Improved piezoelectric nano generator based on two-dimensional NbOCl2 and action recognition system combined with deep learning technology

Abstract

The invention provides an improved piezoelectric nano generator based on two-dimensional NbOCl2 and an action recognition system combined with a deep learning technology, the improved piezoelectric nano generator based on the two-dimensional NbOCl2 comprises a flexible substrate, and a two-end type gold-indium asymmetric microstructure electrode is deposited on the flexible substrate. The gold-indium asymmetric microstructure electrode is in an orthogonally distributed array layout, an NbOCl2 nanosheet is arranged on the gold-indium asymmetric microstructure electrode and located between the gold electrode and the indium electrode, and the extending direction of the gold-indium asymmetric microstructure electrode is consistent with the long side direction of the NbOCl2 nanosheet. The flexible substrate, the gold-indium asymmetric microstructure electrode and the NbOCl2 nanosheet are all packaged by a PDMS material, and are used for solving the technical problems that motion recognition equipment is low in precision, inconvenient to carry, poor in flexibility and difficult in real-time monitoring, and particularly, strain field reconstruction on a complex surface is limited. By combining with a deep learning technology, different wrist joint motion states can be identified, and the method is used for motion monitoring.

Description

Technical Field

[0001] This invention relates to the technical field of piezoelectric nanogenerators, and more particularly to an action recognition system that combines piezoelectric nanogenerators with deep learning technology. Background Technology

[0002] With the booming development of the Internet of Things (IoT), wearable devices are becoming increasingly important in various fields. As a core component of wearable devices, improving the performance of wearable sensors is urgently needed. Currently, mainstream wearable sensors (such as piezoresistive and capacitive sensors) collect information by detecting changes in pressure or deformation. However, a problem remains: the information collected by a single device is local and one-dimensional, and its information density is insufficient to reflect the spatial distribution characteristics of the monitored signal. This greatly limits the development of new, multifunctional smart sensors.

[0003] Among numerous energy conversion technologies, the piezoelectric effect is widely used due to its advantages such as good stability and cleanliness. Compared with traditional piezoelectric bulk materials (such as PZT and ZnO), two-dimensional piezoelectric materials have atomic-level thickness and excellent mechanical flexibility, making them more suitable for the miniaturization and integration of devices. Due to their unique structural and performance advantages, two-dimensional piezoelectric materials are also suitable for integration into the human body; however, their relatively weak electrical output limits their application scenarios. Research on two-dimensional piezoelectric energy harvesters typically employs external field modulation methods such as applying ferroelectric polarization to improve output; however, this not only requires the material to possess ferroelectric polarization properties but also results in limited enhancement effects.

[0004] Currently, insufficient output performance remains the main factor limiting the functionality of 2D piezoelectric nanogenerators (PENGs). Composite film-based PENGs typically exhibit high output performance, with open-circuit voltages (VOCs) reaching tens of volts, so the interface state between the piezoelectric composite film and the electrodes has a negligible impact on output performance. However, for 2D PENGs with VOCs at the millivolt level, the built-in electric field at the metal / semiconductor interface has a significant impact on output performance. This indicates that both the piezoelectric polarization field inside the device and the built-in electric field at the interface significantly affect output performance; however, previous studies have often been conducted in isolation. Therefore, it is urgent to clarify the synergistic relationship between the piezoelectric polarization field inside the two-dimensional piezoelectric material and the built-in electric field at both interfaces to regulate the output performance.

[0005] On the other hand, the signals generated by a single device are localized and singular, and cannot be used to analyze the distribution characteristics of spatial signals. With the continuous development of micro-nano fabrication technology, integrated array designs of devices can achieve richer functions within a smaller size, meeting the demands of modern society for miniaturized and multifunctional equipment. 2D PENG integrated arrays, while reducing interference from external factors and improving stability and reliability, hold promise for developing unique sensing functions and providing more comprehensive and accurate spatial information.

[0006] Patent application CN202211040733.9 relates to a high and low temperature vibration testing device and its method for use in the bending state of flexible piezoelectric composite materials. The device includes a fixed base, a high and low temperature test chamber, an excitation system, a displacement sensing system, and an arc-shaped sample holder. The fixed base is fixed to the side of the high and low temperature test chamber and connected to the excitation system and the displacement sensing system. The excitation system and the displacement sensing system are located in the working chamber inside the high and low temperature test chamber. The arc-shaped sample holder is fixedly connected to the excitation system, and the displacement sensing system is positioned directly above the sample to be tested on the arc-shaped sample holder. This device can simultaneously perform vibration fatigue tests on one or more identical or different bending samples at fixed temperatures or alternating high and low temperature cycles, and at fixed or varying frequencies. It has advantages such as high efficiency, good repeatability, high reliability, and low testing cost, making a positive contribution to exploring the fatigue generation and failure causes of flexible piezoelectric composite materials in space environments. However, due to the high output performance of piezoelectric devices made of piezoelectric composite films, the influence of the interface state between the piezoelectric composite film and the electrode on the output performance is often neglected. In addition, due to the heavy and inflexible structure of the device, the invention patent with application number CN202211040733.9 cannot be used for body movement monitoring and motion recognition. Summary of the Invention

