Intelligent research and development system for cement materials

Abstract

According to the cement material intelligent research and development system provided by the invention, the cement material sample data is stored through the data set, and the experimental formula data for cement production is automatically designed based on the cement material sample data in the data set through the formula design device; and automatically executing a preparation process and a maintenance process of the cement material based on the experimental formula data through automatic experimental equipment, automatically testing the performance of the cement test piece through a performance verification device to obtain performance actual measurement data corresponding to the experimental formula data, and correspondingly storing the performance experimental data and the experimental formula data in a data set. The formula design device is driven to optimize experiment formula design logic based on the updated data set, more accurate and reliable formula data are provided, a closed-loop automatic process from experiment formula prediction to performance evaluation is achieved, the intelligent degree of material research and development is improved, and therefore the research and development efficiency of materials can be improved.

Description

Technical Field

[0001] This application relates to the field of cement research and development technology, and in particular to an intelligent research and development system for cement materials. Background Technology

[0002] Deep and ultra-deep formations have become the main battleground for oil and gas exploration, playing a crucial role in expanding reserves and stabilizing production. However, venturing into deeper areas and drilling wells to depths of 10,000 meters presents immense challenges. In particular, the strength of cementing materials deteriorates significantly at ultra-high temperatures above 200°C, making it difficult to guarantee wellbore integrity, effective cement sheath sealing, and the safety of increased and stable oil and gas production. This is one of the biggest challenges facing deep cementing. Therefore, optimizing the mechanical properties of cement and developing efficient and intelligent ultra-high temperature cement slurry systems are of great importance for supporting cementing in ultra-deep wells.

[0003] Currently, few formulations for ultra-deep, ultra-high temperature cementing for wells above 200℃ fully meet the requirements in terms of both mechanical and engineering properties, especially at 260℃ and above. This challenge is particularly severe, primarily due to the immense difficulty of the research and development process and the anticipated long development cycle. Specifically, the development of ultra-high temperature cement slurry formulations requires navigating various filler types, proportioning schemes, selection of high-temperature additives, and different curing durations (e.g., 7 days, 14 days, 28 days, or even long-term), while also being constrained by safety considerations and quantity limitations of high-temperature curing reactors. For example, while increasing the amount of sand or silica fume can improve the mechanical properties of cement, the origin, gradation (ranging from 200 to 1500 mesh), and proportion of the sand significantly affect the final cement performance. Similarly, adding alumina can enhance the stability of the gel structure at high temperatures, thereby strengthening the cement, but the selection of raw materials such as alumina powder, metakaolin, and fly ash, as well as their amounts, also have a decisive impact on the final performance. In short, numerous factors collectively constitute the crux of the inefficiency of traditional experimental methods in cement slurry development. Therefore, efficiently advancing the research and development of cement slurry and quickly and accurately selecting the optimal formula to meet the stringent requirements of ultra-deep wells and ultra-high temperature environments has become an urgent and crucial task. Summary of the Invention

[0004] This specification provides an intelligent R&D system for cement materials to solve the problem of low efficiency in the R&D of existing cement materials such as cementing cement.

[0005] To address the aforementioned technical problems, this specification provides, in a first aspect, an intelligent R&D system for cement materials. The system includes: a dataset for storing cement material sample data, including sample formulation data for each component of the cement material and corresponding measured performance data of cement specimens; a formulation design device for automatically designing experimental formulation data for cement production based on the cement material sample data and R&D objectives in the dataset; the input data of the formulation design device is the target performance value of the cement material, and the output data is the experimental formulation data; an automated experimental device for automatically executing the cement material preparation and curing processes based on the experimental formulation data to obtain cement specimens corresponding to the experimental formulation data; and a performance verification device for performing performance tests on the cement specimens to obtain measured performance data corresponding to the experimental formulation data, and storing the measured performance data and the experimental formulation data correspondingly in the dataset.

[0006] In some embodiments, the formulation design device includes: a reverse formulation design module, configured to output simulated formulation data corresponding to the input target performance value; a forward performance prediction module, configured to predict simulated performance values ​​corresponding to cement materials based on the simulated formulation data; an adjustment module, configured to repeatedly perform the following operations until the difference between the simulated performance value corresponding to the simulated formulation data and the target performance is less than or equal to a preset threshold: adjusting the design logic of the reverse formulation design module based on the difference between the simulated performance value corresponding to the simulated formulation data and the target performance value, and driving the adjusted reverse formulation design module and the forward performance prediction module to re-execute; and an output module, configured to output the final simulated formulation data and the corresponding simulated performance values ​​after the loop ends.

[0007] In some embodiments, the reverse recipe design module employs a generative adversarial network to implement the design logic.

[0008] In some embodiments, the input data of the positive performance prediction module includes: formulation composition data, environmental process conditions, and multi-scale material properties; the multi-scale material properties include at least one of molecular properties, microscopic properties, and macroscopic properties.

