Capacitive three-dimensional force sensor array, three-dimensional force data set acquisition system and data acquisition and information decoupling method

Abstract

The invention discloses a capacitive three-dimensional force sensing array, a three-dimensional force data set acquisition system and a data acquisition and information decoupling method, and the capacitive three-dimensional force sensing array is characterized in that a contact layer transmits an external force to an electrode layer and a dielectric layer under the action of the external force, so that the capacitance value of a capacitive pressure sensing unit is changed; besides, uniform contact / deformation is generated between each electrode layer in matrix arrangement and the middle dielectric layer due to normal pressure applied to the contact layer, and differential contact / deformation is generated between the electrode layers and the middle dielectric layer due to shear force applied to the contact layer, so that three-dimensional force is represented. A three-dimensional force data set acquisition system composed of a capacitive three-dimensional force sensing array is used for acquisition of a three-dimensional force data set, and an adopted data acquisition and information decoupling method is an information decoupling method based on a random forest and realizes application verification of three-dimensional force real-time decoupling.

Description

Technical Field

[0001] The present invention relates to the technical field of sensors, and in particular to a capacitive three-dimensional force sensing array, a three-dimensional force data set acquisition system, and a data acquisition and information decoupling method. Background Art

[0002] At present, sensors based on the capacitance principle have been widely used in pressure sensing and acceleration sensing, and it has been proposed to combine such a capacitively pressure sensor arranged in a matrix with other mechanical structures for three-dimensional force sensing. Specifically: a contact structure is placed above a capacitively pressure sensing unit arranged in a certain regular pattern, and the three-dimensional force acting on the contact structure causes the contact to deform, generating pressure and torsional moment, so that each of the capacitively pressure sensing units below has a differential output, and the three-dimensional force information is decoupled by methods such as mathematical modeling and machine learning.

[0003] With the development of fields such as robot intelligence, electronic skin, and wearable health detection devices, higher requirements are put forward for the sensing of pressure and touch, and an arrayed three-dimensional force sensor deployment is required. However, the current three-dimensional force sensors cannot meet the requirements of the above fields for large-area sensing, high-speed data acquisition, and three-dimensional force information decoupling. Summary of the Invention

[0004] The technical problem to be solved by the present invention is to provide a capacitive three-dimensional force sensing array.

[0005] Another technical problem to be solved by the present invention is to provide a three-dimensional force data set acquisition system applying the above capacitive three-dimensional force sensing array.

[0006] Another technical problem to be solved by the present invention is to provide a data acquisition and information decoupling method for the above capacitive three-dimensional force sensing array and three-dimensional force data set acquisition system.

[0007] To solve the above technical problems, the technical solutions proposed by the present invention are as follows:

[0008] A capacitive three-dimensional force sensing array, which successively includes a contact layer, an upper electrode layer, a dielectric layer, and a lower electrode layer from top to bottom. Among them, the contact layer is made of a material with complete deformation recovery ability, the upper electrode layer and the lower electrode layer are both composed of an electrode and a substrate material, the electrode is fixed on the substrate material, the electrode includes electrode units and wires connecting each electrode unit, the dielectric layer is made of sheet dielectric material. Among them, the top electrode and the bottom electrode and the dielectric material therein form a capacitive sensing unit, a plurality of top electrodes are arranged and connected to form the upper electrode layer, a plurality of bottom electrodes are arranged and connected to form the lower electrode layer, and each sensing unit in the sensing array includes four capacitive sensing units arranged in a 2×2 matrix and a top contact, and is extended through the capacitive sensing unit.

[0009] Preferably, for the above capacitive three-dimensional force sensing array, the material of the contact layer is PDMS or Ecoflex material.

[0010] Preferably, for the above capacitive three-dimensional force sensing array, the dielectric layer is an ionic gel made of polyvinyl alcohol and phosphoric acid or a porous PDMS elastomer doped with barium titanate.

[0011] Preferably, for the above capacitive three-dimensional force sensing array, the upper electrode layer and / or the lower electrode layer is a gold electrode deposited on a PET film by magnetron sputtering, where the shape of the electrode unit is square, and the row and column electrode units are connected by deposited gold wires; or a nano-silver electrode is formed on a polyimide film by screen printing, where the shape of the electrode unit is square, and the row and column electrode units are connected by spiral lines.

[0012] Preferably, for the above capacitive three-dimensional force sensing array, it is extended through the capacitive sensing unit. For an N×M point capacitive three-dimensional force sensing array, it includes 2N×2M capacitive sensors and N×M top contacts.

[0013] Preferably, for the above capacitive three-dimensional force sensing array, in each of the capacitive sensing units, both the upper electrode layer and the lower electrode layer are 4 electrode units arranged in a 2×2 matrix form, and the electrode units in each column of the upper electrode layer are connected, and the electrode units in each row of the lower electrode layer are connected.

[0014] Preferably, for the above capacitive three-dimensional force sensing array, among the top electrodes of the 2N×2M capacitive sensors, the electrode units in each column are interconnected to form a 2M-column 1×N electrode arrangement; among the bottom electrodes, the electrode units in each row are interconnected to form a 2N-row 1×M electrode arrangement.

