A machine learning-based PDC bit design method, system, and storage medium

Abstract

The application provides a machine learning-based PDC bit design method, system and storage medium, and belongs to the field of oil and gas drilling engineering equipment and technology. The method comprises the following steps: data acquisition and fusion; prediction model construction; multi-objective optimization; output of recommended bit optimal parameter combination and its predicted performance and wear index. The system comprises: a three-dimensional scanning data acquisition and processing module; an intelligent identification module; a model management module; an optimization solving module; a man-machine interaction and output module. The storage medium has a computer program stored thereon, and the program is executed by a processor to realize the above method. The application realizes a data-driven intelligent design closed loop, effectively solves the industry problem that the drilling bit design efficiency and wear resistance are difficult to balance under complex formation conditions. The method can significantly improve the rate of penetration and total footage of a single bit, reduce the non-production time and comprehensive drilling cost, and is especially suitable for complex formation and deep well drilling operations.

Description

Technical Field

[0001] This invention relates to the field of oil and gas drilling engineering equipment and technology, and in particular to a PDC drill bit design method, system and storage medium based on machine learning. Background Technology

[0002] PDC drill bits are key rock-breaking tools in modern oil and gas exploration and development, and their working efficiency and lifespan directly affect the cycle and cost of drilling projects. The structural design parameters of the drill bit, such as the back rake angle of the inner cone, nose, and outer cone, are the core factors that determine its cutting efficiency, mechanical drilling speed, directional control, impact resistance, and wear resistance.

[0003] Currently, the design of PDC drill bits relies heavily on engineers' limited experience, analogies to historical designs, and simulations based on logging data such as sonic transit time or single physical models. Traditional methods have significant limitations: First, conventional PDC drill bit design relies heavily on human experience, making the experience of drill bit design engineers a core asset of the company. Data silos and long validation cycles exist: drilling data for a single well is scattered across oil companies, service companies, and drill bit companies, making it extremely difficult to obtain complete, high-quality data for modeling and analysis. Furthermore, the success or failure of a drill bit design requires validation through field drilling, resulting in extremely high trial-and-error costs. Human experience is also highly subjective, difficult to quantify and pass on, and cannot effectively handle complex and variable formation conditions. Secondly, laboratory indoor tests and numerical simulations are costly, time-consuming, and have a significant gap with the actual working conditions downhole. Finally, existing methods typically consider efficiency and wear resistance as sequential or independent objectives, lacking systematic automated optimization methods that can simultaneously address multiple competing objectives.

[0004] With the development of big data and artificial intelligence technologies, existing research has attempted to predict drill bit performance using data-driven methods. However, most studies only use operating condition data or single-type data, failing to deeply integrate the precise structural characteristics of the drill bit before use, the objective wear state after use, and the real performance data during drilling to form a closed-loop optimization design system. Therefore, developing a PDC drill bit design method that can integrate multi-source heterogeneous data and automatically optimize through intelligent algorithms is of great practical significance for improving drilling technology and reducing costs and increasing efficiency. Summary of the Invention

[0005] In view of this, the present invention aims to propose a PDC drill bit design method, system and storage medium based on machine learning. By integrating advanced measurement technology, machine learning and optimization algorithms, and fusing multi-source data (parameters, wear levels, drilling performance), a predictive model is established, and the drill bit design parameters are automated, intelligent and optimized through genetic algorithms and multi-objective optimization.

[0006] To achieve the above objectives, the technical solution of the present invention is implemented as follows: a PDC drill bit design method based on machine learning, comprising the following steps: Step S1, Data Acquisition and Fusion; The three-dimensional point cloud parameters of the PDC drill bit before use, the wear level data of the PDC drill bit after use, and the drilling performance parameters of the PDC drill bit are obtained. The above multi-source data are cleaned, aligned and normalized to form a standard dataset for model training. Step S2, prediction model construction; Based on machine learning algorithms, using random forest regression or ensemble learning algorithms, with the aforementioned 3D point cloud as input features and drilling performance parameters and wear levels as prediction targets, a high-performance regression and classification model is trained to establish a nonlinear mapping relationship between the 3D geometric features corresponding to the drill bit design and drilling performance parameters and wear levels. Step S3, multi-objective optimization; A multi-objective optimization function is constructed with the goal of minimizing wear level and maximizing drilling performance; a multi-objective evolutionary algorithm is applied to perform a global search within the feasible region of drill bit design parameters to obtain a set of Pareto optimal solutions; Step S4, Decision and Output; Based on actual drilling conditions, the final solution is selected from the Pareto optimal solution set, and the recommended optimal combination of drill bit parameters and its predicted performance and wear indices are output.