[0009] In some embodiments, the automated experimental equipment includes: a controller for acquiring experimental formula data and mapping the experimental formula data into a sequence of experimental actions of a mechanical device, wherein each experimental action in the sequence constitutes a cement material preparation process and a curing process; and the mechanical device for automatically executing each experimental action to achieve the target experimental action sequence and obtaining cement specimens corresponding to the experimental formula data.

[0010] In some embodiments, the mechanical device includes at least one robotic arm having multiple degrees of freedom, and the end of the robotic arm is provided with grippers adapted to grasp various experimental instruments in the cement material preparation process and curing process.

[0011] In some embodiments, the automated experimental equipment further includes: an arc-shaped experimental platform and various experimental instruments, wherein the experimental instruments are placed at designated positions on the arc-shaped experimental platform according to their usage sequence in the cement material preparation and curing processes; the robotic arm is located in the middle of the arc-shaped experimental platform; in the cement material preparation and curing processes, the robotic arm picks up the required experimental instruments from at least one designated position and performs at least one experimental action in the experimental action sequence, and returns the experimental instruments to their original positions after the execution is completed.

[0012] In some embodiments, the robotic arm is also equipped with a weighing sensor.

[0013] In some embodiments, the automated experimental equipment further includes a human-computer interaction device connected to the controller for user intervention in the automated preparation process of cement specimens.

[0014] In some embodiments, the system further includes: an evaluation device for analyzing the contribution of each component of the cement material to the cement performance based on cement material sample data in the latest dataset, so as to screen the required components for design based on the contribution before the formulation design device automatically designs experimental formulation data for cement production.

[0015] The intelligent cement material R&D system provided in this manual stores cement material sample data through a dataset. A formula design device automatically designs experimental formulas for cement production based on this sample data. Automated experimental equipment automatically executes the cement material preparation and curing processes based on the formula data. A performance verification device automatically tests the performance of cement specimens, obtaining the measured performance data corresponding to the experimental formula data. This performance data and the corresponding formula data are stored in the dataset. This drives the formula design device to optimize the experimental formula design logic based on the updated dataset, providing more accurate and reliable formula data. This achieves a closed-loop automated process from experimental formula prediction to performance evaluation, improving the intelligence level of material R&D and thus increasing its efficiency. Furthermore, the automated experimental equipment automatically executes the preparation process based on the formula data, leveraging the high precision of the automated equipment to achieve small-dose, high-precision preparation, improving the accuracy of the preparation process and further enhancing the accuracy, reliability, and efficiency of the R&D process. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this application. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0017] Figure 1 This is a structural diagram of the intelligent material research and development system provided in this specification. Figure 2 A schematic diagram of a formula design device provided in this specification; Figure 3 This is yet another structural diagram of the intelligent material research and development system provided in this specification. Figure 4 A schematic diagram of the cement slurry preparation process; Figure 5 This is a schematic diagram of a robotic arm transitioning from one posture to another. Figure 6 This is a diagram illustrating the tilting angle; Figure 7 This is a schematic diagram showing the connection between the human-computer interaction device and the robotic arm. Detailed Implementation

[0018] To enable those skilled in the art to better understand the technical solutions in this application, the technical solutions in the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of this application.

[0019] This specification provides an intelligent research and development system for cement materials. Specifically, the cement material can be used for cementing operations in oil and gas exploration drilling.

[0020] like Figure 1 As shown, the system includes a data set A, a formula design device B, automated experimental equipment C, and a performance verification device D.

[0021] Data set A is used to store cement material sample data. The cement material sample data includes sample formula data for each component of the cement material and the corresponding measured performance data of the cement specimens.

[0022] Cement materials are formula-driven materials. Through precise proportioning of multiple components and process control, the components can produce synergistic or complementary effects, ultimately obtaining cement specimens with preset properties.

[0023] The initial sample data in dataset A can be obtained through manual experiments; or it can be obtained from publicly available documents by proportionally scaling up or down the dosage of each component in the publicly available formula to the experimental level, and by processing the dosage of each component according to the characteristics of the performance parameters (for example, the strength value of cement does not improve with the dosage, so the strength value of cement does not need to be changed).

[0024] Formula design device B is used to automatically design experimental formulas for cement production based on cement material sample data and research and development goals in dataset A.

[0025] In some embodiments, the input data of the formulation design device B may be the components of cement material and the range of values ​​for the dosage of each component, and the output data may be experimental formulation data for cement production. Accordingly, the formulation design device B can generate the designed experimental formulation data by traversing the dosage values ​​of each component at predetermined intervals within the range of dosage values ​​for each component.

[0026] This formulation design method lacks specific targeting, typically resulting in a large volume of experimental formulation data. This places significant pressure on automated testing equipment to automatically execute the cement material preparation and curing processes. To address this, before the automated testing equipment executes each experimental formulation, a positive performance prediction module (such as the one described below) can be used to predict the performance data of the cement specimens corresponding to each formulation, eliminating formulation data whose predicted performance does not meet expectations. This places high demands on the accuracy of the performance prediction module; if the module is inaccurate, it may discard formulation data corresponding to better performance data.