[0015] The shape of the above electrode units can be any machinable geometric pattern, and the electrode units can be connected by straight wires or spiral lines.

[0016] Preferably, for the above capacitive three-dimensional force sensing array, the top contact can be a frustum structure, a cuboid structure or a hemispherical structure.

[0017] A three-dimensional force data set acquisition system includes the above capacitive three-dimensional force sensing array, a multi-channel capacitance acquisition circuit, and a dual-serial-port acquisition system developed based on LabVIEW; the multi-channel capacitance acquisition circuit includes a multiplexing circuit, a capacitance digital conversion circuit, a micro-control unit MCU and a 485 communication circuit, where,

[0018] The multiplexing circuit includes two multiplexing chips (CD74HC4051). The input pins of the multiplexing chips are respectively connected to the output ends of the upper electrode layer and the lower electrode layer of the capacitive three-dimensional force sensing array. Specifically: two output pins of the multiplexing chips are connected to the corresponding two input pins of the capacitance digital conversion circuit; the multiplexing chips select different capacitance sensor signals through a polling method and sequentially transmit the output signals of the sensors to the capacitance digital conversion circuit;

[0019] The capacitance digital conversion circuit includes a capacitance digital conversion chip (FDC2214) and an LC oscillation circuit. The capacitance digital conversion chip is connected to the LC oscillation circuit by a circuit. The LC oscillation circuit is connected to the multiplexing chip by a circuit. The capacitance digital conversion chip is connected to the microcontroller unit MCU by a circuit. Specifically: both ends of the inductor L1 of the LC oscillation circuit are connected to the input pins INOA and INOB of the capacitance digital conversion chip (FDC2214). Both ends of the inductor L1 are respectively connected to the output pins of two multiplexing chips. The input of the multiplexing chip is connected to the capacitance sensing array. In this way, the inductor L1 and the capacitance sensing array form an LC oscillation circuit. The LC oscillation circuit is connected in parallel to INOA and INOB of the capacitance digital conversion chip (FDC2214); one output pin of the capacitance digital conversion chip and two I2C communication pins are respectively connected to the microcontroller unit MCU (the two I2C communication pins are used to establish a connection with the MCU to ensure the security and accuracy of data transmission), ensuring the correct transmission and processing of data, and converting the capacitance signal of the capacitive three-dimensional force sensing array into a digital signal; the microcontroller unit MCU is used to control the polling time and period of the multiplexing circuit, control the configuration of the capacitance digital conversion circuit, process and calculate the capacitance digital signal, and control the signal transmission of the 485 communication circuit;

[0020] The 485 communication circuit is connected to the microcontroller unit MCU by a circuit and performs data interaction with the host computer using the RS485 communication protocol, and is used to transmit the digital signal to the host computer through the 485 communication protocol for further processing or display.

[0021] The above-mentioned microcontroller unit MCU includes: a main control chip STM32F407, which is used to control the operation and data processing of the circuit; a crystal oscillator circuit, which is used to provide a stable clock signal to ensure the accuracy of the system timing; a reset circuit, which is used to reset the MCU when the system starts or a fault occurs to ensure the stable operation of the system; a spare serial port communication circuit, which is used to provide an additional data communication interface to support backup data transmission or debugging.

[0022] In the above three-dimensional force data set acquisition system, multi-channel measurement can be achieved through a multi-channel capacitance acquisition circuit. The capacitance sensing array is connected in series with a small capacitor to achieve wide-range measurement (500 nF). Among them, the large range is achieved by connecting a small capacitor in series with the capacitance sensing array in the LC oscillation circuit. Specifically, a small capacitor C1 is connected in series on the output pin Outl of the multiplexing chip. In this way, the C in the LC oscillation circuit is the total measured capacitance Ctotal = C1 * C2 / (C1 + C2), and then the capacitance value C2 of the actual capacitance sensing array is obtained through subsequent calculations. The realization of multi-channel measurement is achieved by connecting two multiplexing chips (CD74HC4051) in series between the capacitance sensing array and the capacitance digital conversion circuit. The input pins of the two multiplexing chips are respectively connected to the upper and lower plates of the capacitance sensing array. The output of the multiplexing chip is connected to Out1 and Out2 of the capacitance digital conversion circuit. The MCU is connected to the control pins SO, S1, and S2 of the multiplexing chip. The MCU controls the input pins of the multiplexing chip to conduct sequentially through polling. The capacitance digital conversion circuit measures the individual capacitance values of the capacitance sensing array in turn. The capacitance values of each row of the capacitance sensing array form a data packet, and the individual capacitance values of the capacitance sensing array are obtained by parsing the subsequent data packet.

[0023] Preferably, in the above three-dimensional force data set acquisition system, the capacitance digital conversion circuit uses the principle of an LC oscillation circuit for capacitance measurement. The oscillation frequency is inversely proportional to the capacitance value. The capacitance digital conversion chip outputs a signal with a frequency of 10 KHz to 10 MHz, and the resonance frequency of the LC circuit is detected. The change in the resonance frequency reflects the change in capacitance.

