A control method and system for an intelligent screwdriver

Abstract

The present invention discloses a control method and system for an intelligent screwdriver, and relates to the field of intelligent control technology. A control system for an intelligent screwdriver includes: a setting initialization module, a real-time data acquisition module, a resistance prediction and material identification module, a dynamic adjustment and execution module, and a feedback and data recording module. The present invention can dynamically adjust the working parameters of the screwdriver according to the real-time collected data through the combination of the resistance prediction model and the material identification model, avoiding the traditional adjustment method that relies on manual experience; through the application of the material identification model, the density, hardness, surface texture and other characteristics of the target material are accurately identified, and the working parameters are adjusted accordingly, so that the screwdriver can adapt to a variety of materials including metal, wood, composite materials, etc., thereby significantly improving the versatility and adaptability of the equipment.

Description

Technical Field

[0001] The present invention relates to the field of intelligent control technology, and in particular to a control method and system for an intelligent screwdriver. Background Art

[0002] As an important equipment in industrial assembly, construction engineering and household tools, the main function of the screwdriver is to drive the drill bit through a power device to achieve the tightening and removal of screws. Traditional screwdrivers are mostly mechanical or simple electronic control, and the operating parameters rely on manual settings. They cannot automatically adjust according to the working environment and target materials, resulting in significant limitations on operating efficiency and accuracy.

[0003] At present, the screwdrivers on the market usually have the following problems and limitations: The adjustment of operating parameters depends on manual experience. Due to the different experience levels of different users, this adjustment method is prone to inaccurate parameter settings, resulting in insufficient tightening, excessive tightening or damage to the material. Insufficient material recognition capabilities. Existing screwdrivers cannot effectively identify the type and characteristics of the target material, and it is difficult to dynamically adjust the working parameters according to the needs of different materials. Insufficient real-time feedback and dynamic adjustment capabilities. Traditional screwdrivers lack real-time data collection and feedback mechanisms and cannot dynamically adjust parameters during operation. The level of equipment intelligence is limited.

[0004] The current screwdrivers are still in the early stages of intelligent control, lacking efficient intelligent algorithm support and unable to meet complex and changing task requirements. At the same time, due to the complexity of the calculation model, existing equipment is difficult to operate in a weak computing environment, further limiting its application scenarios. In summary, traditional screwdrivers have significant deficiencies in parameter adjustment, material adaptability, real-time feedback, and intelligence levels, which have an adverse impact on operational efficiency, task accuracy, and equipment life. Summary of the invention

[0005] In response to the above problems, the present invention proposes a control method and system that can collect data in real time, dynamically adjust working parameters and have intelligent feedback capabilities to adapt to the needs of different materials and task scenarios, while improving the reliability and operational efficiency of the equipment.

[0006] A control method for an intelligent screwdriver comprises the following steps:

[0007] Step S1, setting initialization: according to the task parameters input by the user, matching initial working parameters are loaded from the preset material database and screw database, and the state of the drill bit is detected to determine whether it meets the current task requirements. If not, the user is prompted to change the recommended drill bit model; the task parameters include screw specifications and target material types; the state of the drill bit is detected by the rotation smoothness, friction and temperature data of the drill bit;

[0008] Step S2, real-time data collection: real-time data during the operation is collected through the built-in torque sensor, feedback current sensor, material property sensor and pressure sensor of the screwdriver; the real-time data includes torque, feedback current, physical properties of the target material and applied pressure data;

[0009] Step S3, resistance prediction and material identification: Based on the real-time collected torque and feedback current data, the resistance change trend is predicted using the resistance prediction model; based on the density, hardness and surface texture data obtained by the material property sensor, the material type and combination structure of the target operation object are judged through the material identification model, and the material type specifically includes wood, metal, composite material and combination structure type;

[0010] Step S4, dynamic regulation and execution: dynamically adjust the working parameters of the screwdriver according to the resistance prediction and material identification results; the adjustment of the working parameters includes adjusting the torque output, increasing or decreasing the torque in advance according to the resistance change; when the resistance exceeds the threshold or the target material is a hard material, enabling the impact mode, and dynamically adjusting the impact frequency and intensity; based on the material identification results and the resistance situation, regulating the pressure applied by the linear drive device, increasing the pressure for hard materials and reducing the pressure for soft materials;

[0011] Step S5, feedback and data recording: monitor the drill bit status during the operation, and prompt the user to replace it if the drill bit is worn or mismatched; record the real-time data of the tightening operation process and store it locally or in the cloud for task tracing and model optimization.

[0012] As a preferred technical solution of the present invention, the acquisition of the matching initial working parameters includes:

[0013] According to the screw specifications input by the user, the corresponding screw characteristic data is obtained from the screw database, including the screw size, material and recommended tightening torque range; according to the target material type input by the user, the physical property data of the target material is obtained from the material database, including density, hardness and surface texture parameters;

[0014] According to the screw characteristic data and the physical characteristic data of the target material, the initial working parameters are matched in combination with the process requirements corresponding to the task parameters. The initial working parameters include the initial torque, the trigger threshold of the impact mode, the application range of the linear pressure and the adapted rotation speed. When the user does not input the task parameters, the default initial working parameters are recommended based on the historical operation data and the task type.

