficient and costly in deep hard rock formations.
Design a rock-breaking evaluation device combining carbon dioxide jet and particle impact, including a gas pressurization system, a gas-solid mixing system, an impact testing system, a material recovery and screening system, and a rock crushing quantitative analysis system. Combined with a machine learning parameter optimization module, it can reveal the influence of multiple parameters and the optimal combination of process parameters.
This study achieved systematic quantification and optimization of the rock-breaking efficiency of carbon dioxide jets, improved drilling efficiency in deep hard rock formations, provided guidance for engineering applications, reduced costs, and improved experimental accuracy and efficiency.
Smart Images

Figure CN123116748A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of high-speed particle impact rock breaking technology, specifically relating to a rock breaking evaluation device and optimization method for carbon dioxide jet combined with particle impact. Background Technology
[0002] In the early stages of oil and natural gas extraction, the exploration scope was limited to shallow rock formations due to the limitations of deep, hard rock extraction technology. With rapid industrial development, shallow surface oil and gas resources can no longer meet the ever-increasing energy demand, making the development of deep formation oil and gas resources a core strategic direction in global oil and gas exploration. As extraction depth increases, the rock hardness of the formation increases exponentially, and traditional mechanical rock breaking technologies face insurmountable technical bottlenecks such as severe drill bit wear, a sharp drop in drilling efficiency, and soaring construction costs. [1] Against this backdrop, particle impact-assisted rock breaking, as a non-contact, high-efficiency rock breaking technology, has emerged, which achieves rock crushing by using high-pressure fluid to carry abrasive particles to impact rocks at high speed.
[0003] The rock-breaking effect of particle impact is affected by a variety of factors, including abrasive particle size, abrasive type, jet velocity, and concentration. [2] The diameter and type of the mixed abrasive are particularly critical: large-diameter particles provide impact kinetic energy, while small-diameter particles enhance grinding, but the ratio between the abrasive particles must be balanced. Among them, high-speed carbon dioxide particle jets have high rock-breaking efficiency, lower energy and water consumption, and can also achieve carbon sequestration and improve oil and gas recovery by injecting and storing carbon dioxide, thus having both environmental and economic benefits.
[0004] Several scholars have conducted targeted studies on the influence of different factors on the rock-breaking effect of particle impact. (Fang Tiancheng, Ren Fushen) [3] The particle jet coupled resonant rock-breaking device and method proposed by Zhou Zongqing et al. (CN202311458132.4) can change the particle ratio, but all impact particles come from the same particle storage tank, without differentiation in particle size and material. It can only adjust the particle concentration within the same particle size range, and cannot achieve independent feeding and mixed gradation tests of two or more different particle sizes. Furthermore, it has not established a quantitative evaluation method for the rock-breaking effect of mixed particle size systems. [4] A numerical simulation framework for particle impact rock breaking was established (CN202211548562), but the simulation medium was conventional drilling fluid, and it did not involve particle jet systems under carbon dioxide-carrying conditions. Furthermore, it used single-size particles as the simulation object and did not establish a predictive model for the rock breaking effect of mixed particle systems. (Li Botao, Zhang Hongtu) [5]A rock-breaking device and method combining liquid nitrogen jet and particle impact was proposed (CN202411833515), but it did not involve carbon dioxide-carried particle jets. In summary, previous studies have made progress in multi-parameter rock-breaking testing, numerical simulation optimization, and combined liquid nitrogen-particle rock breaking, but there are still shortcomings such as the inability to systematically quantify the rock-breaking effect under mixed particle conditions, the lack of an engineering-applicable prediction model, and insufficient consideration of carbon dioxide-driven particle impact.
[0005] To address the challenge of systematically quantifying rock-breaking efficiency, a rock-breaking evaluation device and optimization method combining carbon dioxide jet and particle impact were designed and developed. This device not only systematically reveals the influence of various parameters on rock-breaking efficiency but also integrates with LightGBM machine learning technology. [6] Deep integration enables the rapid provision of optimal process parameter combinations for engineering sites.
[0006] This study has significant theoretical value and engineering guidance for promoting the engineering application of carbon dioxide particle jet rock breaking technology, improving drilling efficiency in deep hard rock formations, and realizing the green and efficient development of oil and gas resources.
[0007] [1] He Manchao, et al. Advances in rock mechanics for deep mining [J]. Journal of Coal Science and Technology, 2024, 49(1):1-23.
[0008] [2] Liu Yong, Chen Changjiang, Wei Jianping, et al. Comparative analysis of rock-breaking pressure of abrasive water jet and abrasive gas jet [J]. Journal of Coal Science and Technology, 2018, 43(9):2510-2517.
[0009] [3] Northeast Petroleum University. Particle jet coupled resonant rock breaking device and rock breaking method: CN202311458132.4[P]. 2024-02-02.
[0010] [4] Shandong University. A simulation and optimization method and system for particle impact rock breaking technology: CN202211548562.0[P]. 2023-06-23.
[0011] [5] Henan Polytechnic University. A rock-breaking device and method of liquid nitrogen jet combined with particle impact: CN202411833515.X[P]. 2025-10-31.
[0012] [6] Wang Ying, Wang Enyuan, et al. A novel metaheuristic LightGBM model for predicting rockburst intensity [J]. Shock and Vibration, 2024. Summary of the Invention
[0013] To achieve the above objectives, the technical solution of the present invention is a rock-breaking evaluation device for carbon dioxide jet combined with particle impact, which includes five major hardware systems: a gas pressurization system, a gas-solid mixing system, an impact test system, a material recovery and screening system, and a rock crushing quantitative analysis system, and also includes a machine learning parameter optimization module;
[0014] The gas pressurization system includes a gas storage tank 1, a gas pump 3, a pipeline heat exchanger 2, a temperature control unit 21, and a monitoring instrument 22. The outlet of the gas storage tank 1 is connected to the inlet of the gas pump 3, and the inlet of the gas pump 3 and the outlet of the gas storage tank 1 are connected through the pipeline heat exchanger 2. The gas pump 3 is connected to the gas-solid mixing system via a pressurization gas pipe 4. The temperature control unit 21 and the monitoring instrument 22 are electrically connected to the pipeline heat exchanger 2 and provide feedback control of its heat exchange power. By adjusting the temperature and pressure of CO2 in the pipeline, phase conversion is achieved, which is used to realize the continuous and controllable conversion of carbon dioxide liquid phase to gas phase, and to precisely control the pressure and flow rate of the output gas.