[0024] The oscillation frequency is inversely proportional to the capacitance value to be measured, as shown in Equation (1):

[0025]

[0026] Preferably, in the above three-dimensional force data set acquisition system, the dual-serial-port acquisition system developed based on LabVIEW has two serial-port resources. Among them, serial port 1 is used to receive data from the multi-channel capacitance acquisition circuit based on multiplexing, and serial port 2 is used to receive the output of a commercial standard three-dimensional force sensor.

[0027] Preferably, in the above three-dimensional force data set acquisition system, the serial-port resources are accessed through the LabVIEW program. The capacitance measurement values from the multi-channel capacitance acquisition circuit based on multiplexing are received in real time using serial port 1. At the same time, the three-dimensional force data of the commercial three-dimensional force sensor is received using serial port 2. A fixed field is sent to the commercial three-dimensional force sensor through the LabVIEW system to request three-dimensional force measurement data, and the capacitance measurement values and three-dimensional force data are written into a file for storage using the data writing module.

[0028] The above-mentioned commercial three-dimensional force sensor is used as a standard three-dimensional force sensor, and its output is used to calibrate the capacitive three-dimensional force sensing array with the standard value.

[0029] In the above-mentioned three-dimensional force dataset acquisition system, Serial Port 1 receives data from the multiplexing-based multi-channel capacitance acquisition circuit in real time, intercepts characters and verifies the correctness of the data frame. Subsequently, data bits are processed according to the pre-confirmed communication protocol, converted into single-precision floating-point numbers, and continuously displayed in the form of numbers and waveforms. Finally, the displayed data is saved in a txt file at a fixed position; Serial Port 2 is used to receive the output from the commercial three-dimensional force sensor. According to the settings of the commercial three-dimensional force sensor, the serial port first sends a request data field to it, and then processes the data bits according to the pre-confirmed communication protocol, converts them into single-precision floating-point numbers, and continuously displays them in the form of numbers and waveforms. Finally, the displayed data is saved in a txt file at a fixed position.

[0030] Preferably, in the above-mentioned three-dimensional force dataset acquisition system, the capacitance measurement value (data from Serial Port 1) and the real-time three-dimensional force data (data from Serial Port 2) are written into the same file at the same time. The two parts share a time counter (that is, the time counter can also be stored in the same file), and LabVIEW allocates resources by itself according to the system resources, matches the capacitance measurement value, the real-time three-dimensional force data, and the system time to form a three-dimensional force dataset.

[0031] Preferably, in the above-mentioned three-dimensional force dataset acquisition system, a three-dimensional force information decoupling module is added to the LabVIEW program for real-time information decoupling: after processing the data received by Serial Port 1, it is input into the script program box of LabVIEW, and a pre-set three-dimensional force information decoupling algorithm is called, and the output is connected to a suitable display control to display the decoupled three-dimensional force information.

[0032] The above-mentioned three-dimensional force information decoupling module includes the code inside the Matlab script node, as well as the input and output variables connected to it, and also includes other display controls connected to the output variables.

[0033] A data acquisition and information decoupling method using the above-mentioned three-dimensional force dataset acquisition system is as follows:

[0034] (1) Data sorting: Sort the responses of the pre-acquired flexible three-dimensional force sensing array and the corresponding real three-dimensional force values to obtain a three-dimensional force dataset;

[0035] (2) Data preprocessing: Remove the outliers in the three-dimensional force dataset and replace them with the average value of the two adjacent values.

[0036] (3) Establish a model for prediction. Using the random forest method in machine learning, divide the data set obtained in step (3) to establish a regression prediction model for the output of the flexible three-dimensional force sensing array and the three-dimensional force; input the data set into the three-dimensional force regression prediction model for data training, testing, and / or prediction.

[0037] Preferably, in the above data acquisition and information decoupling method, in step (1), Origin software is used to screen the required data.

[0038] Preferably, in the above data acquisition and information decoupling method, the data preprocessing in step (2) includes the following steps:

[0039] For a certain flexible three-dimensional force sensing unit in the flexible three-dimensional force sensing array, its output variables include <C1, C2, C3, C4>; the three-dimensional force corresponding to each group of outputs is <F x , F y , F z >. Perform min-max normalization on the above 7 variables. The calculation method is shown in Equation (2):

[0040]

[0041] Where X represents any variable in <C1, C2, C3, C4> or <F x , F y , F z >, X max and X min represent the maximum and minimum values of any variable respectively, and X normalized is the value after normalization.

[0042] Preferably, in the above data acquisition and information decoupling method, after step (2), feature engineering is further included: According to the working principle of the flexible three-dimensional force sensing array, further deduce the relationship between its output and the three-dimensional force to enrich the feature relationship between the input and output; improve the prediction accuracy of the model by adding additional features.

[0043] Preferably, in the above data acquisition and information decoupling method, the feature engineering includes the following steps:

[0044] (I) Deduce the further relationship between the vector <C1, C2, C3, C4> and <F x , F y , F z > by using methods including but not limited to theoretical analysis and mathematical modeling and solution;

[0045] (II) Deduce the quantitative relationship between C1C2C3C4 and F x , Fy , F z The quantitative relationship between them.