[0007] Furthermore, in step S1, the parameters of the PDC drill bit before use include at least the inner cone backslope angle, the nose backslope angle, and the outer cone backslope angle; the wear level of the PDC drill bit after use includes at least the wear level of the inner cone cutting teeth, the wear level of the nose cutting teeth, and the wear level of the outer cone cutting teeth; the drilling performance parameters of the PDC drill bit include at least the total drilling footage and the average mechanical drilling speed.

[0008] Furthermore, in step S1, the initial 3D point cloud is acquired using a non-contact 3D laser scanner, and the 3D point cloud reconstruction error is less than 0.05mm.

[0009] Furthermore, in step S1, the wear level is identified through an image recognition algorithm combined with expert review; drilling performance parameters are automatically extracted and parsed from the daily drilling report or real-time drilling database through a data interface.

[0010] Furthermore, the machine learning algorithm in step S2 is a random forest regression algorithm, gradient boosting tree, or support vector machine; When training high-performance regression and classification models, the criteria for judging high performance include: The high-performance criterion includes the regression model's coefficient of determination R on the test set.2 The mean square error or average absolute error is greater than or equal to 0.8, and the mean square error or average absolute error is reduced by a preset percentage compared to the baseline model; the accuracy of the classification model in the wear level identification task is greater than or equal to 85%, and the macro average F1 value is greater than or equal to the preset threshold.

[0011] Furthermore, the multi-objective optimization algorithm in step S3 employs a fast non-dominated sorting genetic algorithm with an elitist strategy.

[0012] Furthermore, the optimization function in step S3 introduces weighting coefficients, allowing users to flexibly adjust the optimization bias based on actual formation lithology, drilling cost structure, or efficiency-first strategy.

[0013] The present invention also provides a PDC drill bit parameter design optimization system for implementing the above method, comprising: The 3D scanning data acquisition and processing module is used to automatically acquire and preprocess the drill bit's parameters, wear, and performance data. The intelligent recognition module integrates an image recognition model for automatically assessing the wear level of cutting teeth; The model management module is used to store, train, and retrieve performance prediction models; The optimization and solution module is used to configure optimization objectives and constraints, execute multi-objective optimization algorithms, and output the optimal parameter combination. The human-computer interaction and output module is used to visualize the Pareto frontier, recommended parameters, and prediction results, and to receive user decision-making instructions.

[0014] Furthermore, the 3D scanning data acquisition and processing module integrates a 3D scanning control unit and a data communication interface.

[0015] The present invention also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the above-described method.

[0016] Compared with existing technologies, the PDC drill bit design method, system, and storage medium based on machine learning described in this invention have the following advantages: (1) Data-driven systematic optimization: For the first time, multi-source data (3D point cloud, performance, wear) throughout the entire life cycle of the drill bit are effectively integrated, breaking down data silos and laying a solid data foundation for intelligent design; (2) Objective and accurate condition assessment: Image recognition technology is used to replace subjective judgment by human eyes, so as to realize rapid, objective and quantitative assessment of the wear level of cutting teeth, which greatly improves the accuracy and consistency of data; (3) High-precision prediction capability: Using advanced machine learning algorithms, it can accurately capture the mapping relationship between design parameters and drill bit performance under complex geological conditions, and the prediction accuracy is significantly higher than that of traditional empirical methods; (4) Multi-objective global optimum: Through multi-objective optimization algorithm, the design space can be automatically explored to find a set of optimal solutions rather than a single one, providing engineers with a variety of feasible and excellent design schemes that take into account conflicting objectives, and making decision-making highly flexible; (5) Significant economic benefits: Practical application shows that the drill bit designed with the parameters recommended by the method of this invention can simultaneously improve mechanical drilling speed and footage, reduce wear, and reduce the number of trips in and out of the hole, thereby significantly reducing the overall drilling cost. Attached Figure Description

[0017] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an undue limitation of the invention. In the drawings: Figure 1 This is a flowchart illustrating the overall implementation of a machine learning-based PDC drill bit design method according to an embodiment of the present invention. Figure 2 This is a schematic diagram of the Pareto front distribution obtained by NSGA-II optimization according to the present invention; Figure 3 This is a schematic diagram comparing the recommended drill bit design parameters, predicted performance, and baseline parameters described in this invention. Detailed Implementation

[0018] It should be noted that, unless otherwise specified, the embodiments and features described in the present invention can be combined with each other.