[0027] In other embodiments, the input data of the formula design device B can be the target performance value of the cement material, and the output data can be experimental formula data. That is to say, the above-mentioned research and development target can be the target performance value of the cement material.

[0028] It is possible to design experimental formulation data for cement production based on the target performance values ​​of cement materials by training a highly accurate design model. This places high demands on the accuracy and reliability of the design model.

[0029] Furthermore, the formula design device B may include a reverse formula design module, a forward performance prediction module, an adjustment module, and an output module.

[0030] To further improve the accuracy and reliability of the designed formulation data, such as Figure 2As shown, the formula design device may include a reverse formula design module B1, a forward performance prediction module B2, an adjustment module B3, and an output module B4.

[0031] The reverse recipe design module B1 is used to output the proposed recipe data corresponding to the input target performance value. The proposed recipe data is temporary recipe data generated by the reverse recipe design module B1 during the design process.

[0032] The target performance value input to the reverse formulation design module B1 can include compressive strength.

[0033] The positive performance prediction module B2 is used to predict the simulated performance values ​​of cement materials based on the simulated formula data.

[0034] The input data for the positive performance prediction module B2 may include: formulation composition data (which describes which components are used and their proportions, including water-cement ratio, filler ratio, and types of additives), environmental process conditions (e.g., temperature, stirring time, curing time), and multi-scale material properties. Multi-scale material properties include at least one of molecular properties, microscopic properties, and macroscopic properties.

[0035] The output data of the positive performance prediction module B2 may include compressive strength.

[0036] The adjustment module B3 is used to repeatedly perform the following operations until the difference between the simulated performance value and the target performance value corresponding to the simulated formula data is less than or equal to a preset threshold: adjust the design logic of the reverse formula design module according to the difference between the simulated performance value and the target performance value corresponding to the simulated formula data, and drive the adjusted reverse formula design module and the forward performance prediction module to be re-executed.

[0037] In some embodiments, the adjustment module B3 can determine the direction of parameter adjustment based on the changing trend of the pseudo-performance value output by the forward performance prediction module B2, and determine the adjustment amount based on the difference between the pseudo-performance value and the target performance value. For example, the adjustment module B3 will first adjust a parameter in the reverse formulation design module B1 in an increasing direction. △ x1, the adjusted reverse formulation design module B1 outputs pseudo-formulation data M1, and the forward performance prediction module B2 predicts the pseudo-performance value corresponding to formulation data M1 as N1; then, this parameter in the reverse formulation design module B1 is first adjusted in the direction of increasing. △x2. The adjusted reverse formulation design module B1 outputs pseudo-formulation data M2, and the forward performance prediction module B2 predicts the pseudo-performance value corresponding to formulation data M2 as N2. If N2 is closer to the target performance value than N1, the next adjustment module B3 can adjust this parameter in the reverse formulation design module B1 in the direction of increasing; if N2 is farther from the target performance value than N1, the parameter should be adjusted in the direction of decreasing. The smaller the difference between the pseudo-performance value and the target performance value, the smaller the adjustment amount when adjusting this parameter in the reverse formulation design module B1 in the next step; the larger the difference between the pseudo-performance value and the target performance value, the larger the adjustment amount of this parameter in the next step.

[0038] In other embodiments, a loss function can be constructed based on the difference between the simulated performance value and the target performance value, and the parameters in the positive performance prediction module B2 can be adjusted with the goal of minimizing the value of the loss function.

[0039] It's important to note that the loss function here is the one used when designing experimental recipe data, and it differs from the loss function used during model training. When the model is used for designing experimental recipe data, the models in the reverse recipe design module and the forward performance prediction module have already been trained.

[0040] Output module B4 is used to output the final simulated formula data and corresponding simulated performance values ​​after the loop ends.

[0041] The reverse recipe design module B1 can implement the design logic using encoders and decoders, reinforcement learning models, neural network models, physical information neural networks, etc.

[0042] In some embodiments, the reverse recipe design module B1 may employ a generative adversarial network to implement the design logic. This generative adversarial network is periodically updated and trained based on sample data from the latest dataset.

[0043] Generative Adversarial Networks (GANs) consist of a generator and a discriminator. The generator produces simulated formula data based on the target performance values ​​of cement materials, while the discriminator determines whether the simulated formula data generated by the generator can achieve the target performance values. Training a GAN is a minimax game process, with the goal of achieving a Nash equilibrium between the generator and the discriminator. This equilibrium occurs when the generator's formula data is so good that the discriminator cannot distinguish whether it achieves the target performance values. At this point, the generator's output represents high-quality simulated formula data that achieves the target performance values.