[0046] Preferably, in the above data acquisition and information decoupling method, the machine learning method in step (3) is Random Forest (RF), and the dataset obtained by the feature engineering is divided into a training dataset and a test dataset according to a ratio of 7:3.

[0047] Preferably, in the above data acquisition and information decoupling method, the parameters for evaluating the prediction ability of the quantitative regression prediction model in step (3) are the overall mean square error MSE, and for the three-dimensional force F x , F y , F z the mean square error MSE_F x , MSE_F y , MSE_F z , and the distribution is shown in formulas (3)-(4):

[0048]

[0049] where Y i is the actual value, is the predicted value, and N represents the number of samples;

[0050] where the value of ω is x, y, or z. F ω represents the actual value, represents the predicted value;

[0051] where MSE and MSE_F ω are error metrics, and the smaller the value, the smaller the prediction error and the higher the model stability;

[0052] The error metrics MSE and MSE_F are introduced ω , MSE is a measure of the error of the overall prediction, while MSE_F ω is used to evaluate the prediction performance for a certain one-dimensional force information.

[0053] Preferably, in the above data acquisition and information decoupling method, the prediction in step (3) includes inputting the capacitance response value in the three-dimensional force dataset into the Random Forest program for prediction to obtain the corresponding three-dimensional force information F x , F y , F z .

[0054] Technical effects

[0055] In the above capacitive three-dimensional force sensing array, under the action of an external force, the contact layer transfers the external force to the electrode layer and the dielectric layer, causing the capacitance value of the capacitive pressure sensing unit to change; in addition, the normal pressure applied to the contact layer causes uniform contact / deformation between each electrode layer arranged in a matrix and the intermediate dielectric layer, while the shear force applied to the contact layer causes differential contact / deformation between the electrode layer and the intermediate dielectric layer, thereby characterizing the three-dimensional force. The three-dimensional force data set acquisition system composed of the capacitive three-dimensional force sensing array is used for the acquisition of the three-dimensional force data set. The data acquisition and information decoupling method adopted is an information decoupling method based on random forest, and the application verification of real-time decoupling of the three-dimensional force is realized.

[0056] In the three-dimensional force data set acquisition system, the multi-channel capacitance acquisition circuit based on multiplexing can realize multi-channel measurement. The capacitance sensing array is connected in series with a small capacitance to realize wide-range measurement (500 nF). Among them, the large range is realized by connecting a small capacitance in series with the capacitance sensing array in the LC oscillation circuit. The three-dimensional force data set acquisition system realizes the one-to-one correspondence between the standard output of the commercial three-dimensional force sensor and the capacitive three-dimensional force sensor, can directly collect the data set for machine learning, simplifies the construction process of the data set through the optimization of the algorithm, can realize continuous process calibration, and has an outstanding prediction effect. Brief Description of the Drawings

[0057] Figure 1 It is a schematic structural diagram of the capacitive three-dimensional force sensing array.

[0058] Figure 2 It is a working flowchart of the multi-channel capacitance acquisition circuit based on multiplexing.

[0059] Figure 3 It is a working flowchart of the dual-serial-port acquisition system developed based on LabVIEW.

[0060] Figure 4 It is a working flowchart of the three-dimensional force decoupling algorithm.

[0061] Figure 5 It is a prediction result diagram of the random forest decoupling algorithm.

[0062] Figure 6 It is a circuit diagram of the capacitance digital conversion circuit. Detailed Description of the Invention

[0063] The following combines the embodiments and the drawings to make a detailed description of the capacitive three-dimensional force sensing array, the three-dimensional force data set acquisition system, and the data acquisition and information decoupling method of the present invention.

[0064] Embodiment 1

[0065] The capacitive three-dimensional force sensing array is constructed as follows Figure 1 As shown, its structure from top to bottom is: a contact layer, an upper electrode layer, a dielectric layer, and a lower electrode layer. Among them, the contact layer is made of a material with complete deformation recovery ability (PDMS material), the upper electrode layer and the lower electrode layer are both composed of an electrode and a substrate material, the electrode is fixed on the substrate material, the electrode includes electrode units and wires connecting each electrode unit, the dielectric layer is made of a sheet-like dielectric material (an ion gel made of polyvinyl alcohol and phosphoric acid), where the top electrode and the bottom electrode and the dielectric material therein form a capacitive sensing unit, multiple top electrodes are arranged and connected to form the upper electrode layer, multiple bottom electrodes are arranged and connected to form the lower electrode layer, each sensing unit in the sensing array includes four capacitive sensing units arranged in a 2×2 matrix and a top contact, each column of electrode units in the upper electrode layer is connected, each row of electrode units in the lower electrode layer is connected, and it is extended through the capacitive sensing unit. For an N×M point capacitive three-dimensional force sensing array, it includes 2N×2M capacitive sensors and N×M top contacts. Among the top electrodes of the 2N×2M capacitive sensors, each column of electrode units is interconnected to form a 2M-column 1×N electrode arrangement; among the bottom electrodes, each row of electrode units is interconnected to form a 2N-row 1×M electrode arrangement.