[0015] As a preferred technical solution of the present invention, the detection of the drill bit status includes:

[0016] Parameter calculation:

[0017] By monitoring the fluctuation of the drill bit's rotation speed, the deviation of its smoothness is calculated , the formula is as follows:

[0018] ,in is the rotation speed of the i-th sampling, is the average value of the rotation speed, is the sampling times, the minimum sampling times is 10;

[0019] By measuring the friction between the drill bit and the material , to judge the degree of drill bit wear, the friction force is calculated as follows: ,in is the real-time measured torque, is the radius of the drill bit, is the friction coefficient of the contact surface, is the applied axial pressure, is the contact area;

[0020] Real-time monitoring of drill bit temperature via built-in temperature sensor ;

[0021] Evaluation parameter quantization:

[0022] Quantify the degree of deviation from the smoothness of the drill bit rotation as a condition score , the calculation formula is: ;in is the maximum allowable deviation of the rotational smoothness setting; if If it is less than the set threshold, it is judged that there is an abnormality in the rotation smoothness;

[0023] Convert friction test results into condition scores , the calculation formula is: ,in is the corresponding theoretical friction force in the target material database; if If it is less than the set threshold, it is judged that the drill bit is severely worn;

[0024] Convert drill bit temperature into a condition score , the calculation formula is: ,in is the maximum operating temperature; if If it is less than the set threshold, the drill is judged to be overheated;

[0025] Comprehensive status score calculation:

[0026] Calculates a comprehensive condition score by combining rotational smoothness, friction and temperature , the formula is: ,in They are the weight coefficients of the corresponding parameters, which need to be adjusted according to the specific scenario, and the sum of the three weight coefficients is 1.2;

[0027] When the comprehensive status score is lower than the set threshold, the drill bit status is judged to be abnormal; and each parameter is evaluated independently to further analyze the cause of the abnormality.

[0028] As a preferred technical solution of the present invention, the resistance prediction model includes the following contents:

[0029] Model structure: The resistance prediction model adopts a lightweight recursive neural network structure, which specifically includes: an input layer for receiving the time series of real-time collected torque data and feedback current data; a hidden layer, which is composed of a group of gated recurrent units, for reducing the number of parameters and computational complexity and retaining the long-term dependencies in the time series; an output layer, which outputs the predicted resistance value at the next moment; the lightweight design of the model is adapted to the weak computing power environment;

[0030] Data preprocessing, real-time acquisition of torque and feedback current data Perform standard normalization processing, the specific formula is:

[0031] ,in and is the corresponding historical mean, and is the corresponding standard deviation; the normalized data As model input;

[0032] Model training, constructing a time series training dataset through historical data of operations during the tightening process; using mean square error as the loss function of the model for training; using an adaptive moment estimation optimizer combined with a learning rate scheduler to reduce the risk of overfitting during training; during model training, using a pruning algorithm to remove neurons that contribute less to prediction, in order to further reduce the amount of model calculation and parameter size.

[0033] As a preferred technical solution of the present invention, the material identification model includes the following contents:

[0034] Model structure: The material recognition model is designed based on a combined architecture of a lightweight convolutional neural network and a decision tree, specifically including:

[0035] The input layer is used to receive the physical property data collected by the material sensor, including the density, hardness and surface texture data of the material;

[0036] The feature extraction module uses a lightweight convolutional neural network to extract the features of surface texture data. The lightweight convolutional neural network includes two convolutional layers and one fully connected layer. Each layer uses a 3*3 filter and a sparse activation function to reduce the computational complexity. Density and hardness are directly used as input features and are combined with texture features after dimensionality reduction.

[0037] In the classification module, the extracted features are used by an integrated decision tree classifier to determine the material type;

[0038] Output layer, the type of output material, including wood, metal, composite materials, and combined structures;

[0039] Data preprocessing, minimum normalization processing of density and hardness data, the formula is: ,in Indicates density or hardness data, represents the data after minimum normalization processing, and are the maximum and minimum values ​​of the data respectively; high-pass filtering and denoising technology is used to preprocess the surface texture data, remove the interference components in the collected data, and retain the key features;

[0040] Model training: collect density, hardness and surface texture data of samples of different material types to construct a material property training data set. During the training process, network pruning and weight quantization strategies are used to reduce the parameter scale of the convolutional neural network, and the weights are quantized to INT8 to reduce the computational complexity. Cross entropy loss is used as the loss function of the model, and the stochastic gradient descent algorithm is used for model training. The learning rate dynamic adjustment mechanism is combined to improve training efficiency. The feature importance evaluation in the random forest is used to select the features that contribute most to material classification.

[0041] As a preferred technical solution of the present invention, the adjustment of the working parameters includes the following contents:

[0042] Torque output adjustment:

[0043] According to the output results of the resistance prediction model, the torque output is adjusted in real time. When the resistance prediction value increases, the torque output is increased in advance to avoid interruption of tightening; when the resistance prediction value decreases, the torque output is reduced to avoid over-tightening or damage to the operation object;

[0044] The torque adjustment is achieved by the controller by adjusting the input current of the motor, and the specific relationship is as follows: T=k*I, where T is the output torque, I is the motor input current, and k is the torque constant of the motor;

[0045] When it is detected that the resistance reaches or exceeds the set impact threshold, the system automatically switches to the impact mode and dynamically adjusts the impact frequency and intensity according to the resistance change in the impact mode;

[0046] Pressure adjustment:

[0047] According to the output result of the material identification model and the real-time resistance prediction data, the linear pressure P applied by the linear drive device is dynamically adjusted, and the pressure adjustment includes:

[0048] For hard materials, increase the applied pressure to ensure that the screw is fully embedded. The pressure is calculated as follows: P = P s +k1*R, where P s is the initial set basic pressure, R is the real-time resistance prediction value, and k1 is the pressure adjustment coefficient of the hard material;

[0049] For soft materials, reduce the applied pressure to avoid surface damage. The pressure calculation formula is as follows: P = P s +k2*R, where k2 is the pressure adjustment coefficient of the soft material;

[0050] The pressure is controlled by a built-in linear drive device, and the pressure is changed by adjusting the input current of the force-applying motor in real time to achieve precise control.