[0015] The gas-solid mixing system includes a first abrasive storage tank 81, a second abrasive storage tank 82, a screw metering feeder 10, and a gas-solid mixing device 9. The outlets of the first abrasive storage tank 81 and the second abrasive storage tank 82 are connected in parallel and then connected to the inlet of the screw metering feeder 10. The outlet of the screw metering feeder 10 is connected to the abrasive storage chamber 91 of the gas-solid mixing device 9. The gas inlet of the gas-solid mixing device 9 is connected to the outlet of the air pump 3. The mixed fluid outlet of the gas-solid mixing device 9 is connected to the impact testing system. The two abrasive storage tanks are arranged in parallel to adjust the abrasive type, particle size, concentration, and gas-solid mass ratio to output a uniform and stable gas-solid two-phase fluid.
[0016] The impact testing system includes a main body 12, a reverse high-speed jet nozzle 13, an autoclave 14, a confining pressure device 15, a high-speed camera 16, and a backflow pipe 17. The inlet end of the main body 12 is connected to the outlet of the injection pipe 11. The mixed fluid outlet of the gas-solid mixing device 9 is connected to the inlet of the main body 12 via the injection pipe. The nozzle 123 of the main body 12 extends into the autoclave 14 and is positioned towards the rock specimen. The reverse high-speed jet nozzle 13 is located on the side wall of the detachable nozzle device 121 of the main body 12, and its jetting direction is towards the inlet of the backflow pipe 17. The confining pressure device 15 is arranged around the outside of the autoclave 14. The high-speed camera 16 is located outside the transparent observation window 141 of the autoclave 14. The outlet of the backflow pipe 17 is connected to the material recovery and screening system. This system is used to simulate the rock impact process under different depth stress environments and to collect impact parameters and images of the rock damage process.
[0017] The material recovery and screening system includes a gas buffer device 18, an ultrasonic vibrating screen 19, and a recovery tank 20. The outlet of the backflow pipe 17 is connected to the inlet of the gas buffer device 18, the outlet of the gas buffer device 18 is connected to the inlet of the ultrasonic vibrating screen 19, and the abrasive outlet of the ultrasonic vibrating screen 19 is connected to the recovery tank 20. This system is used to separate rock chips from abrasives and realize the recycling of abrasives.
[0018] The rock breaking quantitative analysis system includes the high-speed camera 16, a three-dimensional image processing module and a rock cuttings weighing module, which are used to process the collected images and rock cuttings data, and to quantitatively calculate the core indicators of rock breaking effect.
[0019] The machine learning parameter optimization module is used to construct a LightGBM prediction model based on quantified data, and output the optimal rock-breaking parameter combination by combining a genetic algorithm or a particle swarm optimization algorithm. Based on this, it controls or prompts the operator to adjust the set values of the gas pressurization system and the gas-solid mixing system for verification experiments. The machine learning parameter optimization module is also used to feed the verification experiment data back to the training set for incremental learning. It is used to guide the parameter reset of the gas pressurization system and / or the gas-solid mixing system. This module is a hardware and software system integrated in the computer, and its function depends on the test data collected by the aforementioned hardware system to be realized.
[0020] Furthermore, the gas pressurization system also includes a temperature sensor, a pressure gauge 5, and a flow rate gauge 6. The temperature sensor is installed on the pressurized gas pipe 4 at the outlet of the pipeline heat exchanger 2. The pressure gauge 5 and the flow rate gauge 6 are sequentially installed on the pressurized gas pipe 4 at a position after the pipeline heat exchanger 2. The temperature regulation unit 21 adjusts the output pressure of the gas pump 3 and the heat exchange power of the pipeline heat exchanger 2 according to the gas pressure and temperature feedback signals to achieve stable phase conversion of carbon dioxide.
[0021] Furthermore, replaceable screens are respectively provided at the outlets of the first abrasive storage tank 81 and the second abrasive storage tank 82 to control the abrasive particle size range entering the screw metering feeder 10; the outlet of the screw metering feeder 10 is connected to the inlet of the abrasive storage chamber 91 of the gas-solid mixing device 9 to precisely adjust the abrasive feeding rate; the gas-solid mixing device 9 is provided with an abrasive storage chamber 91 and an emission pipe 92, the outlet of the abrasive storage chamber 91 is connected to the inlet of the emission pipe 92, and the air inlet of the emission pipe 92 is connected to the outlet of the pipeline heat exchanger 2 to fully mix the abrasive with pressurized carbon dioxide.
[0022] Furthermore, the main body 12 of the device includes a detachable nozzle device 121, a detachable nut 122 and a replaceable nozzle 123. The replaceable nozzle 123 is fixedly installed at the outlet end of the main body 12 of the device through the detachable nut 122. The detachable nozzle device 121 integrates an integrated air pressure and flow rate gauge 124 for real-time monitoring of the impact air pressure and flow rate at the nozzle inlet.
[0023] Furthermore, the reverse high-speed jet nozzle 13 is fixedly installed on the side wall of the detachable nozzle device 121, its inlet is connected to an external compressed air source, and its outlet faces the inlet direction of the reverse flow pipe 17, which is used to create negative pressure in the impact area and drive rock chips and residual abrasive into the reverse flow pipe 17.