[0007] To address the limitations of existing motion recognition devices, such as low accuracy, inconvenience in portability, poor flexibility, and difficulty in real-time monitoring, especially in reconstructing strain fields on complex surfaces, this invention proposes an improved piezoelectric nanogenerator based on two-dimensional NbOCl2. This is further integrated with a wearable sensing array for regional motion recognition and strain field reconstruction. A performance-optimized wearable PENG array was fabricated using overlay technology, successfully detecting the piezoelectric response and strain field during wrist joint movement. Combined with deep learning, it can identify different wrist joint movement states and be used for motion monitoring.

[0008] To achieve the above objectives, the technical solution of the present invention is implemented as follows:

[0009] An improved piezoelectric nanogenerator based on two-dimensional NbOCl2 includes a flexible substrate on which two-end gold-indium asymmetric microstructure electrodes are deposited. The gold-indium asymmetric microstructure electrodes are arranged in an orthogonal array. NbOCl2 nanosheets are disposed on the gold-indium asymmetric microstructure electrodes between the gold and indium electrodes. The flexible substrate, the gold-indium asymmetric microstructure electrodes and the NbOCl2 nanosheets are all encapsulated by PDMS material.

[0010] Furthermore, the NbOCl2 nanosheets are synthesized from Nb powder, Nb2O5 powder and NbCl5 powder using a chemical vapor transport method to form NbOCl2 bulk material, which is then obtained by mechanical exfoliation.

[0011] Furthermore, the NbOCl2 nanosheets are rectangular in shape, and the extension direction of the gold-indium asymmetric microstructure electrode is consistent with the long side direction of the NbOCl2 nanosheets.

[0012] The thickness of the NbOCl2 nanosheets is 5-66 nm; the flexible substrate is a PET substrate.

[0013] Furthermore, the gold-indium asymmetric microstructure electrode is fabricated using an ultraviolet lithography machine. The fabrication method is as follows: first, photoresist is uniformly spin-coated on a flexible substrate, and then photolithography is performed using a mask to etch the gold electrode shape on the flexible substrate; then, a chromium adhesion layer and a gold conductive layer are sequentially deposited using a resistive evaporation coating machine and ion sputtering, and the photoresist is removed, so that the electrode array composed of gold and chromium is deposited on the flexible substrate; then, photoresist is uniformly spin-coated on the flexible substrate, and then photolithography is performed using a mask to etch the indium electrode shape on the flexible PET substrate; then, a chromium adhesion layer and an indium conductive layer are deposited using a resistive evaporation coating machine, and the photoresist is removed, so that the electrode array composed of indium and chromium is deposited on the flexible substrate.

[0014] Furthermore, the thickness of the chromium adhesion layer is 10-20 nm, the thickness of the gold conductive layer is 30-50 nm, and the thickness of the indium conductive layer is 30-50 nm.

[0015] Furthermore, the maximum piezoelectric polarization direction of NbOCl2 nanosheets was determined by second harmonic SHG testing, and the optimal electrode composition under the maximum piezoelectric polarization direction of NbOCl2 nanosheets was determined by stepper motor bending test.

[0016] Furthermore, a piezoelectric sensor is used to sense various wrist joint movements and obtain electrical signals of wrist joint motion.

[0017] Furthermore, the piezoelectric sensor is connected to a multi-channel nanogenerator analyzer for synchronous real-time evaluation of wrist joint motion. The application method is as follows: the piezoelectric sensor collects the electrical signals generated by wrist joint motion in real time and transmits them to the multi-channel nanogenerator analyzer. The multi-channel nanogenerator analyzer processes the electrical signals to filter out noise and calculates and statistically analyzes the wrist joint motion information using the noise-filtered signal. The strain at each point is obtained by sensitivity conversion of the electrical signals generated by wrist joint motion, and strain contour maps are plotted by interpolation to reconstruct the strain field and obtain the strain distribution map.

[0018] Furthermore, a deep learning-based wrist motion recognition system uses the aforementioned application method to acquire wrist joint motion signals with different bending amplitudes and reconstruct strain distribution maps. The reconstructed strain distribution maps are then input into a convolutional neural network for training to obtain a wrist motion recognition model. This model is then used to classify wrist joint motion signals with different bending amplitudes.

[0019] Furthermore, the convolutional neural network includes a first convolutional layer, an average pooling layer, a second convolutional layer, a max pooling layer, a flattening layer, a first fully connected layer, a second fully connected layer, and a softmax output layer connected in sequence. The first convolutional layer contains 32 convolutional kernels, and the second convolutional layer contains 16 convolutional kernels.