[0051] As a preferred technical solution of the present invention, it also includes the activation of the impact mode:

[0052] The activation conditions include: the resistance prediction value output by the resistance prediction model exceeds the preset impact threshold, and the material identification model determines that the target material is a hard material or a high resistance layer in a composite material;

[0053] In impact mode, the impact frequency and impact intensity are dynamically adjusted according to the real-time resistance prediction value;

[0054] After the impact mode is enabled, the resistance changes are monitored in real time. When the resistance drops below the set threshold and the material is identified as a soft material, the impact mode is automatically exited and the normal tightening mode is restored.

[0055] As a preferred technical solution of the present invention, the feedback includes the following contents:

[0056] Real-time feedback and parameter adjustment: During operation, the torque sensor, feedback current sensor and pressure sensor are used to collect key data of the tightening process in real time, including torque value, feedback current value and applied pressure value. When it is detected that the deviation between the actual operating parameter and the set value exceeds the preset tolerance range, the corresponding parameters are automatically adjusted to optimize the current operation.

[0057] Abnormal feedback and prompts: if an abnormal situation is detected during the tightening process, the task will be interrupted immediately, and an alarm will be issued to the user through the display screen or sound and light signals, and the cause of the fault will be prompted;

[0058] In case of drill bit abnormality, the system generates replacement suggestions and reminds users to take action by monitoring the friction and temperature of the drill bit;

[0059] Feedback optimization loop, after each task is completed, adjusts the internal parameter model, including the resistance prediction model and the material identification model, based on the feedback data of the tightening process; the optimization loop dynamically adjusts the prediction threshold and control strategy by taking the historical feedback data as the input of the model.

[0060] As a preferred technical solution of the present invention, the data record includes the following contents:

[0061] Local storage mode: the screwdriver is equipped with a local data storage module to record the key data of each operation task, including torque value, feedback current value, applied pressure value, resistance prediction result, material identification result and operation time; the data is stored in a task-by-task recording mode, and an independent data file is generated after each task is completed, which supports users to export the recorded data through the device-side interface;

[0062] Offline analysis mechanism supports uploading local data to the cloud by connecting to external devices, viewing operation records, generating data reports and conducting quality traceability through the cloud;

[0063] Circular storage mechanism: when the storage module capacity is about to run out, the earliest task record will be overwritten first; the system will remind the user to export data before overwriting;

[0064] Data files are stored in a compressed format to reduce storage space, while a verification mechanism is used to ensure data integrity.

[0065] A control system for an intelligent screwdriver includes the following modules:

[0066] Setting the initialization module: It is used to load the matching initial working parameters from the preset material database and screw database according to the task parameters input by the user, and detect the status of the drill bit to determine whether it meets the current task requirements. If not, the user is prompted to change the recommended drill bit model;

[0067] Real-time data acquisition module: used to collect real-time data during the tightening process through built-in torque sensor, feedback current sensor, material property sensor and pressure sensor;

[0068] Resistance prediction and material identification module: used to predict the resistance change trend based on the torque and feedback current data collected in real time using the resistance prediction model; based on the density, hardness and surface texture data obtained by the material property sensor, the material type of the target operation object is determined through the material identification model;

[0069] Dynamic adjustment and execution module: used to dynamically adjust the working parameters of the screwdriver according to the resistance prediction results and material identification results;

[0070] Feedback and data recording module: used to monitor the tightening status during operation, including monitoring drill wear, real-time torque, applied pressure and impact mode parameters during the tightening process; generates prompt information when an abnormality is detected; used to record real-time data of tightening operations and store them in the local data storage module, and supports exporting data files through the interface for offline analysis and task tracing.

[0071] The present invention has the following advantages:

[0072] The present invention, through the combination of the resistance prediction model and the material identification model, can dynamically adjust the working parameters of the screwdriver according to the real-time collected torque, feedback current, physical properties of the target material and applied pressure data, avoiding the traditional adjustment method that relies on manual experience, significantly improving operating efficiency and accuracy, and reducing the risk of human error. Through the application of the material identification model, the density, hardness, surface texture and other characteristics of the target material are accurately identified, and the working parameters are adjusted accordingly, so that the screwdriver can adapt to a variety of materials including metals, woods, composite materials, and even complex task scenarios with different material combination structures, thereby significantly improving the versatility and adaptability of the equipment.

[0073] The present invention uses multiple sensors to collect real-time data during the operation process, combined with a dynamic feedback mechanism, to provide timely reminders or automatic adjustments to abnormal situations to avoid mission failure or equipment damage. In addition, the feedback data can also be used to optimize the resistance prediction model and material identification model to further improve the performance of the system.

[0074] The present invention adopts lightweight recursive neural network and convolutional neural network structures, combined with model pruning and weight quantization techniques, which significantly reduces computational complexity and energy consumption, enabling the device to operate efficiently in a weak computing power environment, and is suitable for a variety of scenarios such as industrial assembly, construction, and home use; at the same time, the introduction of the intelligent control module greatly improves the automation level of the equipment. BRIEF DESCRIPTION OF THE DRAWINGS

[0075] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings required for use in the embodiments or the prior art description are briefly introduced below. Obviously, the drawings in the following description are only schematic diagrams of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on the provided drawings without creative work.

