Multi-mode electronic device AI self-adaptive embedding and weaving process for intelligent textile fabric fibers

Abstract

The invention discloses a multi-mode electronic device AI self-adaptive embedding weaving process of intelligent textile fabric fibers, and relates to the technical field of electronic information, comprising process pretreatment, AI self-adaptive positioning and path planning, self-adaptive embedding weaving and post-treatment and performance optimization. AI self-adaptive positioning and path planning comprises real-time position acquisition, a path planning algorithm and digital twin auxiliary path planning, and the method has the advantages that position information of textile fabrics and electronic devices is accurately acquired through visual identification system enhancement and a multi-modal data fusion positioning technology; in combination with innovation means such as A * algorithm optimization and multi-objective decision, reinforcement learning path exploration and digital twin auxiliary path planning, the embedding and weaving precision of a multi-mode electronic device in intelligent textile fabric fibers is remarkably improved, and the problem that the device and the fibers are not tightly combined due to positioning deviation is effectively avoided; the problem that in the traditional technology, connection failure is prone to occurring when fabric deforms is fundamentally solved.

Description

Technical Field

[0001] This invention relates to the field of electronic information technology, specifically to an AI adaptive weaving process for multimodal electronic devices in smart textile fibers. Background Technology

[0002] With the rapid development of smart wearable devices, the integration of multimodal electronic devices and smart textiles has become an important development direction. Traditional electronic device and textile weaving processes have many problems: on the one hand, the combination of electronic devices and textile fibers lacks precision, which makes the connection of devices prone to failure when the fabric is deformed; on the other hand, in the face of complex scenarios such as human movement and environmental changes, it is impossible to adaptively adjust the weaving state and device performance. Existing technologies are difficult to meet the requirements of smart textiles for high integration, high stability and adaptive functions of multimodal electronic devices. To this end, we propose an AI adaptive weaving process for multimodal electronic devices in smart textile fibers. Summary of the Invention

[0003] The purpose of this invention is to provide an AI-adaptive weaving process for multimodal electronic devices in smart textile fibers.

[0004] To achieve the above objectives, the present invention provides the following technical solution: a multimodal electronic device AI adaptive weaving process for intelligent textile fibers, including process preprocessing, AI adaptive positioning and path planning, adaptive weaving, and post-processing and performance optimization. The AI ​​adaptive positioning and path planning includes real-time position acquisition, path planning algorithms, and digital twin-assisted path planning. The specific operation method of the multimodal electronic device AI adaptive weaving is as follows: Step 1: After cleaning and treating the textile fibers, test their basic properties and check the status of multimodal electronic devices; Step 2: Use a high-precision visual recognition device to collect the texture feature points of the textile and the edge contours of the electronic devices, use an algorithm to plan the weaving path, and balance the path length according to the generation requirements. Step 3: Set initial weaving parameters, and monitor and adjust the parameters in real time; Step 4: After weaving is completed, the textile is subjected to hot pressing treatment. After the treatment is completed, the operating parameters of the device are adjusted according to environmental parameters to dynamically optimize the performance of the electronic device.

[0005] As a further aspect of the present invention: In step one, plasma cleaning technology is used to treat the fabric for 3-5 minutes at a power of 100W-150W and an argon gas flow rate of 50sccm-80sccm. After the treatment, the tensile strength and elastic modulus parameters of the textile are tested using a mechanical property testing device. At the same time, the multimodal electronic devices are functionally tested, including the sensitivity of the pressure sensor and the measurement accuracy of the temperature sensor. After the test is passed, the device is coded. The coding information includes the device type, functional parameters and the embedding position.

[0006] As a further aspect of the present invention: In step two, real-time location acquisition includes enhanced visual recognition system and multimodal data fusion positioning. The enhanced visual recognition system uses a high-precision visual recognition device to perform real-time image acquisition of textiles and multimodal electronic devices, including texture feature points and edge contours. At the same time, a semantic segmentation algorithm based on deep learning is introduced to semantically annotate the material distribution of textiles, fiber orientation, and functional areas of electronic devices. The multimodal data fusion positioning is based on the idea of ​​a multimodal large model of power artificial intelligence, which integrates sensor data to achieve accurate positioning. A laser rangefinder and an ultrasonic sensor are added to verify and calibrate the location information acquired by visual recognition.