[0066] In this embodiment, the PDMS material is used to prepare the contact layer, and a gold electrode is deposited on the PET film by magnetron sputtering. The shape of the electrode unit is square, and the row and column electrode units are connected by deposited gold wires. The dielectric layer uses an ion gel made of polyvinyl alcohol and phosphoric acid (the molecular weight Mw of polyvinyl alcohol is 145000, the analytical purity of phosphoric acid is 85%, and the ratio of polyvinyl alcohol to phosphoric acid is 1:1.2).

[0067] Example 2

[0068] The capacitive three-dimensional force sensing array has the same structure as in Example 1, except that:

[0069] The Ecoflex material is used to prepare the contact layer, and a nano-silver electrode is formed on the polyimide film by screen printing. The shape of the electrode unit is square, and the row and column electrode units are connected by a spiral wire. The dielectric layer uses a porous PDMS elastomer doped with barium titanate (the mass ratio of PDMS to barium titanate is 10:1, see Deep-learning enabled smartinsole system aiming for multifunctional foot-healthcare applications, https: / / onlinelibrary.wiley.com / doi / 10.1002 / EXP.20230109).

[0070] Example 3

[0071] The three-dimensional force dataset acquisition system includes a multiplexing-based multi-channel capacitance acquisition circuit and the capacitive three-dimensional force sensing array described in Embodiment 1 or Embodiment 2. The three-dimensional force dataset acquisition system uses the multiplexing-based multi-channel capacitance acquisition circuit to obtain the force-capacitance response changes of a 3×3 flexible three-dimensional force sensing array. Specifically:

[0072] The multiplexing-based multi-channel capacitance acquisition circuit mainly includes the following components: a multiplexing circuit, a capacitance digital conversion circuit, a microcontroller unit MCU, and a 485 communication circuit. Its working principle is as Figure 2 shown.

[0073] Multiplexing circuit: It includes two multiplexing chips CD74HC4051. The input pins of these two chips are respectively connected to the plate output terminals of the upper electrode layer and the lower electrode layer of the capacitive three-dimensional force sensing array. Each multiplexing chip has 1 output pin, and they are connected to the corresponding input pins of the capacitance digital conversion circuit; the multiplexing chips sequentially select different capacitance sensor signals through polling and transfer the output signals of the sensors to the capacitance digital conversion circuit for further processing.

[0074] Capacitance digital conversion circuit( Figure 6 ): It includes a capacitance digital conversion chip FDC2214 and an LC oscillation circuit. The two ends of the inductor L1 of the LC oscillation circuit are connected to the input pins INOA and INOB of the capacitance digital conversion chip FDC2214. The two ends of the inductor L1 are respectively connected to the output pins of 2 multiplexing chips, and the inputs of the multiplexing chips are connected to the capacitance sensing array. In this way, the inductor L1 and the capacitance sensing array form an LC oscillation circuit, and the LC oscillation circuit is connected in parallel to INOA and INOB of the FDC2214; 1 output pin of the capacitance digital conversion chip FDC2214 and 2 I2C communication pins are respectively connected to the microcontroller unit MCU to ensure the security and accuracy of data transmission. This circuit uses the LC oscillation principle to measure the capacitance value, where the oscillation frequency is inversely proportional to the capacitance value to be measured, as shown in Equation (1):

[0075]

[0076] The capacitance digital conversion chip can generate frequency signals in the range of 10KHz to 10MHz to detect the resonance frequency changes of the LC circuit, thereby reflecting the capacitance changes.

[0077] Microcontroller Unit (MCU): As the "brain" of the system, the MCU is responsible for tasks such as controlling the selection logic of the multiplexing circuit, configuring the parameters of the capacitance digital conversion circuit, and processing the data from the capacitance digital conversion circuit. In addition, it manages the operation of the 485 communication circuit to ensure that data can be accurately sent to the host computer. Specifically, the STM32F407 main control chip is selected in this embodiment, which features high performance, low power consumption, and rich peripheral resources. At the same time, to ensure the stable operation of the system, there is also a crystal oscillator circuit to provide an accurate clock source, a reset circuit for initializing or restoring the system state, and a spare serial port communication circuit to support additional data transmission requirements or debugging purposes.

[0078] Communication Circuit: The 485 communication circuit is a key part for realizing remote monitoring and data analysis. It operates based on the RS-485 standard protocol, allowing long-distance transmission of capacitance acquisition data and maintaining good anti-interference performance in a noisy environment. The collected digital information is transmitted to the host computer through the 485 communication line.

[0079] Embodiment 4

[0080] Based on Embodiment 1, 2, or 3, data acquisition is carried out. The method of obtaining the measurement values of the multi-channel capacitance acquisition circuit based on multiplexing and the output of the corresponding commercial standard three-dimensional force sensor through a dual-serial-port acquisition system developed based on LabVIEW. Its working process is as Figure 3 shown.

[0081] The baud rate of Serial Port 1 is 460800. It receives data from the multi-channel capacitance acquisition circuit based on multiplexing, intercepts the first 2 bytes of the data frame and checks the correctness of the data frame. Subsequently, every two bytes form a unit of capacitance value, which is converted from hexadecimal to single-precision floating-point number and continuously displayed in the form of numbers and waveforms. At the same time, the converted data is saved in a txt file at a fixed position.