[0044] The training process of a Generative Adversarial Network (GAN) is as follows: S1. Fix the generator and train the discriminator. Train the discriminator using recipe data that can achieve the target performance value and recipe data that cannot, optimizing its classification ability (i.e., determining whether a sample recipe can achieve the target performance value) to maximize its accuracy in distinguishing whether a sample recipe can achieve the target performance value. S2. Fix the discriminator and train the generator. Adjust the generator parameters through backpropagation so that the generated sample recipe data elicits a higher probability value from the discriminator (this probability refers to the probability that the sample recipe can achieve the target performance value), minimizing the discriminator's discriminative power. Repeat S1 and S2 until the abilities of the generator and discriminator reach a balance, at which point the generator can stably output high-quality sample recipe data.

[0045] The development of ultra-high temperature cementing is a key technological bottleneck in deep well oil and gas exploration. Its core challenges lie in selecting feasible fillers from a vast array of materials, data scarcity, weak extrapolation capabilities of traditional machine learning, low development efficiency, and the difficulty of adapting existing systems to the demands of refined laboratory formulation development. By employing generative adversarial networks to implement the design logic of the reverse formulation design module B1, and integrating it into the cementing slurry development system, a breakthrough innovation in the development model has been achieved. The core technological effects are as follows: 1. Overcoming the bottleneck of small sample data and reducing reliance on R&D data. The high cost and stringent testing conditions of ultra-high temperature cement preparation lead to a scarcity of real samples, making it difficult to implement traditional deep learning models. Generative adversarial networks (GANs), through an adversarial training mechanism between the generator and discriminator, can learn the inherent patterns of data from a limited set of 50-200 real samples. Simultaneously, they generate synthetic data that conforms to physical characteristics to expand the training set. Combined with transfer learning, knowledge from the ordinary cement industry can be reused, significantly reducing the need for ultra-high temperature-specific data and solving the core challenge of small sample learning in this field.

[0046] 2. Precisely Achieve Two-Way Mapping Between Performance and Formulation, Enhancing Design Accuracy. Addressing the complex nonlinear relationship between "formulation-microstructure-macroscopic performance" in cement at ultra-high temperatures, an adversarial network is generated to achieve end-to-end inverse mapping between "target performance value and formulation data," eliminating the need for manual feature decomposition. By using target performance such as compressive strength as input constraints, formulations meeting performance standards can be directly generated. Combined with verification feedback from the positive performance prediction module, this further improves formulation design accuracy and avoids performance deviation problems associated with traditional trial-and-error methods.

[0047] 3. Generate diverse feasible formulations and expand the space for innovative research and development. Under the same performance target, the generative adversarial network can output a variety of differentiated feasible formulations (such as different additive combinations and proportioning schemes), ensuring that the formulations comply with the principles of materials science (such as water-cement ratio range and additive compatibility), while providing engineers with multi-dimensional choices for cost optimization and process adaptation. At the same time, it can explore innovative formulations that go beyond the existing data distribution, break through the limitations of traditional experience, and contribute to the innovation of ultra-high temperature cementing materials.

[0048] 4. Construct a closed-loop optimization system to improve R&D efficiency and cost-effectiveness. Formulas generated by generative adversarial networks can be quickly verified through the system's built-in positive performance prediction module. Error feedback signals optimize the model in reverse, forming an intelligent closed loop of "generation-verification-iteration." This model compresses the traditional "months of R&D cycle and hundreds of experiments" to a few days, generating tens of thousands of samples in just tens of seconds. It significantly reduces the number of high-temperature laboratory experiments, lowering R&D costs to 1 / 10 of existing methods, and adapts to the refined R&D needs of small-dose, multi-stage processes in laboratories.

[0049] In summary, the synergistic integration of generative adversarial network technology with precise robot control and full-process monitoring can effectively overcome the bottlenecks of low R&D efficiency and difficulty in performance optimization of ultra-high temperature cement, promote the transformation of the cementing materials field from "experience-based trial and error" to "intelligent and precise", and provide key technical support for deep well oil and gas exploration.

[0050] To address the small sample size issue encountered in scenarios such as the development of ultra-high temperature cementing, the forward performance prediction module B2 employs a neural network model incorporating an attention mechanism. This self-attention mechanism automatically identifies key influencing factors in the input formula data (such as water-cement ratio and additive types in the formula parameters, and curing temperature in the process conditions) and assigns differentiated weights to different factors. Specifically, features with a significant impact on cement compressive strength (such as water-cement ratio) are given higher weights, while features with a smaller impact are given lower weights. This setting allows the model to focus on core variables and avoid interference from irrelevant features, thereby improving the accuracy of performance value predictions.