[0007] As a further aspect of the present invention: in step two, the path planning algorithm includes A* algorithm optimization and multi-objective decision-making, path planning strategy, and path exploration using reinforcement learning, wherein the cost function of the path planning algorithm... The risk coefficient is calculated based on the above, where Indicates the first inlay path Length of the path segment Indicates the first inlay path The risk coefficient corresponding to the segment path is assessed based on the performance differences of the textile and mechanical performance parameters. When planning the weaving path, a dynamic adjustment strategy is adopted to correct the weaving path in real time based on the real-time monitoring of the textile's movement status.

[0008] As a further aspect of the present invention: the path exploration of the reinforcement learning introduces a reinforcement learning mechanism, which works in conjunction with the A* algorithm. The path planning process is regarded as an interaction between an agent and the environment. The agent learns the optimal path planning strategy by continuously trying different path choices and based on the reward signals fed back by the environment.

[0009] As a further aspect of the present invention: the digital twin-assisted path planning includes constructing a digital twin model and virtual path verification and optimization. The digital twin model uses digital twin technology to construct a virtual model of textiles and multimodal electronic devices, models the physical properties of the textiles and the geometry and functional characteristics of the electronic devices, and simulates the weaving process in a virtual environment. The model contains static information of the textiles and electronic devices, while reflecting the dynamic changes in the production process in real time. The virtual path verification and optimization involves simulating and verifying the planned path in the digital twin model before the actual weaving operation.

[0010] As a further aspect of the present invention: in step three, initial weaving parameters are set according to the material of the textile and the type of electronic device, with a weaving speed of 100mm / s-200mm / s and a weaving force of 0.5N-1N.

[0011] As a further aspect of the present invention: In step three, during the weaving process, the connection resistance between the electronic devices and the textile is acquired in real time through a built-in resistance monitoring device. Using displacement sensors to detect device position deviation ,exist Out of range At that time, the AI ​​system processes according to the preset adjustment rules, where the LSTM network structure is input layer - hidden layer (32 layers) - output layer.

[0012] As a further aspect of the present invention: In step four, after the weaving is completed, the textile is subjected to hot pressing treatment at a temperature of 110℃-130℃ and a pressure of 0.4MPa-0.6MPa for 2-4 minutes. The dynamic performance optimization is achieved by the AI ​​system using a long short-term memory network to predict the impact of environmental changes on the performance of multimodal electronic devices based on environmental parameters such as temperature and humidity collected by environmental sensors and human motion data obtained by other sensors on the textile. Based on the prediction results, the operating parameters of the devices are adjusted, and the sensitivity adjustment value of the sensors is adjusted accordingly. According to the formula The calculation shows that, among which The hidden state output by the LSTM network. and For weights and biases, This is the activation function.

[0013] Compared with the prior art, the beneficial effects of the present invention by adopting the above technical solution are as follows: 1. This invention enhances the positioning technology of visual recognition system and multimodal data fusion to accurately obtain the position information of textiles and electronic devices. Combined with innovative methods such as A* algorithm optimization and multi-objective decision-making, reinforcement learning path exploration and digital twin-assisted path planning, it significantly improves the embedding accuracy of multimodal electronic devices in smart textile fibers, effectively avoids the problem of loose bonding between devices and fibers due to positioning deviation, and fundamentally solves the problem of connection failure of devices when the fabric is deformed in traditional processes. 2. This invention uses a real-time monitoring device to dynamically monitor the connection resistance and position deviation. The AI ​​system quickly adjusts the weaving parameters according to preset rules to ensure that the electronic devices and textile fibers always maintain a stable connection. At the same time, after weaving is completed, the device operating parameters are dynamically optimized based on the prediction of environmental changes by the long short-term memory network, so that the multimodal electronic devices can adaptively adjust their performance in real time, effectively cope with complex environmental changes, and meet the needs of smart textiles for the adaptive function of multimodal electronic devices. 3. Through systematic process design, this invention achieves intelligent and precise control of the entire process from pre-processing to post-processing and performance optimization. This not only improves the integration and stability of multimodal electronic devices and smart textile fibers, but also enhances production efficiency. Compared with traditional processes, it can reduce the probability of device connection failure, improve production efficiency, effectively reduce production costs, and promote the development of smart textiles in the field of smart wearable devices towards higher performance and practicality. Attached Figure Description

[0014] Figure 1 This is a process flow diagram of the present invention. Detailed Implementation

[0015] The specific embodiments of the present invention will be further described below with reference to the accompanying drawings. It should be noted that the description of these embodiments is for the purpose of helping to understand the present invention, but does not constitute a limitation of the present invention.