[0082] The baud rate of Serial Port 2 is 115200, which is used to receive the output from the commercial standard three-dimensional force sensor. According to the instructions of the commercial three-dimensional force sensor, the serial port first sends a request data field to it, and then the data sent by the commercial three-dimensional force sensor can be received.

[0083] The data frame structure sent by the commercial three-dimensional force sensor is as follows: The first two bytes represent the device code, the third and fourth bytes represent the device address, and the subsequent 6 bytes are the measurement values of the three-dimensional forces fx, fy, and fz, with each measurement value being two bytes.

[0084] Using the string truncation function and data type coercion, the measured values in the data frame are converted into single-precision floating-point numbers and continuously displayed in the form of numbers and waveforms. At the same time, the converted data is saved in a txt file at a fixed location.

[0085] The data of Serial Port 1 and Serial Port 2 can be saved in the same file, and the time counter can also be stored in the same file. LabVIEW matches the data of the two serial ports with the system time according to the system resource allocation.

[0086] Example 5

[0087] The construction, training, and decoupling process of the three-dimensional force information decoupling algorithm are as follows:

[0088] By preprocessing the three-dimensional force data set obtained in Example 4, a random forest is constructed to achieve regression prediction of the three-dimensional force. The flow chart is as Figure 4 shown, including the following steps:

[0089] (1) Data collation:

[0090] The responses of the pre-collected flexible three-dimensional force sensing array are collated with the corresponding true three-dimensional force values to obtain a three-dimensional force data set.

[0091] (2) Data preprocessing: Use Matlab to perform maximum-minimum normalization on the three-dimensional force data set. The calculation method is shown in Equation (2):

[0092]

[0093] Perform a preliminary verification check on the obtained data information to ensure the authenticity and effectiveness of the data.

[0094] (3) Feature engineering: According to the working principle of the flexible three-dimensional force sensing array, further deduce the relationship between its output and the three-dimensional force to enrich the feature relationship between the input and output:

[0095] Derive additional feature quantities according to the empirical formula to improve the prediction performance of the random forest model. The derived additional variables <C x , C y , C z > and <C1, C2, C3, C4> have the following quantitative relationship:

[0096]

[0097] Therefore, for each three-dimensional force sensing unit, after feature engineering, its output is <C1, C2, C3, C4, C x , C y , C z >.

[0098] (4) Establish a model for prediction: Using the random forest method in machine learning, divide the data set obtained in step 3 to establish a regression prediction model for the output of the flexible three-dimensional force sensing array and the three-dimensional force. Input the data set into the three-dimensional force regression prediction model for data training, testing, and / or prediction.

[0099] First, divide the obtained data set into a training set and a test set in a ratio of 7:3, and use the random forest (RF) method to construct a regression prediction model between the capacitance response and the three-dimensional force.

[0100] Establish a model through machine learning methods and calculate the overall average error and the overall mean square error MSE, as well as the mean square error MSE_F for the three-dimensional force F x ,F y ,F z of x ,MSE_F y ,MSE_F z .

[0101] The results are shown in Table 1 and Figure 5 , and it is found that the good stability and prediction accuracy of the model can be seen from the values of the random forest model.

[0102] The above parameters for evaluating the prediction ability of the quantitative regression prediction model are the overall mean square error MSE and the mean square error MSE_F for the three-dimensional force F x ,F y ,F z of x ,MSE_F y ,MSE_F z , and the distributions are shown in Eqs. (3)-(4):

[0103]

[0104] where Y i is the actual value, is the predicted value, and N represents the number of samples.

[0105] where the value of ω is x, y, or z. F ω represents the actual value, represents the predicted value.

[0106] where MSE and MSE_F ω are error indicators, and the smaller the value, the smaller the prediction error and the higher the model stability;

[0107] Introduce the error indicators MSE and MSE_F ω , MSE is a measure of the error of the overall prediction, while MSE_Fω For evaluating the prediction performance for a certain one-dimensional force information.

[0108] Table 1

[0109] MSE <![CDATA[MSE_F x > <![CDATA[MSE_F y > <![CDATA[MSE_F z > Training set 0.0283 0.0214 0.0441 0.0195 Test set 0.0598 0.0431 0.1014 0.0349

[0110] (5) Prediction:

[0111] Analyze the results of the actual values and the predicted values: Using the random forest model, based on the visualization between the predicted values and the actual values, the regression prediction accuracy of the model for decoupling three-dimensional force information is clarified.

[0112] Example 6

[0113] The process of constructing, training the three-dimensional force information decoupling algorithm and decoupling the three-dimensional force information refers to Example 5, the difference is that step (3) feature engineering is not adopted.

[0114] Example 7

[0115] The regression prediction of the three-dimensional force is realized through the random forest constructed and trained in Example 5. When the error indicators MSE and MSE_F ω After meeting the requirements of the applied scenario, the pre-trained random forest model can be imported through the Matlab Script Node in Labview to predict the real-time data. The specific operation method is as follows:

[0116] (1) Construct the Matlab Script Node

[0117] On the Labview program panel, construct the Matlab Script Node, and construct 36 input nodes to receive 36 capacitance output values from a 3×3 capacitive sensing array, and construct 27 output nodes for outputting 9 three-dimensional force output information of a 3×3 capacitive three-dimensional force sensing array.