[0019] The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0020] This invention relates to a machine learning-based PDC drill bit design method. It is an intelligent, data-driven design method for PDC (Polycrystalline Diamond Compact) drill bits that combines 3D measurement, machine vision, machine learning, and multi-objective optimization algorithms. Specifically, it relates to drill bit design and optimization technology. The method first uses 3D laser scanning technology to accurately acquire the structural parameters of the drill bit before use, including the inner cone backslope angle, nose backslope angle, and outer cone backslope angle, and records this data into the drilling daily report system. Second, it employs a machine learning-based image recognition model, combined with an expert knowledge base, to automatically and quantitatively assess the wear level of the cutting teeth at various parts of the drill bit after use. Simultaneously, it automatically extracts post-use performance indicators such as total drilling footage and average mechanical drilling speed from the drilling daily report system. Based on this, it uses machine learning algorithms such as random forests to construct an accurate predictive model of drill bit structural parameters, performance, and wear, and couples this model with multi-objective genetic algorithms such as NSGA-II to perform global optimization with the goal of simultaneously maximizing drilling efficiency and minimizing wear. Finally, it outputs a Pareto-optimal combination of drill bit design parameters that balances wear resistance and drilling efficiency. Figure 1 The diagram illustrates the complete closed-loop process from data acquisition, processing, modeling, optimization to output, specifically including the following steps: Step S1, Data Acquisition and Fusion; By comprehensively utilizing 3D laser scanning, machine vision, and data mining technologies, high-precision 3D point clouds of the drill bit before use, wear levels of multiple parts after use, and actual drilling performance indicators are obtained. The above multi-source data is then cleaned, aligned, and normalized to form a standard dataset for model training, thus constructing a comprehensive drill bit design-performance-wear database. Details are as follows: (1) Parameter acquisition: The parameters of the drill bit before use are measured using a three-dimensional laser scanner, including the inner cone back tilt angle, the nose back tilt angle, and the outer cone back tilt angle.

[0021] (2) Wear level identification: The wear characteristics of the drill bit after use are automatically extracted by image recognition algorithm and combined with expert review to obtain the wear level of the inner cone cutting teeth, the wear level of the nose cutting teeth, and the wear level of the outer cone cutting teeth.

[0022] (3) Drilling performance parameters: automatically extracted and parsed from the drilling daily report software or real-time drilling database through the data interface.

[0023] The initial 3D point cloud was acquired using a non-contact 3D laser scanner, and the 3D point cloud reconstruction error was less than 0.05mm.

[0024] This invention achieves the objective quantification of structural parameters of drill bits before use and the automatic assessment of wear levels after use through high-precision three-dimensional laser scanning and machine learning-based image recognition technology. This combination of technologies solves the problems of strong subjectivity and poor consistency caused by traditional reliance on human experience, and provides an accurate and traceable data foundation for model construction.

[0025] Step S2, prediction model construction; Based on machine learning algorithms, using random forest regression or ensemble learning algorithms, and taking the aforementioned 3D point cloud as input features, drilling performance parameters and wear levels as prediction targets, a high-performance regression and classification model is trained to establish a complex nonlinear mapping relationship between design parameters, drilling performance parameters, and wear levels. This constructs a high-precision, robust prediction model to achieve forward-looking performance predictions for new design schemes. The machine learning algorithms used are random forest regression, gradient boosting tree (GBDT), or support vector machine (SVM). The criteria for judging high performance in training the regression and classification model include: the coefficient of determination R0 of the regression model on the test set. 2 The mean square error or average absolute error is greater than or equal to 0.8, and the model's accuracy in wear level identification is greater than or equal to 85% compared to the baseline model, with the macro-average F1 value greater than or equal to a preset threshold. This ensures the model's stability and generalization ability in drilling performance prediction and wear level identification.