[0024] Furthermore, the inlet of the gas buffer device 18 is connected to the outlet of the reflux pipe 17 to reduce the flow rate of the mixed fluid; the outlet of the gas buffer device 18 is connected to the feed inlet of the ultrasonic vibrating screen 19, and the undersize outlet of the ultrasonic vibrating screen 19 is connected to the inlet of the recovery tank 20 to separate rock chips and abrasives and collect the screened abrasives.
[0025] Furthermore, the rock fragmentation quantitative analysis system performs three-dimensional reconstruction of the images acquired by the high-speed camera 16, and combines rock fragment sieving analysis to quantitatively calculate the core indicators of rock impact damage depth and fragmentation volume.
[0026] A method for evaluating and optimizing the rock-breaking parameters of carbon dioxide jet combined with particle impact includes the following steps:
[0027] S1 Test preparation: Fix the rock specimen in the autoclave 14 and apply the preset confining pressure through the confining pressure device 15. Load the specified abrasive into the abrasive storage tanks 81 and 82 and install the matching replaceable nozzle 123.
[0028] S2 Phase Adjustment: Turn on the air pump 3, and adjust the gas pressure and temperature to the preset value through the phase conversion controller to achieve stable output of carbon dioxide in the preset phase.
[0029] S3 Impact Test: The abrasive is mixed with pressurized carbon dioxide to form a gas-solid two-phase flow, which is then ejected through replaceable nozzle 123 to impact a rock specimen. Impact parameters and rock damage images are collected in real time.
[0030] S4 Material Recovery and Data Quantification: Separate rock cuttings and abrasives, calculate the quantitative index of rock breaking effect through the rock crushing quantitative analysis system, and form a standardized test dataset;
[0031] S5 Parameter Optimization: With the dual optimization objectives of maximizing rock damage depth and fracture volume, the LightGBM prediction model is trained based on the experimental dataset. The Pareto optimal combination of rock breaking parameters is output by combining the genetic algorithm or particle swarm optimization algorithm, and the validation experimental data is fed back to the model for incremental learning.
[0032] The intelligent optimization in step S5 specifically involves: taking the maximization of impact fracture volume and impact depth as the optimization objective, setting the feasible range of each test parameter, using a genetic algorithm or particle swarm optimization algorithm for global optimization, and outputting the Pareto optimal parameter combination; it also includes a model update step: conducting verification experiments based on the optimal parameter combination, adding the verification experiment data to the training set, incrementally training the LightGBM model, and continuously improving the model's prediction accuracy under target lithology and working conditions.
[0033] Furthermore, step S5 also includes a data preprocessing step: using the 3σ criterion to identify and remove outliers, replacing outliers with the mean of parallel test data under the same working conditions, and using the Min-Max standardization method to map the data to the [0,1] interval.
[0034] Further, the model training in step S5 specifically involves: dividing the preprocessed dataset into training, validation, and test sets according to a certain ratio; building a LightGBM multi-output prediction model based on the training set using a chain regression framework; modeling the correlation between rock damage depth and fracture volume using regression chains; simultaneously predicting the two indicators; optimizing the model hyperparameters using a grid search method or Bayesian optimization algorithm; minimizing the validation set error and maximizing the coefficient of determination R² as the optimization objective; and verifying the model's prediction accuracy using the test set.
[0035] Compared with existing technologies, the impact device and machine learning parameter optimization system of this invention work together to form an integrated mode of "gas pressurization-gas-solid mixing-impact test-recovery screening-damage quantification-intelligent optimization". This effectively combines the controllable conversion of carbon dioxide multiphase state, independent adjustment of multi-dimensional parameters and intelligent optimization of rock breaking parameters. It can efficiently carry out carbon dioxide particle impact rock breaking tests under different working conditions and quickly obtain the optimal combination of engineering parameters, significantly improving test efficiency and parameter optimization accuracy. Attached Figure Description
[0036] The accompanying drawings are provided to further illustrate the embodiments of this application and form part of the specification. They are used together with the following detailed description to explain the embodiments of this application, but do not constitute a limitation on the embodiments of this application. In the drawings:
[0037] Figure 1 This is a process flow diagram.
[0038] Figure 2Schematic diagram of the overall structure of the carbon dioxide jet combined with particle impact rock breaking device.
[0039] Figure 3 for Figure 2 A partially enlarged schematic diagram of the impact system.
[0040] Figure Labels
[0041] Detailed Implementation
[0042] In this application example, the gas pressurization system can realize the continuous and controllable conversion of carbon dioxide between the liquid and gas phases, and the gas pressure and flow rate can be continuously adjusted within a preset range; the gas-solid mixing system can independently adjust the abrasive type, concentration, and particle size parameters, covering a variety of gas-solid ratio conditions; the impact testing system can apply different levels of confining pressure to simulate geostress environments at different depths; and the machine learning parameter optimization system can perform incremental training based on historical test data and new test data to continuously improve the prediction accuracy of the model under different lithologies and working conditions.
[0043] In this application example, it is necessary to optimize the range and step size of the test parameters based on actual engineering needs and rock characteristics to ensure the comprehensiveness and representativeness of the test data; at the same time, it is necessary to adjust the parameter settings of the intelligent optimization algorithm in combination with different optimization objectives to obtain the optimal parameter combination that meets the engineering requirements.
[0044] Other features, advantages, and processes of this application example will be described in detail in the following detailed description section.
[0045] The present invention will now be described in detail with reference to the accompanying drawings and embodiments.
[0046] This invention provides a carbon dioxide particle impact rock-breaking device and a machine learning optimization method. The device includes:
[0047] The device consists of five major hardware systems: a gas pressurization system, a gas-solid mixing system, an impact testing system, a material recovery and screening system, and a rock crushing and quantitative analysis system, along with a supporting machine learning parameter optimization module.