[0118] (2) Program configuration

[0119] Use matlab to implement feature engineering and prediction for 36 capacitance values. And transfer the results to the output variables.

[0120] (3) Visualization of output results

[0121] The 27 output nodes contain the three-dimensional force information sensed by 9 three-dimensional force sensing units, and connect the output control to visualize the results.

[0122] The above are only the preferred embodiments of the present invention. It should be noted that for those of ordinary skill in the art, without departing from the principle of the present invention, several improvements and refinements can be made, and these improvements and refinements should also be regarded as the protection scope of the present invention.

Claims

1. A capacitive three-dimensional force sensing array, characterized in that: From top to bottom are the contact layer, the upper electrode layer, the dielectric layer, and the lower electrode layer. Among them, both the upper electrode layer and the lower electrode layer are composed of electrodes and substrate materials. The electrodes are fixed on the substrate materials. The electrodes include electrode units and wires connecting each electrode unit. Among them, the top electrode and the bottom electrode and the dielectric material therein form a capacitive sensing unit. Multiple top electrodes are arranged and connected to form the upper electrode layer, and multiple bottom electrodes are arranged and connected to form the lower electrode layer. Each sensing unit in the sensing array includes four capacitive sensing units arranged in a 2×2 matrix and a top contact, and is extended through the capacitive sensing unit.

2. The capacitive three-dimensional force sensing array according to claim 1, wherein: The upper electrode layer and / or the lower electrode layer is a gold electrode deposited on a PET film by magnetron sputtering, where the shape of the electrode unit is square, and the row and column electrode units are connected by deposited gold wires; or a nano-silver electrode is formed on a polyimide film by screen printing, where the shape of the electrode unit is square, and the row and column electrode units are connected by spiral wires.

3. The capacitive three-dimensional force sensing array according to claim 1, characterized in that: Extended through the capacitive sensing unit, for an N×M dot capacitive three-dimensional force sensing array, it includes 2N×2M capacitive sensors and N×M top contacts; Preferably, in each of the capacitive sensing units, both the upper electrode layer and the lower electrode layer are 4 electrode units arranged in a 2×2 matrix, and each column of electrode units in the upper electrode layer is connected, and each row of electrode units in the lower electrode layer is connected; More preferably, among the top electrodes of the 2N×2M capacitive sensors, each column of electrode units is interconnected to form a 2M-column 1×N electrode arrangement; among the bottom electrodes, each row of electrode units is interconnected to form a 2N-row 1×M electrode arrangement.

4. A three-dimensional force data set acquisition system, characterized in that: Comprising the capacitive three-dimensional force sensing array according to any one of claims 1-3, a multi-channel capacitance acquisition circuit, and a dual-serial-port acquisition system developed based on Iabview; the multi-channel capacitance acquisition circuit includes a multiplexing circuit, a capacitance digital conversion circuit, a microcontroller unit MCU, and a communication circuit, where, The multiplexing circuit includes two multiplexing chips. The input pins of the multiplexing chips are respectively connected to the output terminals of the upper electrode layer and the lower electrode layer of the capacitive three-dimensional force sensing array: the 2 output pins of the multiplexing chips are connected to the corresponding 2 input pins of the capacitance digital conversion circuit; the multiplexing chips select different capacitive sensor signals in a polling manner and sequentially transmit the output signals of the sensors to the capacitance digital conversion circuit; The capacitance digital conversion circuit includes a capacitance digital conversion chip and an LC oscillation circuit. The capacitance digital conversion chip is connected to the LC oscillation circuit by wires. The LC oscillation circuit is connected to a multiplexing chip by circuit. The capacitance digital conversion chip is connected to a microcontroller unit (MCU) by circuit: Both ends of the inductor L1 of the LC oscillation circuit are connected to the input pins INOA and INOB of the capacitance digital conversion chip. Both ends of the inductor L1 are respectively connected to the output pins of two multiplexing chips. The inputs of the multiplexing chips are connected to a capacitance sensing array. In this way, the inductor L1 and the capacitance sensing array form an LC oscillation circuit. The LC oscillation circuit is connected in parallel to INOA and INOB of the capacitance digital conversion chip; One output pin of the capacitance digital conversion chip and two I2C communication pins are respectively connected to the microcontroller unit (MCU); The communication circuit is connected to the microcontroller unit (MCU) by wires and performs data interaction with a host computer using a communication protocol.