[0026] The prediction model is shown below: f:(inner_angle,nose_angle,outer_angle)→(footage,ROP,inner_wear,nose_wear,outer_wear) in, inner_angle: Inner cone back tilt angle (°); nose_angle: Nasal posterior tilt angle (°); outer_angle: external cone back tilt angle (°); Footage: Maximum drill bit advance (in meters); ROP: Maximize mechanical drilling rate (m / h); inner_wear: Minimizes inner cone wear, dimensionless; nose_wear: Minimizes wear on the nose, dimensionless; outer_wear: Minimizes outer cone wear, dimensionless; Step S3, multi-objective optimization; A multi-objective optimization function is constructed with the goal of minimizing wear level and maximizing drilling performance; a multi-objective evolutionary algorithm is applied to perform a global search within the feasible region of drill bit design parameters to obtain a set of Pareto optimal solutions; Among them, the multi-objective optimization algorithm adopts the fast non-dominated sorting genetic algorithm (NSGA-II) with elitist strategy.

[0027] The optimization function introduces weighting coefficients, allowing users to flexibly adjust the optimization bias based on actual formation lithology, drilling cost composition, or efficiency-first strategy.

[0028] Specifically as follows: Based on the established prediction model, the drill bit design problem is formalized into a multi-objective optimization problem, which is solved using evolutionary algorithms such as NSGA-II to generate a series of Pareto optimal design schemes that achieve the best balance between efficiency and wear resistance. This invention takes "improving drill bit efficiency and reducing wear level" as the optimization objective, and constructs a multi-objective optimization problem with the objective function defined as follows; F( )= in, Minimizes internal cone wear, dimensionless; Minimizes nasal abrasion; dimensionless. Minimizes external cone wear, dimensionless; Total drill bit advance, in meters; Maximize the mechanical drilling rate (meters per hour); Weights of each objective, dimensionless.

[0029] Genetic algorithms are used to search for the optimal solution within the parameter range.

[0030] Step S4, Decision and Output, namely, the output and application of optimization results; Based on actual drilling conditions, a final solution is selected from the Pareto optimal solution set, and the recommended optimal combination of drill bit parameters and their predicted performance and wear indices are output. Specifically, the optimization results are presented to the designers in a visual manner (such as a Pareto front plot) to assist them in making final decisions based on specific engineering requirements. The recommended optimal output parameters (inner cone backslope angle, nose backslope angle, outer cone backslope angle) are output, and their corresponding total drill bit footage, average mechanical drilling rate, and wear level are predicted.

[0031] The present invention also provides a PDC drill bit parameter design optimization system for implementing the above method, comprising: The 3D scanning data acquisition and processing module is used to automatically acquire and preprocess the drill bit's parameters, wear, and performance data; the 3D scanning data acquisition and processing module integrates a 3D scanning control unit and a data communication interface.

[0032] The intelligent recognition module integrates an image recognition model for automatically assessing the wear level of cutting teeth; The model management module is used to store, train, and retrieve performance prediction models; The optimization and solution module is used to configure optimization objectives and constraints, execute multi-objective optimization algorithms, and output the optimal parameter combination. The human-computer interaction and output module is used to visualize the Pareto frontier, recommended parameters, and prediction results, and to receive user decision-making instructions.

[0033] The present invention also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the above-described method.

[0034] The core innovation of this invention lies in constructing a data-driven intelligent design closed loop, where a clear causal logic chain exists between its technical features and effects: First, by using high-precision three-dimensional laser scanning and machine learning-based image recognition technology, the objective quantification of the structural parameters of the drill bit before use and the automatic assessment of the wear level after use were realized respectively. This combination of technologies solves the problems of strong subjectivity and poor consistency caused by traditional reliance on human experience, and provides an accurate and traceable data foundation for model construction. Secondly, by using machine learning algorithms such as random forest, multi-source heterogeneous data (structural parameters, performance indicators, wear levels) are deeply fused and jointly modeled, establishing a nonlinear mapping relationship from design parameters to performance and wear. This feature breaks through the limitations of traditional single data source or physical simulation, enabling the prediction model to have high accuracy and strong generalization ability. Furthermore, by introducing multi-objective optimization algorithms such as NSGA-II, the competing objectives of efficiency and wear resistance are simultaneously incorporated into the optimization function for global search. This technical mechanism overcomes the shortcomings of traditional sequential optimization or independent optimization and can automatically generate a series of Pareto optimal solutions. Finally, the optimization results are visualized in the form of Pareto front through the human-computer interaction interface, and users are allowed to adjust the weights and make scheme decisions according to actual working conditions and preferences. This feature enables engineers to flexibly weigh multiple objectives, thereby outputting the optimal combination of drill bit parameters that meets both efficiency requirements and wear resistance.