[0048] The gas pressurization system provides the entire device with carbon dioxide power sources of different phases, pressures, and flow rates, enabling precise control of the carbon dioxide phase. The gas pressurization system of this device is primarily composed of a gas storage tank 1, a gas pump 3, a pressurization pipe 4, a pipeline heat exchanger 2, a temperature control unit 21, a phase conversion controller, a safety valve 7, and monitoring instruments 22. The gas storage tank 1 stores liquid or gaseous carbon dioxide, and the gas pump 3 pressurizes and delivers the carbon dioxide. A pressure gauge 5, a flow rate gauge 6, and a safety valve 7 are sequentially installed on the pressurization pipe 4 to monitor the gas pressure and flow rate in real time and ensure safe system operation. The temperature sensor and phase conversion controller work together; based on the feedback signals from the pressure gauge 5 and the temperature sensor, the system adjusts the output pressure of the gas pump 3 and the heat exchange power of the pipeline heat exchanger 2 to achieve stable phase conversion of carbon dioxide. This system can achieve multi-dimensional independent adjustment of carbon dioxide phase, pressure, and impact velocity, and can systematically analyze the rock-breaking effect of various jets under different initial conditions, providing comprehensive experimental data support for the study of rock-breaking mechanisms.
[0049] The gas-solid mixing system achieves uniform mixing of abrasive and pressurized carbon dioxide, and allows independent adjustment of the abrasive type, concentration, and particle size parameters. The core components of this gas-solid mixing system are an abrasive storage tank 8, a control valve 101, a gas-solid mixing device 9, and a screw feeder 10. The outlet of the abrasive storage tank 8 is equipped with a replaceable screen to control the abrasive particle size range entering the screw feeder 10. The screw feeder 10 is used to precisely adjust the abrasive feed rate, thereby controlling the gas-solid mixing concentration. The gas-solid mixing device 9 contains an abrasive storage chamber 91 and a launch pipe 92, ensuring thorough and uniform mixing of the abrasive and pressurized carbon dioxide within the launch pipe 92. This system allows for multi-dimensional independent adjustment of abrasive type (quartz sand, garnet, white corundum), concentration (0%-30%), particle size (0.3mm-1mm), and gas-solid mass ratio. It can systematically analyze the rock-breaking effect under various mixing conditions, providing comprehensive experimental data support for the study of rock-breaking mechanisms.
[0050] The impact test system simulates the rock impact process under different confining pressure conditions and collects impact parameters and rock damage data in real time. The impact test system of this device is composed of an injection tube 11, a detachable nozzle device 121, a reverse high-speed jet nozzle 13, a high-pressure vessel 14, a confining pressure device 15, a high-speed camera 16, and a backflow tube 17. The injection pipe 11 is connected to the emission pipe 92 of the gas-solid mixing device 9, and delivers the mixed fluid to the detachable nozzle device 121. The detachable nozzle device 121 includes a device body 12, a detachable nut 122, and a nozzle 123. The device body 12 is provided with a gas flow channel, and an integrated pressure and velocity gauge 124 is installed on the flow channel to monitor the impact pressure and velocity at the nozzle tip in real time. The nozzle 123 can be quickly replaced with different types and diameters to meet different test requirements. The reverse high-speed jet nozzle 13 is located on the side wall of the detachable nozzle device 121, forming a negative pressure in the impact area, which drives rock chips and residual abrasive into the reverse flow pipe 17. The autoclave 14 is used to contain the rock specimen, and its side wall is provided with a transparent observation window 141. The high-speed camera 16 acquires images of the rock impact damage process through the observation window 141. The confining pressure device 15 is set around the autoclave 14 and can apply adjustable confining pressure to the rock specimen to simulate the geostress conditions at different depths. The system can achieve multi-dimensional independent adjustment of confining pressure (0-60MPa), nozzle type (straight orifice nozzle, Venturi nozzle and Laval nozzle), nozzle diameter (2mm-4mm), jet distance (3mm-9mm), and rock type (granite, sandstone and shale). It can systematically analyze the rock breaking effect under various conditions and provide comprehensive experimental data support for the study of rock breaking mechanism.
[0051] The material recovery and screening system achieves efficient separation of rock cuttings and abrasives, as well as the recycling and reuse of the abrasives. The core components of this system are a gas buffer device 18, an ultrasonic vibrating screen 19, and a recovery tank 20. The gas buffer device 18 is connected to the reflux pipe 17 to reduce the flow rate of the mixed fluid, preventing high-speed fluid from damaging subsequent equipment. The ultrasonic vibrating screen 19 screens the buffered mixture, achieving efficient separation of rock cuttings and abrasives. The recovery tank 20 collects the screened abrasives, enabling their recycling and reuse, and reducing testing costs.
[0052] The rock fragmentation quantitative analysis system enables quantitative assessment of rock impact damage, providing precise quantitative data support for rock breaking effect analysis. This system is primarily composed of a high-speed camera 16, 3D image processing software, a rock cuttings collection device, and a data analysis module. The high-speed camera 16 acquires images of the rock impact damage process in real time through a transparent observation window 141 on the side wall of the autoclave 14. The 3D image processing software performs 3D reconstruction of the images and calculates the impact damage depth on the rock surface. The rock cuttings collection device weighs the rock cuttings after screening by the ultrasonic vibrating screen 19 to obtain fragmentation quality data. The data analysis module comprehensively processes the image and rock cuttings data to calculate core indicators such as rock impact damage depth and fragmentation volume, providing high-quality target label data support for machine learning model training.
[0053] The test operation method for this device is as follows:
[0054] Step 1: Test Preparation. Fix the rock specimen inside the autoclave 14 and apply a preset confining pressure through the confining pressure device 15; select and install the corresponding nozzle 123 according to the test requirements;
[0055] Step 2: Install a matching replaceable screen to control the abrasive particle size range, adjust the feed rate of the two abrasive storage tanks to achieve different ratios, and adjust the screw feeder 10 to precisely control the abrasive feed rate.
[0056] Step 3: Carbon dioxide phase adjustment. Turn on the gas pump 3, adjust the gas pressure and temperature to the preset values through the phase conversion controller, and monitor the gas parameters through the pressure gauge 5 and the flow rate gauge 6 to achieve stable output of carbon dioxide in the preset phase.