[0016] Furthermore, the technical features involved in the various embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0017] Please see the appendix Figure 1 The present invention relates to a multimodal electronic device AI adaptive weaving process for smart textile fibers, comprising process preprocessing, AI adaptive positioning and path planning, adaptive weaving, and post-processing and performance optimization. AI adaptive positioning and path planning includes real-time location acquisition, path planning algorithms, and digital twin-assisted path planning. The specific operation method of multimodal electronic device AI adaptive weaving is as follows: Step 1: After cleaning and treating the textile fibers, test their basic properties and check the status of multimodal electronic devices; Step 2: Use a high-precision visual recognition device to collect the texture feature points of the textile and the edge contours of the electronic devices, use an algorithm to plan the weaving path, and balance the path length according to the generation requirements. Step 3: Set initial weaving parameters, and monitor and adjust the parameters in real time; Step 4: After weaving is completed, the textile is subjected to hot pressing treatment. After the treatment is completed, the operating parameters of the device are adjusted according to environmental parameters to dynamically optimize the performance of the electronic device.

[0018] In one embodiment of the present invention: In step one, the pressure sensor sensitivity test is performed by applying a pressure of 0N-5N using a standard pressure source, recording the output voltage change, and calculating the sensitivity. The temperature sensor measurement accuracy test is performed in a constant temperature chamber, comparing the sensor measurement value with the standard thermometer value within the range of 0℃-50℃, and the error must be controlled within ±0.2℃.

[0019] In one embodiment of the present invention: In step two, during the visual recognition system enhancement, the constructed semantic model library is periodically updated and iterated based on newly acquired samples to improve recognition adaptability. In the multimodal data fusion positioning, the laser ranging sensor has a sampling frequency of 100Hz and a measurement accuracy of ±0.01mm, while the ultrasonic sensor has a detection angle of 30°. Through a weighted fusion algorithm combining the data from both sensors with visual data, the fusion weights are dynamically adjusted based on the real-time error assessment of each sensor, and the cost function... ,in The value range is 0.3-0.6. The value range is 0.4-0.7, and ; Based on the tensile strength of the textile, areas below 5 MPa Values ​​ranging from 1.2 to 1.5, for regions exceeding 10 MPa. Values ​​range from 0.5 to 0.8; In the dynamic adjustment strategy, the weaving path is corrected when the displacement of the textile exceeds 0.1mm, and the correction response time does not exceed 50ms.

[0020] In one embodiment of the present invention: In step three, the initial weaving parameters are: Weaving speed: 100mm / s-150mm / s for cotton fibers, and 150mm / s-200mm / s for polyester fibers; Embedding force: 0.5N-0.7N for devices weighing less than 0.1g, and 0.8N-1.0N for devices weighing more than 0.3g.

[0021] Real-time detection and parameter adjustment: The resistance monitoring device has a sampling frequency of 50Hz and a measurement accuracy of ±0.1Ω; the displacement sensor has a measurement range of ±0.5mm and an accuracy of ±0.005mm. Normal connection resistance range: 50Ω-100Ω for pressure sensors, 80Ω-150Ω for temperature sensors; allowable position deviation error. ±0.05mm; Adjustment rules: When This means there is a risk of the connection being too tight or a short circuit. When the value exceeds 10Ω, it indicates a loose connection or poor contact. For every 10Ω increase, the weaving force increases by 0.1N, up to a maximum of 1.5N. At the same time, the position is corrected according to the direction of the deviation, and the correction amount is 1.2 times the deviation value. Simultaneously, the weaving speed is reduced by 20%. This represents the minimum threshold value for connection resistance. This indicates the maximum threshold value for the connection resistance.

[0022] In one embodiment of the present invention: In step four, the hot pressing parameters are: For cotton fiber, the temperature range is 110℃-120℃; for polyester fiber, it's 120℃-130℃. For devices smaller than 5mm×5mm, hot pressing takes 2-3 minutes; for larger than 5mm×5mm, it takes 3-4 minutes. Performance is currently undergoing dynamic optimization. The environmental sensor's temperature range is -10℃-50℃ with an accuracy of ±0.1℃, and its humidity range is 20%RH-90%RH with an accuracy of ±2%RH. The sensor sensitivity adjustment formula is as follows. middle, It is a 32-dimensional hidden state. It is a 32×1 weight matrix. The value range is (-0.2, 0.2). It is the sigmoid activation function. Within the range of 0.8-1.5 times the initial sensitivity.