[0035] The synergistic effect of the above-mentioned technical features systematically solves the industry problem of balancing drill bit design efficiency and wear resistance in complex formations, ultimately achieving a significant increase in mechanical drilling speed and total footage, as well as an effective reduction in non-productive time and overall costs.

[0036] Compared to traditional design methods that rely on manual experience, this invention achieves a data-driven intelligent design closed loop, deeply integrating drill bit structural features, operating data, and performance feedback. This effectively solves the industry challenge of balancing drill bit design efficiency and wear resistance under complex formation conditions. This method can significantly improve mechanical drilling speed and total footage per drill bit, while reducing non-productive time and overall drilling costs. It is particularly suitable for drilling operations in complex formations and deep wells, and has promising engineering application prospects and widespread application value.

[0037] Taking a drilling project in an oilfield in the Bohai Sea of ​​China as an example, 3,000 sets of complete PDC drill bit data samples were collected from multiple wells.

[0038] (1) During the data acquisition (step S1) stage: A high-precision 3D laser scanner (such as GOM ATOS) is used to acquire 3D point cloud data of the drill bit before use. After reconstruction, the inner cone backslope angle, nose backslope angle, and outer cone backslope angle are accurately measured with a measurement error of less than ±0.1°.

[0039] After the drill bit is retrieved and cleaned, it is placed in a standard imaging chamber where high-resolution images of the cutting teeth at various locations are captured by an industrial camera. A pre-trained ResNet convolutional neural network model is used for feature extraction and preliminary wear level classification. The preliminary wear level classification results are transmitted to an expert review interface for final confirmation and correction by drilling experts, resulting in labels.

[0040] The developed data interface program automatically extracts the total footage (meters) and average mechanical drilling speed (meters / hour) data of the corresponding drill bit from the drilling daily report software.

[0041] (2) In the model building and training (step S2) stage: The cleaned and aligned data were divided into training and test sets in a 7:3 ratio.

[0042] The prediction model was trained using a random forest regression algorithm. It used three backdip angle parameters as input to simultaneously predict five target variables: drilling footage, drilling speed, and three wear levels.

[0043] The model performance was evaluated using 10-fold cross-validation. The results show that the model's coefficient of determination (R²) for predicting drill bit footage and mechanical drilling speed on the test set is [value missing]. 2 The F1-Scores for wear level prediction reached 0.87 and 0.84 respectively; the macro-average F1-Score for wear level prediction reached over 0.85, which meets the criteria for high-performance regression and classification models, thus the model shows excellent predictive ability.

[0044] (3) In the multi-objective optimization stage (step S3) Define optimization objectives: maximize footage, maximize rate of drilling (ROP); minimize inner wear, minimize nose wear, and minimize outer wear.

[0045] Set constraints: The backslope angle parameter value range is set between 10° and 35° depending on the drill bit size and type.

[0046] The NSGA-II algorithm was used for optimization, with a population size of 100 and 200 generations.

[0047] After optimization, the result is as follows: Figure 2 The Pareto optimal solution set is shown. Figure 2 In the diagram, the horizontal axis represents wear-related objectives, the vertical axis represents drilling efficiency-related objectives, and each point represents a non-dominated optimal solution.

[0048] (4) In the result output stage (step S4): Based on the "speed-first" strategy for this well section, the drill bit design engineer selected an optimal solution from the Pareto front.

[0049] The corresponding recommended design parameters are: internal kyphosis angle = 22°, nasal kyphosis angle = 18°, external kyphosis angle = 25°; The model predicts the following performance under this parameter combination: drill bit footage ≈ 380 meters, average mechanical drilling speed ≈ 21 meters / hour, and wear levels of the inner cone / nose / outer cone are 2 / 3 / 2 respectively. Figure 3 The recommended approach and the baseline approach were compared intuitively to demonstrate their performance improvements across various objectives. Figure 3 The improvement in various indicators such as efficiency and wear is displayed intuitively in the form of bar charts.

[0050] Application Results: The recommended parameters were applied to the same formation sections in four subsequent wells. Field practice showed that, compared with drill bits designed using traditional methods, the drill bits designed using the present invention achieved an average increase in footage of 15.2%, an average mechanical drilling speed of 17.8%, and an average wear level reduction of approximately 20%, resulting in significant economic benefits.