[0076] Figure 1 The present invention is a schematic diagram of the structure of a control system of an intelligent screwdriver used in an embodiment of the present invention. DETAILED DESCRIPTION

[0077] In order to make the purpose, technical scheme and advantages of the present invention clearer, the present invention will be further described in detail below in conjunction with the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, rather than all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0078] Embodiment 1, a control method of an intelligent screwdriver, comprising the following steps:

[0079] Step S1, setting initialization: according to the task parameters input by the user, matching initial working parameters are loaded from the preset material database and screw database, and the state of the drill bit is detected to determine whether it meets the current task requirements. If not, the user is prompted to change the recommended drill bit model; the task parameters include screw specifications and target material types; the state of the drill bit is detected by the rotation smoothness, friction and temperature data of the drill bit;

[0080] The acquisition of the matching initial working parameters includes:

[0081] According to the screw specifications input by the user, the corresponding screw characteristic data is obtained from the screw database, including the screw size, material and recommended tightening torque range; according to the target material type input by the user, the physical property data of the target material is obtained from the material database, including density, hardness and surface texture parameters;

[0082] According to the screw characteristic data and the physical characteristic data of the target material, the initial working parameters are matched in combination with the process requirements corresponding to the task parameters. The initial working parameters include the initial torque, the trigger threshold of the impact mode, the application range of the linear pressure and the adapted rotation speed. When the user does not input the task parameters, the default initial working parameters are recommended based on the historical operation data and the task type.

[0083] The detection of the drill bit status includes:

[0084] Parameter calculation:

[0085] By monitoring the fluctuation of the drill bit's rotation speed, the deviation of its smoothness is calculated , the formula is as follows:

[0086] ,in is the rotation speed of the i-th sampling, is the average value of the rotation speed, is the sampling times, the minimum sampling times is 10;

[0087] By measuring the friction between the drill bit and the material , to judge the degree of drill bit wear, the friction force is calculated as follows: ,in is the real-time measured torque, is the radius of the drill bit, is the friction coefficient of the contact surface, is the applied axial pressure, is the contact area;

[0088] Real-time monitoring of drill bit temperature via built-in temperature sensor ;

[0089] Evaluation parameter quantization:

[0090] Quantify the degree of deviation from the smoothness of the drill bit rotation as a condition score , the calculation formula is: ;in is the maximum allowable deviation of the rotational smoothness setting; if If it is less than the set threshold, it is judged that there is an abnormality in the rotation smoothness;

[0091] Convert friction test results into condition scores , the calculation formula is: ,in is the corresponding theoretical friction force in the target material database; if If it is less than the set threshold, it is judged that the drill bit is severely worn;

[0092] Convert drill bit temperature into a condition score , the calculation formula is: ,in is the maximum operating temperature; if If it is less than the set threshold, the drill is judged to be overheated;

[0093] Comprehensive status score calculation:

[0094] Calculates a comprehensive condition score by combining rotational smoothness, friction and temperature , the formula is: ,in They are weight coefficients of the corresponding parameters, and the sum of the three weight coefficients is 1.2. They need to be adjusted according to specific scenarios. For example, for operations that are more sensitive to rotation smoothness, w1>w2,w3 can be set; for high-precision operations (such as hard materials), the weight of friction can be increased to w2>w1,w3;

[0095] When the comprehensive status score is lower than the set threshold, the drill bit status is determined to be abnormal; and each parameter is evaluated independently to further analyze the cause of the abnormality, such as:

[0096] Abnormality 1: Unsteady rotation, which may be caused by drill bit deformation, loose assembly or foreign matter;

[0097] Abnormality 2: The friction deviation is too large, which may be caused by drill bit wear or material mismatch;

[0098] Abnormality 3: The temperature is too high, which may be caused by overload, insufficient lubrication or long-term continuous use;

[0099] Provide users with specific abnormality causes and solutions through the display or voice, such as: "The drill bit rotates unsteadily, please check the drill bit installation or replace the drill bit"; "The drill bit is severely worn, it is recommended to replace the drill bit with model [X]"; "The drill bit temperature is too high, please stop using it and check the cooling system."

[0100] Automatic adjustment: If the anomaly is a minor deviation, the system can extend the service life of the drill bit through real-time parameter adjustments (such as reducing torque and reducing the frequency of impact mode).

[0101] Emergency Stop: When a serious anomaly is detected, such as a possible drill bit breakage or a high risk of overheating, the system can automatically stop operation to avoid damage to equipment or materials.

[0102] Step S2, real-time data collection: real-time data during the operation is collected through the built-in torque sensor, feedback current sensor, material property sensor and pressure sensor of the screwdriver; the real-time data includes torque, feedback current, physical properties of the target material and applied pressure data;

[0103] The details of each sensor are as follows:

[0104] The torque sensor is installed on the power output shaft of the screwdriver and is used to collect torque data in real time during operation.

[0105] The feedback current sensor is integrated into the motor control module of the screwdriver and is used to monitor the feedback current when the motor is running.

[0106] Material property sensor: It adopts a combination of laser scanning sensor and optical sensor, installed on the fixed frame near the drill bit at the front end of the screwdriver, keeping a safe distance from the drill bit. The detection direction of the sensor is coaxial with the contact point of the drill bit to scan the surface of the target material. In a non-contact way, the following physical properties of the target material are obtained in real time, including density: calculated by scanning the reflectivity characteristics of the material and the intensity of the laser scattering signal; hardness: calculated by combining the material texture data and the preset hardness database; surface texture: using optical sensors to analyze the surface morphology of the material. Before the drill bit contacts the material, the laser sensor completes the surface scan of the material and obtains the preliminary characteristic values ​​of density and hardness. The optical sensor records the surface texture data through high-precision imaging, and combines the material database for pattern matching to determine the material type (such as metal, wood or composite material). All collected data are denoised and normalized by the signal processing module and transmitted to the material recognition model.

[0107] Pressure sensor: Installed in the linear drive to detect the pressure applied by the screwdriver to the target object.