[0020] The beneficial effects of this invention are as follows:

[0021] 1. The improved two-dimensional NbOCl2 PENG of this invention has high piezoelectricity and can realize the piezoelectric effect of a single NbOCl2 nanosheet material. By introducing compressive strain into the device, a piezoelectric signal is generated. Under 4.56% compressive strain, its open circuit voltage (VOC) and peak-to-peak short circuit current (ISC) can reach up to 164.7mV and 8nA, respectively, which are much higher than those of ordinary two-dimensional PENG with symmetrical electrode structure.

[0022] 2. The improved two-dimensional NbOCl2 PENG of this invention has a small-area piezoelectric layer, making it suitable for integration. Furthermore, to realistically simulate the sensing capabilities of human skin, integrating a high-density sensor network onto the human body surface and ensuring its power supply is crucial. Compared to electronic skin, a two-dimensional NbOCl2 PENG, as a single piezoelectric sensing unit, often falls short in terms of detection area and accuracy. However, as the number of piezoelectric sensing units increases, richer strain characteristics and more accurate motion information can be obtained.

[0023] 3. The improved two-dimensional NbOCl2 PENG sensor array of this invention can monitor wrist joint movements at different locations on the skin of the human hand, and the wrist joint movement signals measured at different locations are significantly different.

[0024] 4. This invention integrates and analyzes multiple characteristic strain signals from a two-dimensional NbOCl2 PENG sensor array, which can be used to reconstruct the micro-stress field of the skin in that area.

[0025] 5. This invention uses a two-dimensional NbOCl2 PENG sensor array to accurately identify wrist joint postures. Combined with deep learning, since the array senses different characteristics in the electrical signals of different wrist joint movements, a large amount of electrical signal data can be obtained for deep learning training and recognition, which can be used for action recognition. Attached Figure Description

[0026] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0027] Figure 1 This is a schematic diagram of the growth of the NbOCl2 piezoelectric material used in this invention. Figure 1 a) and its morphology observed under a microscope ( Figure 1 b).

[0028] Figure 2 This is a schematic diagram of the piezoelectricity of the NbOCl2 nanosheets of the present invention characterized by pressure force microscopy (PFM); wherein, a is the relationship between applied voltage and PFM intensity, b is the relationship between the amplitude of the NbOCl2 nanosheet and voltage, c is the result of second-harmonic generation (SHG) test, and d is the output voltage of the PENG of the gold-gold symmetric electrode structure under compressive strain.

[0029] Figure 3 The diagram shows a schematic of the two-dimensional NbOCl2PENG structure and a schematic of the basic piezoelectric performance characterization of the present invention; wherein, a is a schematic of the two-dimensional NbOCl2PENG structure, b is an output voltage diagram, c is a stability test diagram, and d is a weak strain test diagram.

[0030] Figure 4This is a schematic diagram of a wearable device for wrist joint motion sensing after arraying operations according to the present invention; wherein, ab is a two-dimensional NbOCl2 PENG sensor array diagram, where b is the distribution coordinate diagram of each array component in the sensor array, c is the electrical output voltage of the array when strain is applied to the sensor array using a high-precision programmable linear stepper motor, d is a chord diagram characterizing the interference between sensor units, where the color of the chord represents the direction of influence and the chord width represents the magnitude of mutual influence between sensor units, e is a schematic diagram of applying strain to the sensor array, f is a strain distribution diagram of various parts of the array when strain is applied to the sensor array using a high-precision programmable linear stepper motor, g is a wearable experimental photograph of the sensor array attached to the wrist joint, h is the electrical output voltage of the array when wrist joint movement applies strain to the sensor array in the wearable experiment, and i is the strain distribution diagram of various parts of the sensor array when wrist joint movement applies strain to the sensor array in the wearable experiment.

[0031] Figure 5 This is a schematic diagram of wrist joint motion perception combined with deep learning in this invention for different ranges of motion; where a is a schematic diagram of wrist joint testing, b is a performance comparison diagram of two-dimensional NbOCl2 PENG sensing array and other two-dimensional piezoelectric devices, cd is the accuracy of training and testing and the training loss function curve, e is the confusion matrix of predicted labels and real labels in the test dataset, and f is the ROC curve. Detailed Implementation