[0057] Step 4: Gas-solid mixing and impact test. Open control valve 101 to allow the abrasive to enter the gas-solid mixing device 9, where it is fully mixed with pressurized carbon dioxide to form a gas-solid two-phase flow. The mixed fluid is ejected through the injection pipe 11 and the detachable nozzle device 121 to impact the rock specimen. At the same time, some gas is ejected through the reverse high-speed jet nozzle 13 to form a negative pressure, which carries rock chips and residual abrasive into the reverse flow pipe 17. During the test, the impact parameters and rock damage images are collected in real time by the integrated pressure and velocity gauge 124 and the high-speed camera 16.
[0058] Step 5: Material Recovery and Data Processing. Rock cuttings and residual abrasive enter the gas buffer device 18 through the reflux pipe 17 for deceleration, and are then sent to the ultrasonic vibrating screen 19 for sieving. The abrasive is recovered to the recovery tank 20, and the rock cuttings are collected separately for analysis of the degree of crushing. The test conditions and impact effect data are processed and entered to form a complete test dataset. Factors such as the phase state, pressure, and impact velocity of the carbon dioxide jet are independently adjusted in multiple dimensions by the gas pressurization system and obtained through experiments. Factors such as the type, particle size, concentration, and gas-solid mass ratio of abrasive are controllable and obtained through experiments by the gas-solid mixing system. Factors such as the magnitude of the confining pressure, nozzle type, nozzle diameter, jet distance, and rock type are obtained through simulation of the impact test system. The rock impact damage depth and crushing volume are obtained by three-dimensional reconstruction of the images using a high-speed camera and three-dimensional image processing software.
[0059] Based on the experimental data obtained above, this invention proposes a rock-breaking evaluation method using carbon dioxide jet combined with particle impact. This method uses jet pressure, abrasive concentration, abrasive particle size, confining pressure, jet distance, nozzle diameter, nozzle type, abrasive type, carbon dioxide phase, and rock type as input parameters, and measures and records the rock impact damage depth and crushing volume as target parameters.
[0060] The operation method is as follows:
[0061] The first step is data preprocessing and data segmentation. Outliers in the experimental data are identified and removed using the 3σ criterion, and replaced with the mean of three parallel tests under the same working conditions. Category characteristics such as nozzle type and abrasive type are coded using labels. The Min-Max standardization method is used to map all numerical data to the [0,1] interval to eliminate the influence of differences in the dimensions of different parameters. The calculation formula is as follows:
[0062] (1)
[0063] in, For standardized data, Raw data, and These represent the minimum and maximum values of the original data for the corresponding feature dimensions. The preprocessed standardized dataset is randomly divided into a training set, a validation set, and a test set in a ratio of 7:2:1. The training set is used for model fitting, the validation set is used for hyperparameter optimization and early stopping detection, and the test set is used to verify the model's generalization ability.
[0064] The second step is model training; the model adopts a leaf-wise tree growth strategy, and each time the node with the largest split gain is selected from all existing leaves for splitting. The split gain calculation formula (2) is as follows:
[0065] (2)
[0066] Among them, G L and G R H represents the sum of the first-order gradients of the left and right child nodes, respectively. L H R Let be the sum of the second-order gradients of the left and right child nodes, respectively, and λ be the L2 regularization coefficient. This is the splitting threshold;
[0067] The optimal output value of the leaf node is calculated using formula (3):
[0068] (3)
[0069] Where Ij is the set of samples contained in the j-th leaf node.
[0070] A multi-output prediction model was constructed using a RegressorChain regression structure and LightGBM as the basic regressor. The damage depth prediction results were introduced as input features for the subsequent fracture volume model by using the regression chain sequence. The intrinsic correlation between the two evaluation indicators was fully considered to achieve simultaneous prediction of the two indicators. The model adopted the form of an additive model (4):
[0071] (4)
[0072] in, For learning rate, Let be the predicted value of the t-th tree; optimize the model hyperparameters using the validation set error as an indicator; verify the model prediction accuracy using the test set, and if the accuracy does not reach the preset threshold, re-optimize the hyperparameters and retrain; its objective function adopts the second-order Taylor expansion approximation (5):
[0073] (5)
[0074] in, For the first-order gradient, It is a second-order gradient. This is a regularization term used to suppress model overfitting; an early stopping strategy is adopted, and training is terminated when the mean square error of the validation set no longer decreases for 10 consecutive rounds;
[0075] The model training module optimizes the hyperparameters of the LightGBM model using Bayesian optimization or grid search. The parameter prediction and optimization module predicts the impact effect based on the trained model and outputs the optimal combination of experimental parameters through an intelligent optimization algorithm. This system deeply integrates multi-parameter adjustable experimental equipment with LightGBM machine learning technology to establish a high-precision mapping model between experimental parameters and impact effects. Combined with an intelligent optimization algorithm, it quickly obtains the optimal combination of engineering parameters, solving the technical problems of low efficiency and insufficient accuracy of traditional experience-based trial-and-error methods.
[0076] Step 3: Intelligent parameter optimization; Using the trained LightGBM model as a proxy model for "process parameters - rock breaking effect", with the dual optimization objectives of maximizing impact crushing volume and impact depth, the feasible domain range and engineering constraints of each parameter are set, and the particle swarm optimization algorithm is used for global optimization to output the Pareto optimal parameter combination.
[0077] Particle swarm optimization algorithm, speed update formula (6):
[0078] v id (t+1)=w⋅v id (t)+c 1* r1(p id -x id (t))+c 2* r2(p gd -x id (t))(6)
[0079] v id (t): The velocity of particle ii in the dd-th dimension at generation tt.
[0080] x id (t): The position of particle ii in the dd-th dimension at generation tt.