[0023] In one embodiment of the present invention: in the path planning strategy of the path planning algorithm, when the displacement of the textile is detected, the AI ​​system quickly calculates the new starting point and target point, replans the path, and ensures that the electronic device can be accurately embedded in the predetermined position, reducing the embedding deviation caused by the movement of the textile.

[0024] In one embodiment of the present invention: In the virtual path verification and optimization of digital twin-assisted path planning, simulation allows for a direct observation of whether the path is reasonable and whether there are issues such as conflicts between electronic components and textile structures, or the smoothness of the weaving process. For example, if the simulation reveals that a certain path segment might cause excessive resistance to electronic components during weaving, the AI ​​system can adjust the path promptly based on the simulation feedback to avoid similar problems in actual production, thereby improving production efficiency and product quality. Simultaneously, numerous virtual experiments using the digital twin model explore the optimal path under different parameter settings, providing richer path planning references for actual production. Example 1: Production of a smart wristband for motion monitoring (polyester fiber textile); Process pretreatment: Polyester fiber textiles with a thickness of 0.3 mm were selected and treated with plasma cleaning technology. The power was set to 130W, the argon flow rate was 65sccm, and the treatment time was 4 minutes. After treatment, the tensile strength was tested with a tensile testing machine and the tensile strength was 12MPa and the elastic modulus was 2.5GPa. A PS200 pressure sensor and an AC100 accelerometer were selected for functional testing: the pressure sensor output voltage change rate was 0.5V / N under 0N-5N pressure, and the sensitivity was qualified; the temperature sensor's measurement error was ≤±0.1℃ within the range of 20℃-40℃. After passing the test, the devices were coded. The pressure sensor was marked as "P-001, embedded position: middle section of the inner side of the wristband", and the accelerometer was marked as "A-001, embedded position: joint on the outer side of the wristband". AI adaptive localization and path planning: Real-time location acquisition: A high-precision visual recognition device with a resolution of 1920×1080 is activated to acquire images of the textile. The fiber direction is marked by a semantic segmentation algorithm. The surface flatness error of the textile is measured to be ±0.03mm using a laser rangefinder. An ultrasonic sensor is used for calibration. After fusion, the positioning error is ≤±0.02mm. Path planning: using a cost function Because the tensile strength of polyester fiber is 12 MPa, which is higher than 10 MPa, With a value of 0.6, during reinforcement learning exploration, the high-risk area within 0.5mm of the wristband edge is avoided. The total planned path length is 150mm, which is 12% shorter than the initial path. The digital twin model simulates 5 weaving operations with an initial error perturbation of ±0.04mm. There are no collisions, and the final path is determined. Adaptive weaving: Initial parameters are set as follows: weaving speed for polyester fiber is 180 mm / s, weaving force of pressure sensor is 0.8 N, and weaving force of acceleration sensor is 0.7 N. Real-time monitoring and adjustment: The resistance monitoring device shows that the initial connection resistance of the pressure sensor is 75Ω and the acceleration sensor is 85Ω. When the fabric is woven into the middle section of the wristband, the displacement sensor detects that the fabric has shifted by 0.12mm. The AI ​​system immediately corrects the position and reduces the weaving speed to 144mm / s. After 3 seconds, the normal parameters are restored. Post-processing and performance optimization: Hot pressing treatment: set temperature 125℃, suitable for polyester fiber, pressure 0.5MPa, hot pressing time 3.5min, device size 6mm×6mm, larger than 5mm×5mm; Performance optimization: The environmental sensor acquires temperatures ranging from -5℃ to 45℃ with an accuracy of ±0.1℃, and humidity data from 20%RH to 80%RH with an accuracy of ±2%RH. When an acceleration greater than 2g is detected, it is considered a violent motion, and the LSTM network outputs a 32-dimensional hidden state. Through formula calculate, It is a 32×1 matrix. =0.1, The sigmoid function is used to adjust the accelerometer sensitivity from the initial 100mV / g to 120mV / g. =1.2 times), to ensure accurate collection of motion data.