[0108] Step S3, resistance prediction and material identification: Based on the real-time collected torque and feedback current data, the resistance change trend is predicted using the resistance prediction model; based on the density, hardness and surface texture data obtained by the material property sensor, the material type and combination structure of the target operation object are determined through the material identification model, and the material types specifically include wood, metal, composite materials and combination structures;

[0109] The resistance prediction model and material identification model are constructed based on artificial intelligence (AI) machine learning technology; the final lightweight AI model is obtained through training, verification and adjustment through historical data and embedded in the control system of the screwdriver;

[0110] The resistance prediction model includes the following contents:

[0111] Model structure: The resistance prediction model adopts a lightweight recursive neural network structure, which specifically includes: an input layer for receiving the time series of real-time collected torque data and feedback current data; a hidden layer, which is composed of a group of gated recurrent units, for reducing the number of parameters and computational complexity and retaining the long-term dependencies in the time series; an output layer, which outputs the predicted resistance value at the next moment; the lightweight design of the model is adapted to the weak computing power environment;

[0112] Data preprocessing, real-time acquisition of torque and feedback current data Perform standard normalization processing, the specific formula is:

[0113] ,in and is the corresponding historical mean, and is the corresponding standard deviation; the normalized data As model input;

[0114] Model training, constructing a time series training dataset through historical data of operations during the tightening process; using mean square error as the loss function of the model for training; using an adaptive moment estimation optimizer combined with a learning rate scheduler to reduce the risk of overfitting during training; during model training, using a pruning algorithm to remove neurons that contribute less to prediction, in order to further reduce the amount of model calculation and parameter size.

[0115] The advantages of the model are that it uses gated recurrent units to replace traditional long-short time neural units, significantly reducing computational complexity while retaining good time series modeling capabilities; for weak computing power environments, it uses a model quantization method to convert floating-point weights into low-precision (INT8) weights to further improve prediction efficiency; in the prediction process, it combines the sliding window mechanism of historical data, and only requires a small amount of real-time data to complete high-precision resistance prediction.

[0116] Model application: During operation, the model takes the real-time collected torque and feedback current data as input, and outputs the resistance prediction value at the next moment. The resistance prediction result is used to dynamically adjust the torque, impact mode and applied pressure to ensure the accuracy and stability of the tightening process.

[0117] The material identification model includes the following contents:

[0118] Model structure: The material recognition model is designed based on a combined architecture of a lightweight convolutional neural network and a decision tree, specifically including:

[0119] The input layer is used to receive the physical property data collected by the material sensor, including the density, hardness and surface texture data of the material;

[0120] The feature extraction module uses a lightweight convolutional neural network to extract the features of surface texture data. The lightweight convolutional neural network includes two convolutional layers and one fully connected layer. Each layer uses a 3*3 filter and a sparse activation function to reduce the computational complexity. Density and hardness are directly used as input features and are combined with texture features after dimensionality reduction.

[0121] In the classification module, the extracted features are used by an integrated decision tree classifier to determine the material type;

[0122] Output layer, the type of output material, including wood, metal, composite materials, and combined structures;

[0123] Data preprocessing, minimum normalization processing of density and hardness data, the formula is: ,in Indicates density or hardness data, represents the data after minimum normalization processing, and are the maximum and minimum values ​​of the data respectively; high-pass filtering and denoising technology is used to preprocess the surface texture data, remove the interference components in the collected data, and retain the key features;

[0124] Model training: collect density, hardness and surface texture data of samples of different material types to construct a material property training data set. During the training process, network pruning and weight quantization strategies are used to reduce the parameter scale of the convolutional neural network, and the weights are quantized to INT8 to reduce the computational complexity. Cross entropy loss is used as the loss function of the model, and the stochastic gradient descent algorithm is used for model training. The learning rate dynamic adjustment mechanism is combined to improve training efficiency. The feature importance evaluation in the random forest is used to select the features that contribute most to material classification.

[0125] The advantages of the model are that it uses a lightweight convolutional neural network to extract surface texture features, combines density and hardness characteristics, and uses a decision tree classifier to identify materials, significantly reducing the complexity of reasoning; it optimizes the model's computational efficiency through network pruning and weight quantization methods to adapt to weak computing power environments; it introduces high-pass filtering technology to optimize texture data input and improve material recognition accuracy.

[0126] Model application. During operation, the material recognition model takes the density, hardness and surface texture characteristic data collected by the sensor as input, and outputs the material type and structure results after feature extraction and classification; the recognition results are used to dynamically adjust the working parameters of the screwdriver, including torque output, impact mode and applied pressure, so as to adapt to the operating requirements of different materials.

[0127] Step S4, dynamic regulation and execution: dynamically adjust the working parameters of the screwdriver according to the resistance prediction and material identification results; the adjustment of the working parameters includes adjusting the torque output, increasing or decreasing the torque in advance according to the resistance change; when the resistance exceeds the threshold or the target material is a hard material, enabling the impact mode, and dynamically adjusting the impact frequency and intensity; based on the material identification results and the resistance situation, regulating the pressure applied by the linear drive device, increasing the pressure for hard materials and reducing the pressure for soft materials;

[0128] The adjustment of the working parameters includes the following:

[0129] Torque output adjustment:

[0130] According to the output results of the resistance prediction model, the torque output is adjusted in real time. When the resistance prediction value increases, the torque output is increased in advance to avoid interruption of tightening; when the resistance prediction value decreases, the torque output is reduced to avoid over-tightening or damage to the operation object;

[0131] The torque adjustment is achieved by the controller by adjusting the input current of the motor, and the specific relationship is as follows: T=k*I, where T is the output torque, I is the motor input current, and k is the torque constant of the motor;

[0132] When it is detected that the resistance reaches or exceeds the set impact threshold, the system automatically switches to the impact mode and dynamically adjusts the impact frequency and intensity according to the resistance change in the impact mode;

[0133] Pressure adjustment:

[0134] According to the output result of the material identification model and the real-time resistance prediction data, the linear pressure P applied by the linear drive device is dynamically adjusted, and the pressure adjustment includes:

[0135] For hard materials, increase the applied pressure to ensure that the screw is fully embedded. The pressure is calculated as follows: P = P s +k1*R, where P s is the initial set basic pressure, R is the real-time resistance prediction value, and k1 is the pressure adjustment coefficient of the hard material;

[0136] For soft materials, reduce the applied pressure to avoid surface damage. The pressure calculation formula is as follows: P = P s +k2*R, where k2 is the pressure adjustment coefficient of the soft material;

[0137] The pressure is controlled by a built-in linear drive device, and the pressure is changed by adjusting the input current of the force-applying motor in real time to achieve precise control.