5. The three-dimensional force data set acquisition system according to claim 4, characterized in that: The capacitance digital conversion circuit measures capacitance using the principle of an LC oscillation circuit. The oscillation frequency is inversely proportional to the capacitance value. The capacitance digital conversion chip outputs a signal with a frequency of 10 KHz to 10 MHz, detects the resonance frequency of the LC circuit, and the change in the resonance frequency reflects the change in capacitance: The oscillation frequency is inversely proportional to the capacitance value to be measured, as shown in Equation (1):

6. The three-dimensional force dataset acquisition system according to claim 4, wherein: The dual-serial-port acquisition system developed based on LabVIEW has two serial-port resources. Among them, serial port 1 is used to receive data from the multiplexing-based multi-channel capacitance acquisition circuit, and serial port 2 is used to receive the output of a commercial standard three-dimensional force sensor; Preferably, the serial-port resources are accessed through a LabVIEW program. The capacitance measurement values from the multiplexing-based multi-channel capacitance acquisition circuit are received in real time using serial port 1. At the same time, the three-dimensional force data of the commercial three-dimensional force sensor is received using serial port 2. A fixed field is sent to the commercial three-dimensional force sensor through the LabVIEW system to request three-dimensional force measurement data, and the capacitance measurement values and three-dimensional force data are written into a file for storage using a data writing module; Further preferably, serial port 1 receives data from the multiplexing-based multi-channel capacitance acquisition circuit in real time. The correctness of the data frame is checked by character intercepting, and then the data bits are processed according to the pre-confirmed communication protocol, converted into single-precision floating-point numbers, and continuously displayed in the form of numbers and waveforms. Finally, the displayed data is saved in a txt file at a fixed position; Serial port 2 is used to receive the output from the commercial three-dimensional force sensor. According to the settings of the commercial three-dimensional force sensor, the serial port first sends a request data field to it, and then the data bits are processed according to the pre-confirmed communication protocol, converted into single-precision floating-point numbers, and continuously displayed in the form of numbers and waveforms. Finally, the displayed data is saved in a txt file at a fixed position; Further preferably, the capacitance measurement value and the real-time three-dimensional force data are written into a file simultaneously, and the two parts share a time counter, which is allocated by LabVIEW according to the system resources. The capacitance measurement value is matched with the real-time three-dimensional force data and the time of the system to form a three-dimensional force data set; Further preferably, a three-dimensional force information decoupling module is added to the LabVIEW program for real-time information decoupling: after processing the data received by serial port 1, it is input into the script program box of LabVIEW, the pre-set three-dimensional force information decoupling algorithm is called, and the output is connected to a suitable display control to display the decoupled three-dimensional force information.

7. A method for data acquisition and information decoupling using the three-dimensional force dataset acquisition system described in claims 4-6, characterized in that: The specific steps are as follows: (1) Data sorting: Sort out the responses of the pre-collected flexible three-dimensional force sensing array and the corresponding real three-dimensional force values to obtain a three-dimensional force data set; (2) Data preprocessing: Remove the outliers in the three-dimensional force data set and replace them with the average value of the two adjacent values; (3) Establish a model for prediction. Using the random forest method in machine learning, divide the data set obtained in step (3), and establish a regression prediction model for the output of the flexible three-dimensional force sensing array and the three-dimensional force; Input the data set into the three-dimensional force regression prediction model for data training, testing, and / or prediction.

8. The data acquisition and information decoupling method according to claim 7, wherein: The data preprocessing in step (2) includes the following steps: For a certain flexible three-dimensional force sensing unit in the flexible three-dimensional force sensing array, its output variables include <C1, C2, C3, C4>; the three-dimensional force corresponding to each group of outputs is <F x , F y , F z >. Perform max-min normalization on the above 7 variables. The calculation method is shown in Equation (2): where X represents any variable in <C1, C2, C3, C4> or <F x , F y , F z >, and X max and X min represent the maximum and minimum values of any one of the variables respectively, and X normalized is the value after normalization.

9. The data acquisition and information decoupling method according to claim 7, wherein: Feature engineering is also included after step (2): According to the working principle of the flexible three-dimensional force sensing array, further deduce the relationship between its output and the three-dimensional force to enrich the feature relationship between the input and the output; Preferably, the feature engineering includes the following steps: (I) Derive the further relationship between the vectors <C1, C2, C3, C4> and <F x , F y , F z > by using methods including but not limited to theoretical analysis and mathematical modeling and solution; (II) Derive the quantitative relationships among C1, C2, C3, and C4 and those among F x , F y , F z by means including but not limited to coordinate substitution.

10. The data acquisition and information decoupling method according to claim 7, wherein: In step (3), the machine learning method is random forest, and the data set obtained from the feature engineering is divided into a training data set and a test data set according to a ratio of 7:3; Preferably, the parameters used to evaluate the prediction ability of the quantitative regression prediction model in step (3) are the overall mean square error MSE, and the mean square error MSE_F for the three-dimensional force F x , F y , F z The mean square error MSE_F x , MSE_F y , MSE_F z , and the distribution is shown in formulas (3)-(4): Among them, Y i is the actual value, is the predicted value, and N represents the number of samples; Among them, the value of ω is x, y or z. F ω represents the actual value, represents the predicted value: Among them, MSE and MSE_F ω are error metrics. The smaller the value, the smaller the prediction error and the higher the model stability; Introduce the error metrics MSE and MSE_F ω , MSE is a measure of the error in the overall prediction, while MSE_F ω is used to evaluate the prediction performance for a certain dimension of force information; Further preferably, in the step (3), the prediction includes inputting the capacitance response values in the three-dimensional force dataset into a random forest program for prediction to obtain the corresponding three-dimensional force information F x , F y , F z .

Images