[0081] p id The individual historical best position of particle ii
[0082] p gd The best historical position (single objective) or a guide selected from external archives (multi-objective).
[0083] w: Inertia weight (usually 0.4~0.9)
[0084] c1, c2: acceleration constants (usually taken as 2)
[0085] r1, r2: Random numbers within the range [0,1]
[0086] Particle swarm optimization algorithm, position update formula (7):
[0087] x id (t+1)=x id (t)+v id (t+1)(7)
[0088] Step 4: Experimental Validation and Model Update. Validation experiments are conducted based on the optimal parameter combination. The validation experiment data is added to the training set, the first 80% of the original model's tree structure is frozen, and incremental training is performed with a learning rate of 0.005 to continuously improve the model's prediction accuracy under target lithology and working conditions.
[0089] Example
[0090] Step 1: Initial preparation for the experiment;
[0091] Before conducting the carbon dioxide jet combined with particle impact rock breaking test, system initialization and rock block installation are completed first. The rock block (such as granite, sandstone, or shale) is fixed inside the autoclave 14, and a preset confining pressure (adjustable within the range of 0–60 MPa) is applied to the block through the confining pressure device 15 to simulate the geostress environment at different depths. According to the test design, the corresponding nozzle 123 is selected and installed; the nozzle type can be a straight-hole nozzle, Venturi nozzle, or Laval nozzle, with a diameter range of 2–4 mm. The jet distance is adjusted to the preset value (3–9 mm). The focus and lighting of the high-speed camera 16 and the transparent observation window 141 are checked to ensure clear recording of the rock impact damage process. After completing the above preparations, the gas pressurization system and the gas-solid mixing system are preheated and self-checked.
[0092] Step 2: Change and adjust particle parameters;
[0093] This device supports independent adjustment of various abrasive types, particle sizes, and concentrations, demonstrating flexibility in particle replacement and gradation. According to the experimental design, replaceable screens (0.3mm-1mm) with corresponding particle size ranges are installed at the outlets of abrasive storage tanks 81 and 82 to control the maximum particle size of the abrasive entering the system. The abrasives can be selected from quartz sand, garnet, or white corundum. The two abrasive storage tanks can be independently loaded with the same or different types (quartz sand, garnet, white corundum) and the same or different particle sizes of abrasives. The discharge rates of the two storage tanks can be controlled separately, thereby achieving arbitrary mass ratios of different abrasives. The outlets of the two storage tanks merge and are connected to control valve 101, which controls the on / off flow of the abrasive feed. This valve is then connected to a screw feeder 10, which precisely sets the abrasive feed rate, thereby adjusting the gas-solid mixing concentration (continuously adjustable within the range of 0% to 30%). The abrasive enters the abrasive storage chamber 91 of the gas-solid mixing device 9 via control valve 101, awaiting mixing with pressurized carbon dioxide. This step enables rapid replacement and precise proportioning of single-particle-size or mixed-particle-size abrasives of different particle sizes, materials, and concentrations.
[0094] Step 3: Change the phase state of carbon dioxide;
[0095] One of the core features of this device is the ability to achieve continuous and controllable conversion between the liquid and gaseous phases of carbon dioxide. The gas pump 3 is turned on, and the gas pressure and temperature are adjusted by the phase conversion controller: if a liquid carbon dioxide jet is required, the pressure is controlled at 5–10 MPa and the temperature below 31°C; if a supercritical or gaseous jet is required, the temperature is increased and the pressure is decreased through the pipeline heat exchanger 2 and the temperature regulation unit 21. Feedback signals from the pressure gauge 5 and flow rate gauge 6 are read in real time, and the phase conversion controller uses a closed-loop adjustment to regulate the gas pump output pressure and heat exchange power to ensure the stability of the carbon dioxide phase. After reaching the preset phase, pressure (10–30 MPa), and flow rate, the device operates stably for 5 minutes, providing a power source for gas-solid mixing and impact tests.
[0096] Step 4: Impact testing and parameter measurement to establish a mapping relationship;
[0097] Control valve 101 is opened, and the abrasive enters the launch pipe 92 of the gas-solid mixing device 9, where it mixes thoroughly with pressurized carbon dioxide to form a gas-solid two-phase flow. The mixed fluid is sent through the injection pipe 11 to the detachable nozzle device 121, and is ejected at high speed from the nozzle 123 to impact the rock sample. Simultaneously, if... Figure 3 As shown, the reverse high-speed jet nozzle 13 ejects part of the gas, creating a negative pressure inside the reverse flow pipe, which draws rock chips and residual abrasive into the material recovery system through the reverse flow pipe 17.
[0098] The measurement parameters include:
[0099] Input parameters: jet pressure (MPa), abrasive concentration (%), abrasive particle size (mm), confining pressure (MPa), jet distance (mm), nozzle diameter (mm), nozzle type, abrasive type, carbon dioxide phase (gas / liquid), rock type.
[0100] Output parameters: The impact process image is acquired in real time by high-speed camera 16, and the rock impact damage depth (mm) is obtained by three-dimensional reconstruction by three-dimensional image processing software; the rock fragments separated by ultrasonic vibrating screen 19 are collected, weighed and combined with the impact pit volume measurement to obtain the crushed volume (cm³).
[0101] Based on multiple sets of experimental data, a nonlinear mapping relationship between "input parameters → output parameters" is established. Specifically, by changing any combination of parameters in steps one through three, the corresponding damage depth and fracture volume are recorded to form a standardized experimental dataset. This dataset is used for subsequent machine learning model training, enabling the model to predict the rock-breaking effect under arbitrary parameter combinations, thereby revealing the influence of each parameter on the crushing efficiency.
[0102] Step 5: Parameter optimization;
[0103] The standardized experimental dataset obtained in step four is used for parameter optimization. Specifically, the method described in any one of claims 8 to 12 is employed.
[0104] As a specific implementation of this example, a total of 150 effective working condition tests were conducted. The range of input parameter values is shown in Table 1. The target parameters are damage depth (mm) and breakage volume (cm³).