[0138] To activate the impact mode:

[0139] The activation conditions include: the resistance prediction value output by the resistance prediction model exceeds the preset impact threshold, and the material identification model determines that the target material is a hard material or a high resistance layer in a composite material;

[0140] In impact mode, the impact frequency f is dynamically adjusted according to the real-time resistance prediction value. i And impact strength T i ; The adjustment formulas are as follows: f i =f s +k3*R,T i =T s +k4*R, where f s is the basic impact frequency, k3 is the impact frequency adjustment coefficient; the greater the resistance, the higher the impact frequency; T s is the basic impact torque, k4 is the impact torque adjustment coefficient; the greater the resistance, the higher the impact strength;

[0141] After the impact mode is enabled, the resistance changes are monitored in real time. When the resistance drops below the set threshold and the material is identified as a soft material, the impact mode is automatically exited and the normal tightening mode is restored.

[0142] Step S5, feedback and data recording: monitor the drill bit status during the operation, and prompt the user to replace it if the drill bit is worn or mismatched; record the real-time data of the tightening operation process and store it locally or in the cloud for task tracing and model optimization.

[0143] The feedback includes the following:

[0144] Real-time feedback and parameter adjustment: During operation, the torque sensor, feedback current sensor and pressure sensor are used to collect key data of the tightening process in real time, including torque value, feedback current value and applied pressure value. When it is detected that the deviation between the actual operating parameter (such as torque or pressure) and the set value exceeds the preset tolerance range, the corresponding parameter is automatically adjusted to optimize the current operation.

[0145] Abnormal feedback and prompts: During the tightening process, if an abnormal situation is detected (such as the screw is not tightened, the screw is over-tightened, or the resistance curve is abnormal), the task will be interrupted immediately, and an alarm will be issued to the user through the display or sound and light signals, and the cause of the fault will be prompted;

[0146] In case of drill bit abnormalities (such as excessive wear or unstable rotation), the system generates replacement suggestions and reminds users to take action by monitoring the friction and temperature of the drill bit;

[0147] Feedback optimization loop, after each task is completed, adjusts the internal parameter model, including the resistance prediction model and the material identification model, based on the feedback data of the tightening process; the optimization loop dynamically adjusts the prediction threshold and control strategy by taking the historical feedback data as the input of the model.

[0148] The data record includes the following:

[0149] Local storage mode: the screwdriver is equipped with a local data storage module to record the key data of each operation task, including torque value, feedback current value, applied pressure value, resistance prediction result, material identification result and operation time; the data is stored in a task-by-task recording mode, and an independent data file is generated after each task is completed, which supports users to export the recorded data through the device-side interface (such as USB interface or memory card);

[0150] Offline analysis mechanism supports uploading local data to the cloud by connecting to external devices (such as PC or mobile terminal), viewing operation records, generating data reports and conducting quality traceability through the cloud;

[0151] Circular storage mechanism: when the storage module capacity is about to run out, the earliest task record will be overwritten first; the system will remind the user to export data before overwriting;

[0152] Data files are stored in a compressed format to reduce storage space, while a verification mechanism is used to ensure data integrity.

[0153] Example 2, a control system for an intelligent screwdriver, see Figure 1 As shown, it includes the following modules:

[0154] Setting the initialization module: It is used to load the matching initial working parameters from the preset material database and screw database according to the task parameters input by the user, and detect the status of the drill bit to determine whether it meets the current task requirements. If not, the user is prompted to change the recommended drill bit model;

[0155] Real-time data acquisition module: used to collect real-time data during the tightening process through built-in torque sensor, feedback current sensor, material property sensor and pressure sensor;

[0156] Resistance prediction and material identification module: used to predict the resistance change trend based on the torque and feedback current data collected in real time using the resistance prediction model; based on the density, hardness and surface texture data obtained by the material property sensor, the material type of the target operation object is determined through the material identification model;

[0157] Dynamic adjustment and execution module: used to dynamically adjust the working parameters of the screwdriver according to the resistance prediction results and material identification results;

[0158] Feedback and data recording module: used to monitor the tightening status during operation, including monitoring drill wear, real-time torque, applied pressure and impact mode parameters during the tightening process; generates prompt information when an abnormality is detected; used to record real-time data of tightening operations and store them in the local data storage module, and supports exporting data files through the interface for offline analysis and task tracing.