[0032] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0033] Example 1

[0034] An improved piezoelectric nanogenerator based on two-dimensional NbOCl2, such as Figure 3As shown in Figure a, the device includes a flexible substrate that can be bent arbitrarily. Two-ended gold-indium asymmetric microstructure electrodes are deposited on the flexible substrate. The two-ended electrode structure facilitates connection between the electrodes and external circuits. The gold-indium asymmetric microstructure electrodes are arranged in an orthogonal array. This orthogonal array layout can effectively detect the strain response of each feature point in the plane, achieving accurate measurement of the strain distribution across the entire plane. This facilitates real-time monitoring of the strain distribution on the device surface under horizontal strain, and allows for the acquisition of key parameters of the strain field through feature extraction, enabling the construction of a high-precision motion sensing system. It is particularly suitable for strain sensing applications in wearable devices. The gold-indium asymmetric microstructure electrodes are located between the gold electrode and the indium electrode. NbOCl2 nanosheets are interspersed and deposited on the microelectrode to allow piezoelectric current to be discharged. The extension direction of the gold-indium asymmetric microstructure electrode is consistent with the long side direction of the nanosheet. When the extension direction of the gold-indium asymmetric microstructure electrode is consistent with the long side (b-axis, the direction of maximum piezoelectric polarization) of the NbOCl2 nanosheet, the maximum piezoelectric polarization field can be aligned with the direction of the built-in electric field in the same direction, and the strongest built-in electric field can be formed. Under compressive strain, the piezoelectric field and the built-in electric field are synergistically enhanced, significantly improving the piezoelectric output performance. The flexible substrate, the gold-indium asymmetric microstructure electrode and the NbOCl2 nanosheet are encapsulated by PDMS material to increase flexibility and adhesion.

[0035] The NbOCl2 nanosheets are synthesized from Nb powder, Nb2O5 powder and NbCl5 powder using chemical vapor transport (CVT) to form NbOCl2 bulk, and then obtained by mechanical exfoliation. The thickness of the NbOCl2 nanosheets is 5-66 nm. The flexible substrate is a PET substrate.

[0036] Specifically, in this embodiment, the preparation method of the NbOCl2 nanosheets is as follows: Nb powder, Nb2O5 powder, and NbCl5 powder are used as raw materials (ratio 1:3:6), which are placed in a quartz test tube, and the quartz test tube is pumped to 10 using a vacuum sealing machine. - 4 The tube was sealed, and then the quartz test tube containing the raw materials was placed in the center of the tube furnace, ensuring that the bottom of the test tube was in the reaction zone and the top of the test tube was in the growth zone. The furnace plugs were placed at both ends of the quartz tube in the tube furnace. The reaction zone was maintained at 500-600℃, and the growth zone at 300-400℃ for 5 days, then slowly cooled to room temperature at a rate of approximately 5℃ / h. Bulk NbOCl2 material was formed at the top of the test tube, and then two-dimensional NbOCl2 nanosheets were obtained by mechanical exfoliation with a precision of 5-66 nm. A schematic diagram of the piezoelectric material growth of NbOCl2 nanosheets is shown below. Figure 1 As shown in Figure a, the optical morphology of the NbOCl2 nanosheets is as follows. Figure 1 As shown in b. Figure 1 a indicates that the growth process has good controllability and stability. Figure 1 The material on surface b presents a regular rectangle, with the longer side of the rectangle being the b-axis.

[0037] To verify the piezoelectricity of the NbOCl2 nanosheets, this embodiment utilizes pressure microscopy (PFM) to characterize their piezoelectric properties. Specifically, the NbOCl2 nanosheets are placed on a PFM instrument. By applying a voltage to the PFM probe, the probe tip drives the NbOCl2 nanosheet sample to vibrate. As the voltage increases, the vibration amplitude of the NbOCl2 nanosheet sample gradually increases, thus obtaining the piezoelectric response of the NbOCl2 nanosheet sample. Figure 2 As shown in Figure a, the in-plane vertical PFM response increases with increasing applied voltage. All responses exhibit a unique reverberation peak at approximately 661.81 kHz. The color change of the amplitude image is almost linearly proportional to the applied tip voltage, indicating that the NbOCl2 nanosheets possess significant piezoelectric properties. The relationship between the amplitude and voltage of the NbOCl2 nanosheets is as follows: Figure 2 As shown in b. In-plane piezoelectric coefficient (d 11 The effective piezoelectric coefficient d can be calculated by the slope of the amplitude relative to the voltage after minimizing the electrostatic contribution. A robust effective piezoelectric coefficient d was obtained in NbOCl2 nanosheets. 11 With a piezoelectric coefficient of approximately 23±1 pm / V, NbOCl2 nanosheets can be fabricated into PENGs to output electrical signals.

[0038] The gold-indium asymmetric microstructure electrode was fabricated using an ultraviolet lithography process. The fabrication method involved: first, uniformly spin-coating photoresist onto a flexible substrate; then, using mask 1, photolithography was performed to etch the gold electrode shape onto the flexible substrate. Figure 4 (b) Each electrode is orthogonally distributed. A chromium adhesion layer and a gold conductive layer are sequentially deposited using a resistive evaporation coating machine and ion sputtering, followed by removal of the photoresist, resulting in a gold and chromium electrode array deposited on a flexible substrate. In this embodiment, chromium is deposited at 10 nm and gold at 50 nm. Photoresist is then uniformly spin-coated onto the flexible substrate, and photolithography is performed using mask 2 to etch the indium electrode shape onto the flexible PET substrate. A chromium adhesion layer and an indium conductive layer are then deposited using a resistive evaporation coating machine, followed by removal of the photoresist, resulting in an indium and chromium electrode array deposited on the flexible substrate. In this embodiment, chromium is deposited at 10 nm and indium at 50 nm, thus fabricating an asymmetric gold-indium asymmetric microstructure electrode. The asymmetric electrode can construct a bilateral, unidirectional built-in electric field, which, in conjunction with the piezoelectric polarization field, significantly improves the piezoelectric output performance.