[0105] Table 1. Input feature parameters and value ranges of the rock-breaking effect regression prediction model.
[0106]
[0107] After data preprocessing, the LightGBM chain regression model was trained using the hyperparameters listed in Table 2 (optimized by grid search).
[0108] Table 2. Optimization parameters for the LightGBM algorithm and optimization results for network search.
[0109]
[0110] Using the trained model as a surrogate model, a particle swarm optimization algorithm (the velocity and position update formulas are given in claim 11) is used for bi-objective optimization to obtain the Pareto optimal frontier. Three sets of optimal parameter combinations on the frontier are selected for verification experiments, and the verification data are used for incremental updates of the model.
[0111] The impact device and the machine learning parameter optimization system work together to form an integrated mode of "gas pressurization-gas-solid mixing-impact test-recovery screening-damage quantification-intelligent optimization". This mode effectively combines the controllable conversion of carbon dioxide in multiple phases, independent adjustment of multi-dimensional parameters and intelligent optimization of rock breaking parameters. It can efficiently carry out carbon dioxide particle impact rock breaking tests under different working conditions and quickly obtain the optimal combination of engineering parameters, significantly improving test efficiency and parameter optimization accuracy.
[0112] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.
[0113] The above are merely examples of this application and are not intended to limit this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of the claims of this application.
Claims
1. A rock-breaking evaluation device combining carbon dioxide jet and particle impact, comprising five hardware systems: a gas pressurization system, a gas-solid mixing system, an impact testing system, a material recovery and screening system, and a rock crushing quantitative analysis system, characterized in that, It also includes a machine learning parameter optimization module; The gas pressurization system includes a gas storage tank (1), a gas pump (3), a pipeline heat exchanger (2), a temperature control unit (21), and a monitoring instrument (22). The outlet of the gas storage tank (1) is connected to the inlet of the gas pump (3), and the inlet of the gas pump (3) is connected to the outlet of the gas storage tank (1) through the pipeline heat exchanger (2). The gas pump (3) is connected to the gas-solid mixing system through a pressurization gas pipe (4). The temperature control unit (21) and the monitoring instrument (22) are electrically connected to the pipeline heat exchanger (2) and feedback control of its heat exchange power. The phase conversion is achieved by adjusting the temperature and pressure of CO2 in the pipeline, which is used to realize the continuous and controllable conversion of carbon dioxide liquid phase to gas phase, and to precisely control the pressure and flow rate of the output gas. The gas-solid mixing system includes a first abrasive storage tank (81), a second abrasive storage tank (82), a screw metering feeder (10), and a gas-solid mixing device (9). The outlets of the first abrasive storage tank (81) and the second abrasive storage tank (82) are connected in parallel and then connected to the inlet of the screw metering feeder (10). The outlet of the screw metering feeder (10) is connected to the abrasive storage chamber (91) of the gas-solid mixing device (9). The gas inlet of the gas-solid mixing device (9) is connected to the outlet of the air pump (3). The mixed fluid outlet of the gas-solid mixing device (9) is connected to the impact test system. The two abrasive storage tanks are set in parallel to adjust the abrasive type, particle size, concentration, and gas-solid mass ratio, and output a uniform and stable gas-solid two-phase fluid. The impact testing system includes a main body (12), a reverse high-speed jet nozzle (13), an autoclave (14), a confining pressure device (15), a high-speed camera (16), and a backflow pipe (17). The inlet end of the main body (12) is connected to the outlet of the injection pipe (11), and the mixed fluid outlet of the gas-solid mixing device (9) is connected to the inlet of the main body (12) via the injection pipe. The nozzle (123) of the main body (12) extends into the autoclave (14) and is positioned towards the rock specimen. The flow nozzle (13) is located on the side wall of the detachable nozzle device (121) of the main body (12) of the device and its spray direction is towards the inlet of the backflow pipe (17). The confining pressure device (15) is arranged around the outside of the high pressure vessel (14). The high-speed camera (16) is located outside the transparent observation window (141) of the high pressure vessel (14). The outlet of the backflow pipe (17) is connected to the material recovery and screening system. It is used to simulate the rock impact process under different depth stress environments and collect impact parameters and rock damage process images. The material recycling and screening system includes a gas buffer device (18), an ultrasonic vibrating screen (19), and a recycling tank (20). The outlet of the backflow pipe (17) is connected to the inlet of the gas buffer device (18), the outlet of the gas buffer device (18) is connected to the inlet of the ultrasonic vibrating screen (19), and the abrasive outlet of the ultrasonic vibrating screen (19) is connected to the recycling tank (20) for separating rock chips and abrasives to achieve the recycling and reuse of abrasives. The rock breaking quantitative analysis system includes the high-speed camera (16), a three-dimensional image processing module and a rock cutting weighing module, which are used to process the collected images and rock cutting data, and to quantify and calculate the core indicators of rock breaking effect. The machine learning parameter optimization module is used to construct a LightGBM prediction model based on quantified data, and output the optimal rock-breaking parameter combination by combining a genetic algorithm or a particle swarm optimization algorithm. Based on this, it controls or prompts the operator to adjust the set values of the gas pressurization system and the gas-solid mixing system for verification experiments. The machine learning parameter optimization module is also used to feed the verification experiment data back to the training set for incremental learning. It is used to guide the parameter reset of the gas pressurization system and / or the gas-solid mixing system. This module is a hardware and software system integrated in the computer, and its function depends on the test data collected by the aforementioned hardware system to be realized.
2. The rock-breaking evaluation device for carbon dioxide jet combined with particle impact according to claim 1, characterized in that, The gas pressurization system also includes a temperature sensor, a pressure gauge (5), and a flow rate gauge (6). The temperature sensor is installed on the pressurized gas pipe (4) at the outlet of the pipeline heat exchanger (2). The pressure gauge (5) and the flow rate gauge (6) are installed sequentially on the pressurized gas pipe (4) at a position after the pipeline heat exchanger (2). The temperature regulation unit (21) adjusts the output pressure of the gas pump (3) and the heat exchange power of the pipeline heat exchanger (2) according to the gas pressure and temperature feedback signals to achieve stable phase conversion of carbon dioxide.