[0159] The specific implementation methods described above further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific implementation method of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A control method for an intelligent screwdriver, characterized in that: The following steps are involved: Step S1, setting initialization: according to the task parameters input by the user, matching initial working parameters are loaded from the preset material database and screw database, and the state of the drill bit is detected to determine whether it meets the current task requirements. If not, the user is prompted to change the recommended drill bit model; the task parameters include screw specifications and target material types; the state of the drill bit is detected by the rotation smoothness, friction and temperature data of the drill bit; Step S2, real-time data collection: real-time data during the operation is collected through the built-in torque sensor, feedback current sensor, material property sensor and pressure sensor of the screwdriver; the real-time data includes torque, feedback current, physical properties of the target material and applied pressure data; Step S3, resistance prediction and material identification: Based on the real-time collected torque and feedback current data, the resistance change trend is predicted using the resistance prediction model; based on the density, hardness and surface texture data obtained by the material property sensor, the material type and combination structure of the target operation object are determined through the material identification model, and the material types specifically include wood, metal, composite materials and combination structures; Step S4, dynamic regulation and execution: dynamically adjust the working parameters of the screwdriver according to the resistance prediction and material identification results; the adjustment of the working parameters includes adjusting the torque output, increasing or decreasing the torque in advance according to the resistance change; when the resistance exceeds the threshold or the target material is a hard material, enabling the impact mode, and dynamically adjusting the impact frequency and intensity; based on the material identification results and the resistance situation, regulating the pressure applied by the linear drive device, increasing the pressure for hard materials and reducing the pressure for soft materials; Step S5, feedback and data recording: monitor the drill bit status during the operation, and prompt the user to replace it if the drill bit is worn or mismatched; record the real-time data of the tightening operation process and store it locally or in the cloud for task tracing and model optimization.

2. The control method of an intelligent screwdriver according to claim 1, characterized in that: The acquisition of the matching initial working parameters includes: According to the screw specifications input by the user, the corresponding screw characteristic data is obtained from the screw database, including the screw size, material and recommended tightening torque range; according to the target material type input by the user, the physical property data of the target material is obtained from the material database, including density, hardness and surface texture parameters; According to the screw characteristic data and the physical characteristic data of the target material, the initial working parameters are matched in combination with the process requirements corresponding to the task parameters. The initial working parameters include the initial torque, the trigger threshold of the impact mode, the application range of the linear pressure and the adapted rotation speed. When the user does not input the task parameters, the default initial working parameters are recommended based on the historical operation data and the task type.

3. The control method of an intelligent screwdriver according to claim 1, characterized in that: The detection of the drill bit status includes: Parameter calculation: By monitoring the fluctuation of the drill bit's rotation speed, the deviation of its smoothness is calculated , the formula is as follows: ,in is the rotation speed of the i-th sampling, is the average value of the rotation speed, is the sampling times, the minimum sampling times is 10; By measuring the friction between the drill bit and the material , to judge the degree of drill bit wear, the friction force is calculated as follows: ,in is the real-time measured torque, is the radius of the drill bit, is the friction coefficient of the contact surface, is the applied axial pressure, is the contact area; Real-time monitoring of drill bit temperature via built-in temperature sensor ; Evaluation parameter quantization: Quantify the degree of deviation from the smoothness of the drill bit rotation as a condition score , the calculation formula is: ;in is the maximum allowable deviation of the rotational smoothness setting; if If it is less than the set threshold, it is judged that there is an abnormality in the rotation smoothness; Convert friction test results into condition scores , the calculation formula is: ,in is the corresponding theoretical friction force in the target material database; if If it is less than the set threshold, it is judged that the drill bit is severely worn; Convert drill bit temperature into a condition score , the calculation formula is: ,in is the maximum operating temperature; if If it is less than the set threshold, it is judged as the drill bit overheating; Comprehensive status score calculation: Calculates a comprehensive condition score by combining rotational smoothness, friction and temperature , the formula is: ,in They are the weight coefficients of the corresponding parameters, which need to be adjusted according to the specific scenario, and the sum of the three weight coefficients is 1.2; When the comprehensive status score is lower than the set threshold, the drill bit status is judged to be abnormal; and each parameter is evaluated independently to further analyze the cause of the abnormality.

4. The control method of an intelligent screwdriver according to claim 1, characterized in that: The resistance prediction model includes the following contents: Model structure: The resistance prediction model adopts a lightweight recursive neural network structure, which specifically includes: an input layer for receiving the time series of real-time collected torque data and feedback current data; a hidden layer, which is composed of a group of gated recurrent units, for reducing the number of parameters and computational complexity and retaining the long-term dependencies in the time series; an output layer, which outputs the predicted resistance value at the next moment; the lightweight design of the model is adapted to the weak computing power environment; Data preprocessing, real-time acquisition of torque and feedback current data Perform standard normalization processing, the specific formula is: ,in and is the corresponding historical mean, and is the corresponding standard deviation; the normalized data As model input; Model training, constructing a time series training dataset through historical data of operations during the tightening process; using mean square error as the loss function of the model for training; using an adaptive moment estimation optimizer combined with a learning rate scheduler to reduce the risk of overfitting during training; during model training, using a pruning algorithm to remove neurons that contribute less to prediction, in order to further reduce the amount of model calculation and parameter size.

5. The control method of an intelligent screwdriver according to claim 1, characterized in that: The material identification model includes the following contents: Model structure: The material recognition model is designed based on a combined architecture of a lightweight convolutional neural network and a decision tree, specifically including: The input layer is used to receive the physical property data collected by the material sensor, including the density, hardness and surface texture data of the material; The feature extraction module uses a lightweight convolutional neural network to extract the features of surface texture data. The lightweight convolutional neural network includes two convolutional layers and one fully connected layer. Each layer uses a 3*3 filter and a sparse activation function to reduce the computational complexity. Density and hardness are directly used as input features and are combined with texture features after dimensionality reduction. In the classification module, the extracted features are used by an integrated decision tree classifier to determine the material type; Output layer, the type of output material, including wood, metal, composite materials, and combined structures; Data preprocessing, minimum normalization processing of density and hardness data, the formula is: ,in Indicates density or hardness data, represents the data after minimum normalization processing, and are the maximum and minimum values ​​of the data respectively; high-pass filtering and denoising technology is used to preprocess the surface texture data, remove the interference components in the collected data, and retain the key features; Model training: collect density, hardness and surface texture data of samples of different material types to construct a material property training data set. During the training process, network pruning and weight quantization strategies are used to reduce the parameter scale of the convolutional neural network, and the weights are quantized to INT8 to reduce the computational complexity. Cross entropy loss is used as the loss function of the model, and the stochastic gradient descent algorithm is used for model training. The learning rate dynamic adjustment mechanism is combined to improve training efficiency. The feature importance evaluation in the random forest is used to select the features that contribute most to material classification.