[0051] To address the small sample size issue encountered in scenarios such as the development of ultra-high temperature cementing, data augmentation, transfer learning strategies, regularization constraints, and ensemble learning methods can be employed. Data augmentation methods can reasonably transform the existing limited formula data (e.g., fine-tuning component ratios) to expand the effective training sample size and alleviate the data shortage problem. Transfer learning strategies can leverage mature model knowledge from other fields (e.g., ordinary concrete, similar cementitious materials) and apply it to the performance prediction task of cementing, reducing the dependence on sample data in scenarios such as the development of ultra-high temperature cementing. Constraining the parameters of neural networks (e.g., L1 / L2 regularization) limits model complexity and prevents overfitting to the limited training samples. Ensemble learning methods construct sub-prediction models with various structures and / or slightly different parameters. By fusing (e.g., weighted fusion) the prediction results of multiple sub-models, the advantages of each sub-model are combined, reducing the prediction bias of a single model, improving the reliability of prediction results, and enhancing overall stability.

[0052] To address the small sample size issue encountered in scenarios such as the development of ultra-high temperature cementing, a cross-validation mechanism can be employed in model training. This involves dividing the limited sample into training and validation sets, and then conducting multiple rounds of training and validation to examine the model's performance on different data subsets. This avoids the overfitting problem where the model performs well on the training set but fails to predict new data.

[0053] The positive performance prediction module B2 can adopt a composite small sample retrospective prediction model with attention mechanism neural network as the core, integrating data augmentation, transfer learning, regularization, ensemble learning and other technologies, and using multi-model fusion and cross-validation to ensure the performance prediction results, so as to adapt to the prediction accuracy requirements in scenarios such as the R&D process of ultra-high temperature cementing.

[0054] The automated experimental equipment C is used to automatically execute the preparation and curing processes of cement materials based on experimental formula data, so as to obtain cement specimens corresponding to the experimental formula data.

[0055] like Figure 3 As shown, the automated experimental equipment C includes a controller C1 and a mechanical device C2.

[0056] Controller C1 is used to acquire experimental formula data and map the experimental formula data into a sequence of experimental actions of mechanical device C2. The various experimental actions in the sequence constitute the preparation and curing process of cement materials.

[0057] The experimental sequence of actions includes the cement slurry preparation process. For example... Figure 4As shown, the overall configuration process includes: uniform stirring, solid-liquid mixing, pouring into the mold, curing, and washing the slurry cup. The experimental action sequence may include: 1. The robotic arm grabs the stirring rod and uniformly stirs the powder in the solid cup, then accurately returns it to its original position; 2. Grabs the slurry cup containing the liquid additive and stably places it in the stirrer's clamping position; 3. Controls the gripper to trigger the stirrer's power switch and starts stirring; 4. Grabs the solid cup and pours the solid material into the slurry cup at a constant speed and height to achieve uniform solid-liquid mixing; 5. After mixing, operates the gripper again to turn off the stirrer; 6. Transfers the slurry cup to the top of the mold and pours the slurry, ensuring that the slurry completely fills the mold without overflowing; 7. Transfers the filled mold to the curing unit; 8. Grabs the used slurry cup and tilts it to fix it to the special hook; 9. Takes the cleaning brush and rotates to scrub the inner wall of the slurry cup to ensure no residue remains; 10. After cleaning, returns all tools to their original positions, and the process ends.

[0058] The experimental formula data can include the dosage of various fillers, stirring parameters, mold slurry capacity, and the timing of the process flow. Controller C1 can map the experimental formula data into an experimental action in the sequence of experimental actions performed by the robotic arm: filling the mold with slurry. The specific parameters of this experimental action are: a 50.8mm mold, the robotic arm tilting angle, speed, and time.

[0059] Mechanical device C2 can be composed of multiple mechanisms that perform different functions, or it can be a robotic arm, or it can be composed of mechanisms that perform different functions and a robotic arm.

[0060] In some embodiments, the mechanical device C2 includes at least one robotic arm. The robotic arm has multiple degrees of freedom, and its end is provided with grippers suitable for grasping various experimental instruments.

[0061] In one embodiment, such as Figure 5 As shown, the robotic arm is a six-axis robotic arm with a working radius of 900mm. Its joint range of motion covers J1±360°, J2±360°, J3±160°, J4±360°, J5±360°, and J6±360°, with a repeatability of ±0.02mm, far exceeding the accuracy of manual operation. It can precisely cover most of the operating area on the experimental platform, accurately control various operations during the slurry preparation process, and ensure the stability and repeatability of the cement slurry performance. The rated voltage is DC 48V, the full-load current is 6.3A, the maximum working speed is 3m / s, and the maximum speed of each joint reaches 180° / s, enabling it to flexibly and efficiently complete various experimental operations.

[0062] The robotic arm's end effector is equipped with a multi-functional electric gripper that supports complex actions such as grasping, tilting, and rotating. It can utilize industrial grippers that combine performance and adaptability, offering a gripping force of 15-50N, a finger opening / closing stroke of 0-35mm, and flexible programmable adjustment. Combined with a high positional repeatability of ±0.03mm and a rapid opening / closing response of 0.7s, operating noise <50dB, and an IP54 protection rating, it can operate stably in precision machining and parts grasping scenarios. It is compatible with various shapes of experimental instruments, including slurry cups, solid cups, stirring rods, and molds.