[0051] Pareto frontier analysis shows that when pursuing higher mechanical drilling speeds, wear levels will increase, but with reasonable weighting, a balance between efficiency and lifespan can still be found.

[0052] Furthermore, this invention also conducted model uncertainty analysis using the Bootstrap sampling method. The results showed that at a 95% confidence level, the error fluctuation range of the key performance indicators predicted by the model was within ±5%, demonstrating the robustness of the method.

[0053] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A PDC drill bit design method based on machine learning, characterized in that, Includes the following steps: Step S1, Data Acquisition and Fusion; The three-dimensional point cloud parameters of the PDC drill bit before use, the wear level data of the PDC drill bit after use, and the drilling performance parameters of the PDC drill bit are obtained. The multi-source data are cleaned, aligned and normalized to form a standard dataset for model training. Step S2, prediction model construction; Based on machine learning algorithms, using random forest regression or ensemble learning algorithms, with the aforementioned 3D point cloud as input features and drilling performance parameters and wear levels as prediction targets, a high-performance regression and classification model is trained to establish a nonlinear mapping relationship between the 3D geometric features corresponding to the drill bit design and drilling performance parameters and wear levels. Step S3, multi-objective optimization; A multi-objective optimization function is constructed with the goal of minimizing wear level and maximizing drilling performance; a multi-objective evolutionary algorithm is applied to perform a global search within the feasible region of drill bit design parameters to obtain a set of Pareto optimal solutions; Step S4, Decision and Output; Based on actual drilling conditions, the final solution is selected from the Pareto optimal solution set, and the recommended optimal combination of drill bit parameters and its predicted performance and wear indices are output.

2. The PDC drill bit design method based on machine learning according to claim 1, characterized in that: In step S1, the parameters of the PDC drill bit before use include at least the inner cone backslope angle, the nose backslope angle, and the outer cone backslope angle; the wear level of the PDC drill bit after use includes at least the wear level of the inner cone cutting teeth, the wear level of the nose cutting teeth, and the wear level of the outer cone cutting teeth; the drilling performance parameters of the PDC drill bit include at least the total drilling footage and the average mechanical drilling speed.

3. The PDC drill bit design method based on machine learning according to claim 1, characterized in that: In step S1, the initial 3D point cloud is acquired by a non-contact 3D laser scanner, and the 3D point cloud reconstruction error is less than 0.05mm.

4. The PDC drill bit design method based on machine learning according to claim 2, characterized in that: In step S1, the wear level is identified through an image recognition algorithm combined with expert review; drilling performance parameters are automatically extracted and parsed from the daily drilling report or real-time drilling database through a data interface.

5. The PDC drill bit design method based on machine learning according to claim 2, characterized in that: The machine learning algorithm in step S2 is a random forest regression algorithm, gradient boosting tree, or support vector machine; When training high-performance regression and classification models, the criteria for judging high performance include: The high-performance criterion includes the regression model's coefficient of determination R on the test set. 2 The mean square error or average absolute error is greater than or equal to 0.8, and the accuracy of the classification model in the wear level identification task is greater than or equal to 85%, and the macro average F1 value is greater than or equal to the preset threshold.

6. The PDC drill bit design method based on machine learning according to claim 1, characterized in that: The multi-objective optimization algorithm in step S3 adopts a fast non-dominated sorting genetic algorithm with an elitist strategy.

7. The PDC drill bit design method based on machine learning according to claim 1, characterized in that: The optimization function in step S3 introduces weighting coefficients, allowing users to flexibly adjust the optimization bias based on actual formation lithology, drilling cost composition, or efficiency-first strategy.

8. A PDC drill bit parameter design optimization system for implementing the method of any one of claims 1 to 7, characterized in that, include: The 3D scanning data acquisition and processing module is used to automatically acquire and preprocess the drill bit's parameters, wear, and performance data. The intelligent recognition module integrates an image recognition model for automatically assessing the wear level of cutting teeth; The model management module is used to store, train, and retrieve performance prediction models; The optimization and solution module is used to configure optimization objectives and constraints, execute multi-objective optimization algorithms, and output the optimal parameter combination. The human-computer interaction and output module is used to visualize the Pareto frontier, recommended parameters, and prediction results, and to receive user decision-making instructions.

9. The system according to claim 8, characterized in that, The 3D scanning data acquisition and processing module integrates a 3D scanning control unit and a data communication interface.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the method as described in any one of claims 1 to 7.

Images