[0039] After depositing gold and indium electrodes on a PET substrate, NbOCl2 nanosheets are transferred onto the electrodes. Finally, a layer of PDMS is spin-coated over the PET substrate, NbOCl2 nanosheets, and gold and indium electrodes to increase flexibility and adhesion, thus fabricating a flexible two-dimensional NbOCl2 piezoelectric nanogenerator. Figure 3 As shown in Figure a.

[0040] like Figure 2 As shown in Figure c, the maximum piezoelectric polarization direction of the NbOCl2 nanosheet was determined to be along the b-axis of the material, i.e., the long side of the nanosheet, through second harmonic generation (SHG) testing. When the maximum piezoelectric orientation is aligned with the built-in electric field direction of the NbOCl2 nanosheet, the device exhibits excellent electrical output at 4.56% compressive strain. Specifically, for the Au-Au symmetric electrode structure PENG, under the same compressive strain, the strongest polarization orientation outputs a peak-to-peak voltage of 65.8 mV, while the weakest polarization orientation outputs only 18.4 mV. Figure 2 As shown in d.

[0041] To further optimize device performance, various asymmetric NbOCl2 PENG electrodes with unidirectional built-in electric fields were designed and fabricated, aligning the strongest piezoelectric polarization orientation. Among them, the Au-In electrode PENG exhibited the strongest output due to its maximum work function difference with NbOCl2. Experimental results show that, under 4.56% compressive strain, the Au-In electrode structure device achieves a peak-to-peak output voltage of 164.7 mV and a peak-to-peak short-circuit current of 8 nA at the strongest polarization orientation. Figure 3 As shown in b, compared to the Au-Au symmetric electrode structure, the synergistic effect of polarization orientation and asymmetric structure resulted in an 8.95-fold enhancement. Through COMSOL simulation analysis, the performance differences of four metal electrodes with different work functions (Au, Cr, Zn, and In) were systematically studied. It was found that the Au-In electrode structure, due to its largest work function difference with NbOCl2, formed the strongest built-in electric field. The output performance test data of the fabricated asymmetric electrode device showed a high degree of agreement with the simulation results, verifying the reliability of the theoretical model. Therefore, this invention selected Au-In electrodes, and the Au-In asymmetric microstructure electrode was fabricated with its extension direction aligned with the long side direction of the nanosheet.

[0042] To characterize the basic piezoelectric properties of the two-dimensional NbOCl2 piezoelectric nanogenerator, this invention employs a stepper motor for bending tests. Since the piezoelectric effect of the two-dimensional NbOCl2 nanosheet generator is introduced by applying strain to the device through bending a flexible PET substrate, this embodiment uses a stepper motor to bend the flexible PET substrate, causing strain on the NbOCl2 nanosheets on the flexible substrate. This strain generates piezoelectric charges on the NbOCl2 nanosheets, thereby producing a piezoelectric signal.

[0043] NbOCl2 nanosheets generate an output voltage when subjected to pressure. The voltage signal change over time after being subjected to pressure can be recorded using a multi-channel nanogenerator analyzer. A two-dimensional NbOCl2 nanosheet generator is bent, and then the output voltage and current are obtained by connecting it to a multi-channel nanogenerator analyzer. Figure 3 As shown in b.

[0044] Among them, the device using gold-indium electrodes aligned with the long side (b-axis) of the nanosheet exhibits the highest electrical output under compressive strain. This phenomenon can be attributed to the strongest piezoelectric polarization orientation and the alignment of the bilaterally oriented built-in electric fields. To achieve the oriented built-in electric fields, electrode materials with specific work functions must be selected; that is, the work function of one electrode must be greater than NbOCl2, while the work function of the other electrode must be less than NbOCl2. The greater the difference in work function, the stronger the built-in electric field and the stronger the output voltage. Furthermore, the strain type can change the direction of the piezoelectric polarization field. Under compressive strain, the piezoelectric polarization field aligns with the built-in electric field of the Schottky junction, resulting in synergistic enhancement of the two fields. Conversely, under tensile strain, the two fields are in opposite directions, hindering each other and suppressing the electrical output.

[0045] In this embodiment, a stepper motor was used to repeatedly bend the two-dimensional NbOCl2 PENG to test its durability. After 20,000 cycles, the signal showed almost no attenuation, demonstrating good stability. Figure 3 As shown in c, applying extremely weak pressure to the device with a finger yields a tiny signal output, demonstrating that the two-dimensional NbOCl2 PENG is suitable for monitoring small signals generated by weak pressure, such as... Figure 3 The d-axis indicates that a preliminary experiment was conducted to test wrist joint movements of different amplitudes in the next step.