[0063] Mechanical installation requires only locating pins and screws for fixing, and electrical connections can be completed via adapter cables, ensuring a rigid connection and synchronized movement between the gripper and the end effector of the robotic arm. Subsequent control can be achieved via PC or mobile terminal, while also supporting programmable status reading and indicator light visualization (e.g., a solid blue light upon initialization completion, a solid green light upon gripping an object), balancing professionalism and convenience.

[0064] In terms of operation, the mechanical installation of the robotic arm only requires locating pins and screws for fixation. The design incorporates a tethered cable to connect the robotic arm's end effector I / O interface to the electric gripper, ensuring a secure connection without any looseness. The cable length is designed with a 10%-15% redundancy allowance based on the maximum range of motion of the robotic arm's end effector to prevent stretching. It can move synchronously with the end effector and will not become tangled due to joint rotation (such as 360° rotation of the J4, J5, and J6 axes). Furthermore, the tethered cable integrates signal and power transmission functions, eliminating the need for additional cables and further simplifying the wiring layout. Compared to conventional long cables, it effectively solves the problem of self-tangling and messy wiring after looping once during operation.

[0065] When the automated experimental equipment C is composed of multiple mechanisms that perform different functions, the mechanisms may include feeding components and experimental apparatus, thus eliminating the need for dedicated experimental tables and apparatus.

[0066] In some embodiments, the automated experimental equipment C includes an arc-shaped experimental platform and experimental instruments, which are placed at designated positions on the arc-shaped experimental platform according to their usage sequence in the cement material preparation and curing processes.

[0067] The robotic arm is positioned in the center of the arc-shaped experimental platform. During the cement material preparation and curing process, the robotic arm retrieves the required experimental equipment from at least one designated location and performs at least one experimental action in the experimental action sequence, and then returns the experimental equipment to its original position after completion.

[0068] The arc length of the circular experimental platform can be further increased to form a ring-shaped experimental platform. The robotic arm is set at the center of the ring-shaped experimental platform, and the experimental instruments are arranged radially on the ring-shaped experimental platform relative to the position of the robotic arm. The experimental instruments are arranged along the circumference of the ring according to the order of use in the cement material preparation and curing process.

[0069] By using a circular experimental platform, radially distributed experimental equipment, and equipment arranged according to the order of use, the spatial positions of solid cups, stirrers, mold placement areas, curing devices, and cleaning devices can be planned more rationally. Each functional area is evenly distributed around the robotic arm within its working radius, minimizing the robotic arm's idle travel and motion interference, and improving experimental operation efficiency.

[0070] In some embodiments, the robotic arm may be equipped with sensors to identify the type, position, and quantity of the implements, as well as the type, placement, and quantity of the component materials. The sensors may be infrared sensors, ultrasonic sensors, vision sensors, etc.

[0071] In some embodiments, a weighing sensor may also be provided on the robotic arm to determine the weight of the component material being collected.

[0072] The automated experimental equipment C may also include a cement curing device for curing cement specimens.

[0073] When using a robotic arm to pour slurry into a mold, overflow or insufficient filling is common. To address this, a quantitative control strategy based on angle calibration can be employed: the optimal pouring angle is determined through multiple trials, and the robotic arm maintains this angle during each pour, thus ensuring consistent slurry volume. This method eliminates the need for additional sensors, reducing system complexity and cost while maintaining accuracy. Figure 6 This is a schematic diagram of the tilting angle, where the angle between dashed lines a and b is the tilting angle. Dashed line a is a straight line extending from the bottom of the appliance toward the opening of the appliance, and dashed line b is a vertical line perpendicular to the horizontal plane.

[0074] Furthermore, the robotic arm employs adjustable speed and trajectory motion control during stirring, pouring, and cleaning processes to avoid splashing and operational impact, further improving the repeatability of operations and sample quality. Taking "pouring" as an example, it can be divided into three stages: approaching the mold, stabilizing the pour, and leaving the mold, with different speed and trajectory combinations used in each stage. The approaching-mold stage uses a "linear motion" mode, moving horizontally above the mold to avoid collision between the slurry cup and the mold; the stabilizing-pouring stage calls the "joint motion" function module, moving slowly to the pouring point at an appropriate joint speed, and the software's "smooth transition" function eliminates joint movement impact during the movement to prevent slurry from overflowing due to vibration; the leaving-mold stage uses a linear motion mode, rising vertically to a safe height before moving horizontally to the next station to prevent residual slurry from dripping outside the mold upon leaving.

[0075] In some embodiments, the automated experimental equipment also includes a human-machine interface device connected to the controller for user intervention in the automated preparation process of cement specimens. For example, the user can verify or modify formula data, and control the start or stop of the automated preparation process. Figure 7 As shown, the human-computer interaction device can be a mobile phone or a computer, which can communicate with the robotic arm via wireless communication technology (such as WiFi).