6. The control method of an intelligent screwdriver according to claim 1, characterized in that: The adjustment of the working parameters includes the following: Torque output adjustment: According to the output results of the resistance prediction model, the torque output is adjusted in real time. When the resistance prediction value increases, the torque output is increased in advance to avoid interruption of tightening; when the resistance prediction value decreases, the torque output is reduced to avoid over-tightening or damage to the operation object; The torque adjustment is achieved by the controller by adjusting the input current of the motor, and the specific relationship is as follows: T=k*I, where T is the output torque, I is the motor input current, and k is the torque constant of the motor; When it is detected that the resistance reaches or exceeds the set impact threshold, the system automatically switches to the impact mode and dynamically adjusts the impact frequency and intensity according to the resistance change in the impact mode; Pressure adjustment: According to the output result of the material identification model and the real-time resistance prediction data, the linear pressure P applied by the linear drive device is dynamically adjusted, and the pressure adjustment includes: For hard materials, increase the applied pressure to ensure that the screw is fully embedded. The pressure is calculated as follows: P = P s +k1*R, where P s is the initial set basic pressure, R is the real-time resistance prediction value, and k1 is the pressure adjustment coefficient of the hard material; For soft materials, reduce the applied pressure to avoid surface damage. The pressure calculation formula is as follows: P = P s +k2*R, where k2 is the pressure adjustment coefficient of the soft material; The pressure is controlled by a built-in linear drive device, and the pressure is changed by adjusting the input current of the force-applying motor in real time to achieve precise control.

7. The control method of the intelligent screwdriver according to claim 6, characterized in that: Also includes the activation of impact mode: The activation conditions include: the resistance prediction value output by the resistance prediction model exceeds the preset impact threshold, and the material identification model determines that the target material is a hard material or a high resistance layer in a composite material; In impact mode, the impact frequency and impact intensity are dynamically adjusted according to the real-time resistance prediction value; After the impact mode is enabled, the resistance changes are monitored in real time. When the resistance drops below the set threshold and the material is identified as a soft material, the impact mode is automatically exited and the normal tightening mode is restored.

8. The control method of an intelligent screwdriver according to claim 1, characterized in that: The feedback includes the following: Real-time feedback and parameter adjustment: During operation, the torque sensor, feedback current sensor and pressure sensor are used to collect key data of the tightening process in real time, including torque value, feedback current value and applied pressure value. When it is detected that the deviation between the actual operating parameter and the set value exceeds the preset tolerance range, the corresponding parameters are automatically adjusted to optimize the current operation. Abnormal feedback and prompts: if an abnormal situation is detected during the tightening process, the task will be interrupted immediately, and an alarm will be issued to the user through the display screen or sound and light signals, and the cause of the fault will be prompted; In case of drill bit abnormality, the system generates replacement suggestions and reminds users to take action by monitoring the friction and temperature of the drill bit; Feedback optimization loop, after each task is completed, adjusts the internal parameter model, including the resistance prediction model and the material identification model, based on the feedback data of the tightening process; the optimization loop dynamically adjusts the prediction threshold and control strategy by taking the historical feedback data as the input of the model.

9. The control method of an intelligent screwdriver according to claim 1, characterized in that: The data record includes the following: Local storage mode: the screwdriver is equipped with a local data storage module to record the key data of each operation task, including torque value, feedback current value, applied pressure value, resistance prediction result, material identification result and operation time; the data is stored in a task-by-task recording mode, and an independent data file is generated after each task is completed, which supports users to export the recorded data through the device-side interface; Offline analysis mechanism supports uploading local data to the cloud by connecting to external devices, viewing operation records, generating data reports and conducting quality traceability through the cloud; Circular storage mechanism: when the storage module capacity is about to run out, the earliest task record will be overwritten first; the system will remind the user to export data before overwriting; Data files are stored in a compressed format to reduce storage space, while a verification mechanism is used to ensure data integrity.

10. A control system for an intelligent screwdriver, characterized in that: The system applies a control method for an intelligent screwdriver according to any one of claims 1 to 9, and comprises the following modules: Setting the initialization module: It is used to load the matching initial working parameters from the preset material database and screw database according to the task parameters input by the user, and detect the status of the drill bit to determine whether it meets the current task requirements. If not, the user is prompted to change the recommended drill bit model; Real-time data acquisition module: used to collect real-time data during the tightening process through built-in torque sensor, feedback current sensor, material property sensor and pressure sensor; Resistance prediction and material identification module: used to predict the resistance change trend based on the torque and feedback current data collected in real time using the resistance prediction model; based on the density, hardness and surface texture data obtained by the material property sensor, the material type of the target operation object is determined through the material identification model; Dynamic adjustment and execution module: used to dynamically adjust the working parameters of the screwdriver according to the resistance prediction results and material identification results; Feedback and data recording module: used to monitor the tightening status during operation, including monitoring drill wear, real-time torque, applied pressure and impact mode parameters during tightening; when an abnormality is detected, a prompt message is generated; It is used to record the real-time data of tightening operations and store it in the local data storage module, and supports exporting data files through the interface for offline analysis and task tracing.

Images