[0046] Example 2

[0047] A piezoelectric sensor is disclosed, comprising an improved piezoelectric nanogenerator based on two-dimensional NbOCl2, used to sense various wrist joint movements, obtain electrical signals of wrist joint motion, and achieve synchronous acquisition of multi-region motion signal characteristic parameters. In this embodiment, 16 two-dimensional NbOCl2 PENGs described in Embodiment 1 are integrated as sensing units to form a flexible sensing array of two-dimensional NbOCl2 PENGs as the piezoelectric sensor. The two-dimensional NbOCl2 PENGs have the characteristics of small size, light weight, and high flexibility, such as... Figure 4 As shown in Figure a, it is suitable for designing and manufacturing integrated arrays.

[0048] like Figure 4As shown in Figure b, the 16 sensing elements are arranged in a cross shape within four circles using a coordinate system with the array center as the origin, in order to analyze the output voltage of the 2.4-inch diameter two-dimensional NbOCl2 PENG sensing array. The black arrows indicate the strain direction applied to the two-dimensional NbOCl2 PENG sensing array. Figure 4 c shows the corresponding output voltage of the four-ring layout elements with strain applied using a stepper motor. The elements in the first ring have the strongest output because the four innermost devices experience the greatest strain when the two-dimensional PENG sensing array is bent.

[0049] The other structures and principles are the same as in Example 1.

[0050] Example 3

[0051] An application of a piezoelectric sensor as a wearable sensor is disclosed. The piezoelectric sensor is connected to a multi-channel nanogenerator analyzer. The piezoelectric sensor collects wrist joint motion signals, and the multi-channel nanogenerator analyzer processes the collected wrist joint motion signals, filters out noise, and uses the results for wrist joint motion information identification and statistics. Based on the different peak values ​​of the motion signals in the statistical information, a micro-strain field is reconstructed.

[0052] In this embodiment, a self-powered wrist joint motion monitoring system was constructed using the piezoelectric sensor and multi-channel nanogenerator analyzer for real-time motion identification and strain field reconstruction. This solves the problems that the information captured by a single device is local, one-dimensional, and its information density is insufficient to reflect the spatial signal distribution, thus promoting the diversified expansion of device functions.

[0053] Testing crosstalk between components within an integrated array is an indispensable part of integrated device design and manufacturing, directly affecting the product's performance ceiling, reliability, and market competitiveness, especially with the trend towards advanced processes and heterogeneous integration. Therefore, analyzing the crosstalk between components within a two-dimensional NbOCl2 PENG sensor array is crucial.

[0054] The method for testing crosstalk between components within an integrated array is as follows: a high-precision programmable linear stepper motor is used to apply strain to a two-dimensional NbOCl2 PENG sensor array. During operation, the output voltage of the target component in the two-dimensional NbOCl2 PENG sensor array is measured, and the result is plotted using a chord diagram. Figure 4 In d), the arc length corresponding to the solid-colored area represents the output amplitude. Subsequently, the output voltage of other components unaffected by strain is simultaneously measured, and the output amplitude is represented by the arc length corresponding to the target component's chord in the chord diagram. The above tests are repeated for all components in sequence. Finally, the experimental data is visualized and presented in the form of a chord diagram, as shown below. Figure 4As shown in d, the color of the string depends on the interaction between the two devices and is consistent with the color of the device that is more strongly affected (i.e., interfered with). In other words, when the target device has a greater impact on other devices, the arc length of the string of the target device is larger, and the string is mostly the color of the affected device. Compared with devices at the edge, devices at the center of the array are more easily affected by mechanical stimulation. Therefore, the arc length of the string of devices closer to the center is larger, and there are fewer strings of their own color. The experimental results presented by the string diagram show that the interference between devices is controlled at a low level. Based on the above analysis of interference, it can be seen that the sensing unit devices in this two-dimensional NbOCl2 PENG sensing array exhibit excellent anti-interference characteristics, and its performance optimization scheme can support higher integration density.

[0055] A schematic diagram of applying strain to a high-precision programmable linear stepper motor is shown below. Figure 4 As shown in Figure e, since the 16 sensing units are orthogonally arranged, when an external strain is applied in the in-plane direction, the two-dimensional NbOCl2 PENG sensing array can simultaneously detect the strain distribution at each location within the coverage area, thereby generating 16 electronic output signals. According to... Figure 4 The dependence of the open-circuit voltage on the applied strain in d can be calculated using sensitivity conversion to determine the strain detected by each of the 16 sensing units. ε i In response, V i S outputs voltage signals for each sensing unit. i The sensitivity coefficient in this embodiment is 32.65mV / %. Then, an interpolation algorithm is used to convert the discrete spatial strain values ​​into a continuous strain field distribution, ultimately constructing a strain distribution map, as shown below. Figure 4 As shown in f.