[0076] The operator needs to accurately set the following basic parameters of the robotic arm on the interactive interface: (1) For the six joints (J1-J6) of the robotic arm, set the speed and acceleration ratio range parameters (1~100) of each joint respectively. The operator needs to set the differentiated speed according to the experimental accuracy requirements. The precision dispensing stage is set to 20-50° / second to ensure stability, and the rapid transfer stage can be increased to 100-150° / second to improve efficiency. (2) The precise coordinates of each functional area need to be marked on the interface, including the solid material cup placement point, liquid additive position, stirrer position, mold array coordinates and cleaning device position. The system records key points through the teaching function to ensure the accuracy of subsequent automated execution. The safety parameter configuration requires setting the collision detection force threshold, emergency stop response time (less than 0.5 seconds) and workspace safety boundary to prevent the robotic arm from exceeding the 900mm working radius.

[0077] The experimental action sequence completed by the robotic arm can be constructed through a graphical programming platform of the control system (which includes the aforementioned controller). This platform, combined with the robotic arm's motion capabilities and precise coordinate positioning, provides modular motion commands and logic control units to create the experimental action sequence, enabling the grasping of objects at specific locations within a fixed area. The object needs to be pre-programmed and placed at predetermined coordinate points. The robotic arm moves to these points according to the program instructions, precisely performing operations such as mixing raw materials, pouring and curing, and cleaning equipment. Simultaneously, safety monitoring integrates collision detection and emergency stop mechanisms.

[0078] The aforementioned controller can be based on a graphical programming platform, providing modular motion commands ("joint motion", "linear motion", "circular motion") and logic control units ("conditional judgment", "loop", "delay"), supporting joint, user, and tool coordinate systems, and using algorithms to generate collision-free trajectories to complete transfers between different areas. The control system includes human-machine interaction configuration, AI model command conversion, and execution control flow.

[0079] The performance verification device D is used to perform performance tests on the cement specimens, obtain the actual performance data corresponding to the experimental formula data, and store the actual performance data and the experimental formula data in the data set A.

[0080] The performance verification device D may include an operating mechanism and a testing mechanism. The operating mechanism is used to perform performance testing operations, and the testing mechanism is used to test parameters used for performance evaluation during the performance testing operations.

[0081] The performance verification device D can verify and evaluate the accuracy and reliability of the formulation performance formed by the formulation design device and automated experimental equipment based on the performance test results.

[0082] In some embodiments, such as Figure 3 As shown, the intelligent R&D system for cement materials also includes: an evaluation device E, used to analyze the contribution of each component of cement materials to cement performance based on the latest cement material sample data in the latest dataset A, so as to screen the required components based on the contribution before the formula design device B automatically designs experimental formula data for cement production.

[0083] The evaluation device E has a built-in model interpretation module. After training, this module calculates the SHAP values ​​for each input parameter dimension using the SHAP analytical model, based on the trained target model and the corresponding training dataset. It then determines the contribution of each input parameter dimension to the model output based on the SHAP values, thus achieving interpretability of the model prediction results. The target model is the model corresponding to the reverse recipe design module or the forward performance prediction module.

[0084] Furthermore, a multidimensional feature dependency graph can be constructed based on the SHAP analytical model to conduct feature importance analysis and accurately quantify the impact of key factors on cement strength. This not only provides interpretability support for the model's prediction results but also verifies whether the formula-performance correlation learned by the model conforms to the basic laws of materials science. Simultaneously, by identifying input variables that play a crucial role in cement performance, a reference basis for weight adjustment can be provided for the next round of optimization work in the formula design module.

[0085] The intelligent cement material R&D system provided in this manual stores cement material sample data through a dataset. A formula design device automatically designs experimental formulas for cement production based on this sample data. Automated experimental equipment automatically executes the cement material preparation and curing processes based on the formula data. A performance verification device automatically tests the performance of cement specimens, obtaining the measured performance data corresponding to the experimental formula data. This performance data and the corresponding formula data are stored in the dataset. This drives the formula design device to optimize the experimental formula design logic based on the updated dataset, providing more accurate and reliable formula data. This achieves a closed-loop automated process from experimental formula prediction to performance evaluation, improving the intelligence level of material R&D and thus increasing its efficiency. Furthermore, the automated experimental equipment automatically executes the preparation process based on the formula data, leveraging the high precision of the automated equipment to achieve small-dose, high-precision preparation, improving the accuracy of the preparation process and further enhancing the accuracy, reliability, and efficiency of the R&D process.

[0086] The various embodiments in this specification are described in a progressive manner. For the same or similar parts between the various embodiments, please refer to each other. The focus of each embodiment is to describe the differences from other embodiments.