[0025] Example 2: Production of a smart medical temperature monitoring patch (cotton fiber textile); Process pretreatment: Medical-grade cotton fiber textiles with a thickness of 0.2mm were selected. The plasma cleaning parameters were: power 110W, argon flow rate 55sccm, and processing time 3min. The tensile strength was 6MPa and the elastic modulus was 1.8GPa. The temperature sensor is model TS300. Functional test: measurement error ≤ ±0.05℃ within the range of 35-42℃. After passing the test, it is coded as "T-001, embedding position: center area of ​​patch". AI adaptive localization and path planning: Real-time location acquisition: The visual recognition device marks the disordered texture of the cotton fibers, the laser rangefinder measures the surface error to be ±0.05mm, and the positioning error after multimodal fusion is ≤±0.03mm; Path planning: cost function The tensile strength of cotton fibers is 6 MPa, which is less than 10 MPa. The value was set to 1.3, and five digital twin simulations were performed. The initial error was ±0.05mm, which met the accuracy requirements. The path length was 80mm. Adaptive weaving: Initial parameters: weaving speed 130mm / s, force 0.6N; Real-time adjustment: The temperature sensor connection resistance is initially 100Ω. When it is embedded to the edge, the resistance rises to 160Ω. The AI ​​system increases the force by 0.1N for every 10Ω difference. After 3 seconds, the resistance drops to 120Ω and returns to normal. Post-processing and performance optimization: Hot pressing treatment: temperature 115℃, pressure 0.45MPa, processing time 2.5min, device size 4mm×4mm, less than 5mm×5mm; Performance optimization: When the ambient temperature is 25-37℃, the LSTM network predicts the body temperature monitoring requirements and adjusts the temperature sensor sensitivity using a formula. =1.1 times.

[0026] Comparative examples and existing textile fiber weaving processes; Process pretreatment: Conventional polyester fiber textiles were selected, and ordinary cleaning agents were used for surface wiping and cleaning. Plasma cleaning technology was not used. The textiles were put directly into production after cleaning, and basic properties such as tensile strength and elastic modulus of the textiles were not tested. Ordinary pressure and temperature sensors are selected, and only simple on / off tests are performed to ensure that the devices can work normally. Detailed performance parameters such as sensor sensitivity and measurement accuracy are not tested, and the devices are not coded or identified. Location and route planning: The embedding positions are marked manually with a ruler and marker, and the approximate placement points of electronic components are determined by visual observation, with a positioning error of approximately ±0.8mm. The weaving sequence is manually planned according to the warp and weft directions of the fabric, and the weaving proceeds from the edge to the center. This does not take into account whether the path passes through weak areas of the fabric, nor does it have a dynamic adjustment mechanism. Inlay process: Using a traditional mechanical weaving machine, set fixed weaving parameters: weaving speed 140mm / s, weaving force 1.1N, and keep the parameters unchanged throughout the process; There is no real-time monitoring device during the weaving process. The connection status of the device is only visually inspected after weaving is completed. The connection resistance and position deviation are not detected. If the device is found to be obviously skewed, it must be manually removed and re-woven. Post-processing: After the weaving is completed, the textile is subjected to hot pressing treatment. It is placed in a hot press, and the temperature is set to 100℃ and the pressure to 0.3MPa. The hot pressing time is 5 minutes to enhance the bonding between the electronic devices and the textile. The electronic components operate with fixed parameters, the pressure sensor maintains an initial sensitivity of 0.3V / N, and the temperature sensor has a fixed measurement accuracy of ±0.3℃, and they operate continuously according to the set parameters.

[0027] Through two sets of examples and one set of comparative examples, it can be seen that the shortcomings of traditional processes can be solved by the intelligent and precise innovation of the entire process, which significantly improves the quality, performance and practicality of intelligent textile products, demonstrating obvious technical advantages and application value.

[0028] While the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the invention. Any variations and modifications can be made by those skilled in the art without departing from the spirit and scope of the invention. Therefore, any modifications, equivalent changes, and alterations made to the above embodiments based on the technical essence of the present invention, without departing from the scope of the invention, fall within the protection scope defined by the claims of the present invention.

Claims

1. A multimodal electronic device AI adaptive weaving process for smart textile fibers, including process preprocessing, AI adaptive positioning and path planning, adaptive weaving, and post-processing and performance optimization, wherein the AI ​​adaptive positioning and path planning includes real-time position acquisition, path planning algorithm, and digital twin-assisted path planning, characterized in that: The specific operation method of the AI ​​adaptive weaving of the multimodal electronic device is as follows: Step 1: After cleaning and treating the textile fibers, test their basic properties and check the status of multimodal electronic devices; Step 2: Use a high-precision visual recognition device to collect the texture feature points of the textile and the edge contours of the electronic devices, use an algorithm to plan the weaving path, and balance the path length according to the generation requirements. Step 3: Set initial weaving parameters, and monitor and adjust the parameters in real time; Step 4: After weaving is completed, the textile is subjected to hot pressing treatment. After the treatment is completed, the operating parameters of the device are adjusted according to environmental parameters to dynamically optimize the performance of the electronic device.