[0056] Figure 4 Image g shows a two-dimensional NbOCl2 PENG sensor array monitoring strain distribution during wrist movement. During testing, the sensor array was encapsulated using PDMS to ensure close contact between the component and the skin. Figure 4 h shows the electrical output amplitude of each element in the two-dimensional PENG sensor array. Figure 4 i shows the corresponding strain distribution. Figure 4 The results show that two-dimensional PENG sensor arrays have broad application prospects in the field of wearable sensors.

[0057] During wrist joint movement, the corresponding skin constantly contracts and relaxes. Factors such as sweat and varying degrees of fat on the hands can interfere with strain sensing. Furthermore, in practical applications such as human skin motion sensing and human-machine interfaces, most traditional strain sensors have significant limitations, such as the ability to detect minute deformations and the ability to detect only single-point strain, thus restricting their application in these scenarios. This invention uses an NbOCl2PENG array as a strain sensor attached to the wrist joint to sense this movement. Wrist flexion causes strain on the attached NbOCl2 PENG array, generating piezoelectric signals. These signals are displayed in real time using a multi-channel nanogenerator analyzer. This invention uses a multi-channel nanogenerator analyzer to collect electrical signals, clearly obtaining the electrical signals generated by the sensor's deformation due to wrist flexion. The signals are displayed in real time and stored as a DATA file on a computer for further analysis. Because the flexion speed and amplitude vary slightly each time, the strain on the skin surface is not uniform, resulting in slightly different peak values ​​in the piezoelectric signal output from a single motion signal. Furthermore, micro-strain field reconstruction can be achieved.

[0058] The other structures and principles are the same as in Example 2.

[0059] Example 4

[0060] A deep learning-based wrist motion recognition system includes the piezoelectric sensor and the multi-channel nanogenerator analyzer. The system is characterized by using the piezoelectric sensor as a wearable sensor to acquire wrist joint motion signals with different bending amplitudes and reconstructing strain distribution maps. The reconstructed strain distribution maps are then input into a convolutional neural network for training to obtain a wrist motion recognition model. This model is then used to classify wrist joint motion signals with different bending amplitudes.

[0061] It is worth noting that, such as Figure 5 As shown in b, compared to previous reports, the performance-enhanced 2D NbOCl2 PENG sensor array exhibits significant advantages in integration, robustness, active region, and piezoelectric coefficient. However, sensitivity may be limited by micro / nano fabrication processes. Compared to traditional data processing techniques, deep learning can classify and accurately identify subtle features in large amounts of data. The 2D NbOCl2 PENG sensor array can simultaneously acquire a large number of electrical signals with classifiable features, thus demonstrating feasibility in combining it with deep learning techniques and potentially compensating for any potential sensitivity limitations.

[0062] In this embodiment, during the acquisition of wrist joint motion states using a two-dimensional NbOCl2 PENG sensor array, the wrist joint flexes within a range of 0-70 degrees in 10-degree increments, resulting in eight motion states. The wrist joint motion signals acquired at different flexion amplitudes are reconstructed into strain distribution maps using a multi-channel nanogenerator analyzer and extracted into DATA files. These are then integrated into an Excel file. Subsequently, all electronic output signals measured at different wrist flexion degrees are divided in a 6:4 ratio and used to construct training and testing sets, respectively. These are input into a convolutional neural network (CNN) to construct a wrist motion recognition model. The model is trained 100 times using the training set to learn the characteristics of different wrist joint motion patterns. The model performance is then evaluated using the testing set, achieving a final recognition rate of 98.44%.

[0063] The Convolutional Neural Network (CNN) model architecture consists of two convolutional layers (Conv 1D), containing 32 and 16 convolutional kernels respectively. Each convolutional layer is followed by a pooling layer to reduce feature dimensionality, decrease computational complexity, and enhance model robustness. The pooling layers have a stride of 2; the first pooling layer is an Average Pooling 1D, and the second is a Max Pooling 1D. A flattening layer is then applied to convert the multidimensional features into a one-dimensional vector, followed by classification through two fully connected (Dense) layers. The output layer uses the Softmax activation function.

[0064] Figure 5 c and 5d show the changes in accuracy and loss function over 100 iterations, indicating that the model converges rapidly and demonstrates good signal resolution. Figure 5 The confusion matrix between predicted and actual values ​​in e shows a prediction accuracy as high as 98.44%. The area under the receiver operating characteristic (ROC) curve (AUC) can intuitively evaluate the discriminative power of the CNN model; the larger the area, the stronger the model's discriminative power. From... Figure 5 As can be seen from f, the AUC is equal to 1 for each wrist flexion, which indicates that the CNN model designed for this training data has excellent discrimination ability, and also reflects that the data collected by the device under different conditions has high specificity.

[0065] This invention utilizes the large area and multi-feature data characteristics of a two-dimensional NbOCl2 PENG sensor array, achieving high precision. The device can be attached to the human wrist joint to sense wrist joint motion signals, and it can identify that wrist joint motion signals differ at different angles, ultimately enabling high-precision wrist joint motion recognition. This invention lays the foundation for the application of two-dimensional NbOCl2 PENG sensor arrays in the field of electronic skin.