[0087] The systems, devices, modules, or units described in the above embodiments can be implemented by computer chips or entities, or by products with certain functions.

[0088] For ease of description, the above devices are described by dividing them into various functional units. Of course, in implementing this application, the functions of each unit can be implemented in one or more software and / or hardware.

[0089] As can be seen from the above description of the embodiments, those skilled in the art can clearly understand that this application can be implemented by means of software plus necessary general-purpose hardware platforms. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute certain parts of the methods of various embodiments of this application.

[0090] This application can be used in a wide variety of general-purpose or special-purpose computer system environments or configurations. For example: personal computers, server computers, handheld or portable devices, tablet devices, multiprocessor systems, microprocessor-based systems, set-top boxes, programmable consumer electronics devices, network PCs, minicomputers, mainframe computers, distributed computing environments including any of the above systems or devices, etc.

[0091] This application can be described in the general context of computer-executable instructions, such as program modules, that are executed by a computer. Generally, program modules include routines, programs, objects, components, data structures, etc., that perform a specific task or implement a specific abstract data type. This application can also be practiced in distributed computing environments where tasks are performed by remote processing devices connected via a communication network. In distributed computing environments, program modules can reside in local and remote computer storage media, including storage devices.

[0092] Although this application has been described through embodiments, those skilled in the art will know that this application has many modifications and variations without departing from the spirit of this application, and it is intended that the appended claims cover such modifications and variations without departing from the spirit of this application.

Claims

1. An intelligent R&D system for cement materials, characterized in that, The system includes: A dataset for storing cement material sample data, which includes sample formula data of each component of the cement material and the actual performance test data of the corresponding cement specimens; A formulation design device is used to automatically design experimental formulation data for cement production based on cement material sample data and research and development goals in the dataset; the input data of the formulation design device is the target performance value of the cement material, and the output data is the experimental formulation data. An automated experimental device is used to automatically execute the preparation and curing processes of cement materials based on the experimental formula data, so as to obtain cement specimens corresponding to the experimental formula data. A performance verification device is used to perform performance tests on the cement specimens, obtain the measured performance data corresponding to the experimental formula data, and store the measured performance data and the experimental formula data in the data set.

2. The system according to claim 1, characterized in that, The formula design device includes: The reverse recipe design module is used to output the proposed recipe data corresponding to the input target performance value based on the input target performance value. A positive performance prediction module is used to predict the corresponding simulated performance values ​​of cement materials based on the simulated formula data; The adjustment module is used to repeatedly perform the following operations until the difference between the simulated performance value corresponding to the simulated formula data and the target performance is less than or equal to a preset threshold: adjust the design logic of the reverse formula design module according to the difference between the simulated performance value corresponding to the simulated formula data and the target performance value, and drive the adjusted reverse formula design module and the forward performance prediction module to be re-executed; The output module is used to output the final simulated formula data and corresponding simulated performance values ​​after the loop ends.

3. The system according to claim 2, characterized in that, The reverse recipe design module uses a generative adversarial network to implement the design logic.

4. The system according to claim 2, characterized in that, The input data for the positive performance prediction module includes: formulation composition data, environmental process conditions, and multi-scale material properties; the multi-scale material properties include at least one of molecular properties, microscopic properties, and macroscopic properties.

5. The system according to claim 1, characterized in that, The automated experimental equipment includes: The controller is used to acquire experimental formula data and map the experimental formula data into a sequence of experimental actions of the mechanical device. The experimental actions in the sequence of experimental actions constitute the preparation process and curing process of cement material in sequence. The mechanical device is used to automatically execute each experimental action to achieve the target experimental action sequence and obtain the cement specimen corresponding to the experimental formula data.

6. The system according to claim 5, characterized in that, The mechanical device includes at least one robotic arm with multiple degrees of freedom, and the end of the robotic arm is provided with grippers suitable for gripping various experimental instruments in the cement material preparation and curing processes.

7. The system according to claim 6, characterized in that, The automated experimental equipment also includes: The circular arc-shaped experimental platform and various experimental instruments are arranged in designated positions on the circular arc-shaped experimental platform according to their usage order in the cement material preparation and curing processes. The robotic arm is located in the middle of the arc-shaped experimental platform. In the cement material preparation and curing process, the robotic arm picks up the required experimental equipment from at least one designated position and performs at least one experimental action in the experimental action sequence, and puts the experimental equipment back to its original position after the execution is completed.

8. The system according to claim 6, characterized in that, The robotic arm is also equipped with a weighing sensor.

9. The system according to claim 5, characterized in that, The automated experimental equipment also includes: A human-computer interaction device, connected to the controller, is used for user intervention in the automated preparation process of cement specimens.

10. The system according to claim 1, characterized in that, The system also includes: An evaluation device is used to analyze the contribution of each component of cement material to cement performance based on cement material sample data in the latest dataset, so as to screen the required components based on the contribution before the formulation design device automatically designs experimental formulation data for cement production.

Images