2. The AI ​​adaptive weaving process for multimodal electronic devices in smart textile fibers according to claim 1, characterized in that: In step one, plasma cleaning technology is used to treat the fabric for 3-5 minutes at a power of 100W-150W and an argon gas flow rate of 50sccm-80sccm. After treatment, the tensile strength and elastic modulus parameters of the textile are tested using mechanical property testing equipment. At the same time, the multimodal electronic devices are functionally tested, including the sensitivity of the pressure sensor and the measurement accuracy of the temperature sensor. After passing the test, the device is coded, and the coding information includes the device type, functional parameters and embedding position.

3. The AI ​​adaptive weaving process for multimodal electronic devices in smart textile fibers according to claim 2, characterized in that: In step two, real-time location acquisition includes enhanced visual recognition system and multimodal data fusion positioning. Enhanced visual recognition system uses a high-precision visual recognition device to acquire real-time images of textiles and multimodal electronic devices, including texture feature points and edge contours. At the same time, a semantic segmentation algorithm based on deep learning is introduced to semantically annotate the material distribution of textiles, fiber direction, and functional areas of electronic devices. Multimodal data fusion positioning is based on the concept of a multimodal large model of power artificial intelligence, which integrates sensor data to achieve accurate positioning. Laser rangefinders and ultrasonic sensors are added to verify and calibrate the location information acquired by visual recognition.

4. The AI ​​adaptive weaving process for multimodal electronic devices in smart textile fibers according to claim 3, characterized in that: In step two, the path planning algorithm includes A* algorithm optimization and multi-objective decision-making, path planning strategy, and path exploration using reinforcement learning. The cost function of the path planning algorithm... The risk coefficient is calculated based on the above, where Indicates the first inlay path Length of the path segment Indicates the first inlay path The risk coefficient corresponding to the segment path is assessed based on the performance differences of the textile and mechanical performance parameters. When planning the weaving path, a dynamic adjustment strategy is adopted to correct the weaving path in real time based on the real-time monitoring of the textile's movement status.

5. The AI ​​adaptive weaving process for multimodal electronic devices in smart textile fibers according to claim 4, characterized in that: The path exploration in the reinforcement learning introduces a reinforcement learning mechanism, which works in conjunction with the A* algorithm. The path planning process is regarded as an interaction between an agent and the environment. The agent learns the optimal path planning strategy by continuously trying different path choices and based on the reward signals from the environment.

6. The AI ​​adaptive weaving process for multimodal electronic devices in smart textile fibers according to claim 5, characterized in that: The digital twin-assisted path planning includes constructing a digital twin model and virtual path verification and optimization. The digital twin model uses digital twin technology to construct virtual models of textiles and multimodal electronic devices, modeling the physical properties of the textiles and the geometry and functional characteristics of the electronic devices, simulating the weaving process in a virtual environment. The model contains static information of the textiles and electronic devices, while reflecting the dynamic changes in the production process in real time. The virtual path verification and optimization involves simulating and verifying the planned path in the digital twin model before the actual weaving operation.

7. The AI ​​adaptive weaving process for multimodal electronic devices in smart textile fibers according to claim 6, characterized in that: In step three, initial weaving parameters are set according to the material of the textile and the type of electronic device, with a weaving speed of 100mm / s-200mm / s and a weaving force of 0.5N-1N.

8. The AI ​​adaptive weaving process for multimodal electronic devices in smart textile fibers according to claim 7, characterized in that: In step three, during the weaving process, the resistance between the electronic components and the textile is acquired in real time using a built-in resistance monitoring device. Using displacement sensors to detect device position deviation ,exist Out of range At that time, the AI ​​system processes according to the preset adjustment rules, where the LSTM network structure is input layer-hidden layer-output layer.

9. The AI ​​adaptive weaving process for multimodal electronic devices in smart textile fibers according to claim 8, characterized in that: In step four, after the weaving is completed, the textile is subjected to hot pressing treatment at a temperature of 110℃-130℃ and a pressure of 0.4MPa-0.6MPa for 2-4 minutes. The dynamic performance optimization involves the AI ​​system using a long short-term memory network to predict the impact of environmental changes on the performance of multimodal electronic devices based on environmental parameters (temperature and humidity) collected by environmental sensors and human motion data obtained by other sensors on the textile. Based on the prediction results, the system adjusts the device's operating parameters and the sensor sensitivity adjustment value. According to the formula The calculation shows that, among which The hidden state output by the LSTM network. and For weights and biases, This is the activation function.

Images