[0066] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. An improved piezoelectric nanogenerator based on two-dimensional NbOCl2, comprising a flexible substrate, characterized in that, The flexible substrate has a two-ended gold-indium asymmetric microstructure electrode deposited on it. The gold-indium asymmetric microstructure electrode is arranged in an orthogonal array. NbOCl2 nanosheets are placed on the gold-indium asymmetric microstructure electrode between the gold electrode and the indium electrode. The flexible substrate, the gold-indium asymmetric microstructure electrode and the NbOCl2 nanosheets are all encapsulated by PDMS material.

2. The improved piezoelectric nanogenerator based on two-dimensional NbOCl2 according to claim 1, characterized in that, The NbOCl2 nanosheets are synthesized from Nb powder, Nb2O5 powder and NbCl5 powder using chemical vapor transport method to form NbOCl2 bulk material, which is then obtained by mechanical exfoliation.

3. The improved piezoelectric nanogenerator based on two-dimensional NbOCl2 according to claim 2, characterized in that, The aforementioned The NbOCl2 nanosheets are rectangular in shape, and the extension direction of the gold-indium asymmetric microstructure electrode is consistent with the long side direction of the NbOCl2 nanosheets. The thickness of the NbOCl2 nanosheets is 5-66 nm; the flexible substrate is a PET substrate.

4. The improved piezoelectric nanogenerator based on two-dimensional NbOCl2 according to claim 2, characterized in that, The gold-indium asymmetric microstructure electrode is fabricated using an ultraviolet lithography process. The fabrication method is as follows: First, photoresist is uniformly spin-coated on a flexible substrate. Then, photolithography is performed using a mask to etch the gold electrode shape on the flexible substrate. Next, a chromium adhesion layer and a gold conductive layer are sequentially deposited using a resistance evaporation coating machine and ion sputtering, and then the photoresist is removed, so that the electrode array composed of gold and chromium is deposited on the flexible substrate. Then, photoresist is uniformly spin-coated on the flexible substrate. Then, photolithography is performed using a mask to etch the indium electrode shape on the flexible PET substrate. Finally, a chromium adhesion layer and an indium conductive layer are deposited using a resistance evaporation coating machine, and then the photoresist is removed, so that the electrode array composed of indium and chromium is deposited on the flexible substrate.

5. The improved piezoelectric nanogenerator based on two-dimensional NbOCl2 according to claim 4, characterized in that, The chromium adhesion layer has a thickness of 10-20 nm, the gold conductive layer has a thickness of 30-50 nm, and the indium conductive layer has a thickness of 30-50 nm.

6. The improved piezoelectric nanogenerator based on two-dimensional NbOCl2 according to claim 1, characterized in that, The maximum piezoelectric polarization direction of NbOCl2 nanosheets was determined by second harmonic SHG testing, and the optimal electrode composition under the maximum piezoelectric polarization direction of NbOCl2 nanosheets was determined by stepper motor bending test.

7. A piezoelectric sensor, characterized in that, The piezoelectric sensor is composed of multiple improved piezoelectric nanogenerators based on two-dimensional NbOCl2 as described in any one of claims 1-6, and is used to sense various wrist joint movements and obtain electrical signals of wrist joint movement.

8. The application of the piezoelectric sensor according to claim 7 as a wearable sensor, characterized in that, The piezoelectric sensor is connected to a multi-channel nanogenerator analyzer for synchronous real-time assessment of wrist joint motion. The application method is as follows: the piezoelectric sensor collects the electrical signals generated by wrist joint motion in real time and transmits them to the multi-channel nanogenerator analyzer. The multi-channel nanogenerator analyzer processes the electrical signals to filter out noise and calculates and statistically analyzes the wrist joint motion information using the noise-filtered signal. The strain at each point is obtained by sensitivity conversion of the electrical signals generated by wrist joint motion, and strain contour maps are plotted by interpolation to reconstruct the strain field and obtain the strain distribution map.

9. A deep learning-based wrist motion recognition system, comprising the piezoelectric sensor and multi-channel nanogenerator analyzer as described in claim 8, characterized in that, The wrist joint motion signals with different bending amplitudes are obtained by using the application method described in claim 8 and strain distribution maps are reconstructed. The reconstructed strain distribution maps are input into a convolutional neural network for training to obtain a wrist motion recognition model. The wrist motion recognition model is used to classify wrist joint motion signals with different bending amplitudes.

10. The wrist motion recognition system based on deep learning according to claim 9, characterized in that, The convolutional neural network includes a first convolutional layer, an average pooling layer, a second convolutional layer, a max pooling layer, a flattening layer, a first fully connected layer, a second fully connected layer, and a softmax output layer connected in sequence. The first convolutional layer contains 32 convolutional kernels, and the second convolutional layer contains 16 convolutional kernels.

Images