3. The rock-breaking evaluation device for carbon dioxide jet combined with particle impact according to claim 1, characterized in that, The first abrasive storage tank (81) and the second abrasive storage tank (82) are respectively provided with replaceable screens at their outlets to control the abrasive particle size range entering the screw metering feeder (10); the outlet of the screw metering feeder (10) is connected to the inlet of the abrasive storage chamber (91) of the gas-solid mixing device (9) to precisely adjust the abrasive feeding rate; the gas-solid mixing device (9) is provided with an abrasive storage chamber (91) and an emission pipe (92), the outlet of the abrasive storage chamber (91) is connected to the inlet of the emission pipe (92), and the air inlet of the emission pipe (92) is connected to the outlet of the pipeline heat exchanger (2) to fully mix the abrasive with pressurized carbon dioxide.
4. The rock-breaking evaluation device for carbon dioxide jet combined with particle impact according to claim 1, characterized in that, The main body (12) of the device includes a detachable nozzle device (121), a detachable nut (122) and a replaceable nozzle (123). The replaceable nozzle (123) is fixedly installed at the outlet end of the main body (12) through the detachable nut (122). The detachable nozzle device (121) integrates an integrated air pressure and flow rate meter (124) for real-time monitoring of the impact air pressure and flow rate at the nozzle inlet.
5. The rock-breaking evaluation device for carbon dioxide jet combined with particle impact according to claim 1, characterized in that, The reverse high-speed jet nozzle (13) is fixedly installed on the side wall of the detachable nozzle device (121). Its inlet is connected to an external compressed air source, and its outlet faces the inlet direction of the reverse flow pipe (17). It is used to form a negative pressure in the impact area and drive rock chips and residual abrasive into the reverse flow pipe (17).
6. The rock-breaking evaluation device for carbon dioxide jet combined with particle impact according to claim 1, characterized in that, The inlet of the gas buffer device (18) is connected to the outlet of the reflux pipe (17) to reduce the flow rate of the mixed fluid; the outlet of the gas buffer device (18) is connected to the feed inlet of the ultrasonic vibrating screen (19), and the undersize outlet of the ultrasonic vibrating screen (19) is connected to the inlet of the recovery tank (20) to separate rock chips and abrasives and collect the screened abrasives.
7. The rock-breaking evaluation device for carbon dioxide jet combined with particle impact according to claim 1, characterized in that, The rock fragmentation quantitative analysis system performs three-dimensional reconstruction of images acquired by the high-speed camera (16) and combines rock fragment sieving analysis to quantitatively calculate the rock impact damage depth and fragmentation volume core indicators.
8. A method for evaluating and optimizing the rock-breaking parameters of a carbon dioxide jet combined with particle impact, applied to the apparatus described in any one of claims 1 to 7, characterized in that, Includes the following steps: S1 Test preparation: Fix the rock specimen in the autoclave (14) and apply the preset confining pressure through the confining pressure device (15). Load the specified abrasive into the abrasive storage tank (81, 82) and install the matching replaceable nozzle (123). S2 Phase adjustment: Turn on the gas pump (3) and adjust the gas pressure and temperature to the preset value through the phase conversion controller to achieve stable output of carbon dioxide in the preset phase state; S3 Impact Test: The abrasive is mixed with pressurized carbon dioxide to form a gas-solid two-phase flow, which is then ejected through a replaceable nozzle (123) to impact a rock specimen. Impact parameters and rock damage images are collected in real time. S4 Material Recovery and Data Quantification: Separate rock cuttings and abrasives, calculate the quantitative index of rock breaking effect through the rock crushing quantitative analysis system, and form a standardized test dataset; S5 Parameter Optimization: With the dual optimization objectives of maximizing rock damage depth and fracture volume, the LightGBM prediction model is trained based on the experimental dataset. The Pareto optimal combination of rock breaking parameters is output by combining the genetic algorithm or particle swarm optimization algorithm, and the validation experimental data is fed back to the model for incremental learning. The intelligent optimization in step S5 specifically involves: taking the maximization of impact fracture volume and impact depth as the optimization objective, setting the feasible range of each test parameter, using a genetic algorithm or particle swarm optimization algorithm for global optimization, and outputting the Pareto optimal parameter combination; it also includes a model update step: conducting verification experiments based on the optimal parameter combination, adding the verification experiment data to the training set, incrementally training the LightGBM model, and continuously improving the model's prediction accuracy under target lithology and working conditions.
9. The rock breaking evaluation and parameter optimization method according to claim 8, characterized in that, Step S5 also includes a data preprocessing step: outliers are identified and removed using the 3σ criterion, outliers are replaced by the mean of parallel test data under the same working conditions, and the data are mapped to the [0,1] interval using the Min-Max standardization method.
10. The rock-breaking evaluation and parameter optimization method according to claim 8, characterized in that, The model training in step S5 is as follows: the preprocessed dataset is divided into training set, validation set and test set according to the proportion. Based on the training set, a LightGBM multi-output prediction model is built using a chain regression framework. The correlation between rock damage depth and fracture volume is modeled using the regression chain. The two indicators are predicted simultaneously. The hyperparameters of the model are optimized using grid search or Bayesian optimization algorithm. The optimization objective is to minimize the error of the validation set and maximize the coefficient of determination R². The prediction accuracy of the model is verified using the test set.
Citation Information
Patent Citations
A method and system for simulating and optimizing a particle impact rock breaking process
CN116306332B
Particle jet coupling resonance rock breaking device and rock breaking method
CN117489264B
A rock-breaking device and method combining liquid nitrogen jet and particle impact.
CN119754765B
