CN110806260A - Ultrasonic levitation three-dimensional manipulation control method and system based on neural network - Google Patents
Ultrasonic levitation three-dimensional manipulation control method and system based on neural network Download PDFInfo
- Publication number
- CN110806260A CN110806260A CN201911008064.5A CN201911008064A CN110806260A CN 110806260 A CN110806260 A CN 110806260A CN 201911008064 A CN201911008064 A CN 201911008064A CN 110806260 A CN110806260 A CN 110806260A
- Authority
- CN
- China
- Prior art keywords
- neural network
- ultrasonic
- matrix
- sound
- control
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Images
Classifications
-
- G—PHYSICS
- G10—MUSICAL INSTRUMENTS; ACOUSTICS
- G10K—SOUND-PRODUCING DEVICES; METHODS OR DEVICES FOR PROTECTING AGAINST, OR FOR DAMPING, NOISE OR OTHER ACOUSTIC WAVES IN GENERAL; ACOUSTICS NOT OTHERWISE PROVIDED FOR
- G10K15/00—Acoustics not otherwise provided for
- G10K15/02—Synthesis of acoustic waves
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01H—MEASUREMENT OF MECHANICAL VIBRATIONS OR ULTRASONIC, SONIC OR INFRASONIC WAVES
- G01H17/00—Measuring mechanical vibrations or ultrasonic, sonic or infrasonic waves, not provided for in the preceding groups
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
Abstract
The invention discloses an ultrasonic suspension three-dimensional control method based on a neural network, which uses MATLAB software to generate a plurality of groups of random phase distribution data, generates sound potential trap depth distribution data corresponding to the phase distribution data one by one, collects the data and establishes a training sample set; constructing a neural network model and training; setting a pre-manipulated path and decomposing the pre-manipulated path into unit paths in the x direction, the y direction and the z direction, and constructing a sound potential trap depth distribution matrix for realizing the movement of the unit paths, wherein the matrix meets the following requirements: generating gradient near the point to be controlled and maximizing the depth of the potential well; simultaneously inputting the sound potential trap depth distribution matrix into three trained neural network models; the three neural network models respectively output a phase control matrix which respectively corresponds to the phase control of the three-direction ultrasonic phased array. The invention also discloses an ultrasonic suspension three-dimensional control system based on the neural network. The method and the system can ensure that the ultrasonic suspension three-dimensional control is more accurate.
Description
Technical Field
The invention relates to an ultrasonic levitation three-dimensional manipulation control method and system, in particular to an ultrasonic levitation three-dimensional manipulation control method and system based on a neural network.
Background
At present, in 1866, german scientists' holes especially first reported experimental phenomena in which sound waves in resonator tubes can suspend dust particles. Gor' kov subsequently proposed the time-averaged behavior of acoustic radiation force in 1962, providing a simple and convenient method for calculating acoustic radiation force. Under the guidance of theory, the first acoustic suspension device for single droplet dynamics study was available. In the next decades, acoustic suspension technology has gradually developed and is now widely used in the fields of cell manipulation, container-less processing, droplet fabrication, and the like. It is worth mentioning that compared with visible light or infrared light adopted by the optical tweezers, the acoustic wave has longer wavelength, so that the generated binding force is 5 orders of magnitude higher than that of the optical tweezers, the acoustic wave can better penetrate through an object, the directional control in the organism is realized, the safety and the nondestructive performance are higher, and the variety of suspended media can be wider. Therefore, further development, transformation and innovation of the acoustic tweezers technology are inevitable trends of future social science and technology development. However, most of the current developments are that the suspended matter is bound in the sound potential trap with extremely high stability and is accurately manipulated, but all the approaches of forward design are adopted to achieve the purpose of manipulation, so that certain limitations exist, namely the known oscillator phase distribution is used for forward deducing the sound field distribution, and the requirement of the change of the corresponding phase distribution caused by the real-time change of the manipulation track in the production activity cannot be met.
Disclosure of Invention
The invention provides an ultrasonic suspension three-dimensional control method and system based on a neural network, which can control the real-time change of a track, and aims to solve the technical problems in the prior art.
The technical scheme adopted by the invention for solving the technical problems in the prior art is as follows: an ultrasonic levitation three-dimensional manipulation control method based on a neural network comprises the following steps: generating a plurality of groups of random phase distribution data by using MATLAB software, generating sound potential trap depth distribution data corresponding to the phase distribution data one by one, collecting the phase distribution data and the sound potential trap depth distribution data and establishing a training sample set; constructing a neural network model and training; setting a pre-manipulated path and decomposing the pre-manipulated path into unit paths in the x direction, the y direction and the z direction, and constructing a sound potential well depth distribution matrix for realizing the movement of the unit paths so that the sound potential well depth distribution matrix meets the following requirements: generating gradient near the point to be controlled and maximizing the depth of the potential well of the point to be controlled; inputting the depth distribution matrix of the sound potential well into the three trained neural network models simultaneously; and the three neural network models respectively output a phase control matrix corresponding to the phase control of the ultrasonic phased array in the x direction, the y direction and the z direction.
Further, the neural network model is constructed based on a U-net neural network model, and the neural network model comprises two parts: the first half part comprises a convolution layer and a pooling layer and is used for carrying out feature extraction on input information and reducing the dimensionality of an input matrix; the second half includes an inverse convolutional layer to unwind the data features and increase the dimensionality of the output matrix.
Further, the training method of the neural network model comprises the following steps: taking a part of data in the training sample set as a training set, and taking a part of data in the training sample set as a verification set; inputting data in a training set into the neural network model for training, verifying the data in a verification set after each round of training and calculating loss; and taking the trained neural network model with the lowest loss in verification as the trained neural network model.
Further, the method for generating random phase distribution data and corresponding sound potential trap depth distribution data by using MATLAB software comprises the following steps: and generating a random phase distribution matrix by using a rand function, inputting the phase distribution matrix into MATLAB software to calculate a sound pressure field, and calculating a sound potential well depth distribution matrix by using the sound pressure field so as to obtain random phase distribution data and corresponding sound potential well depth distribution data.
Further, when the neural network model is trained, the loss function adopts a mean square error loss function, and a gradient descent momentum method is adopted to optimize parameters of the neural network model.
Further, the method for constructing the sound potential well depth distribution matrix for realizing the unit path movement comprises the following steps: the wave path of the sound wave sent by each vibrator at the point is consistent, namely sound focusing is carried out, so that the corresponding sound potential at the point to be suspended reaches the peak value, and a sound potential matrix is obtained; multiplying the sound potential matrix by-1 to maximize the depth of the sound potential trap at the point to be suspended to obtain a target matrix, calculating the sound potential matrix at different heights during sound focusing, and multiplying the sound potential matrix by-1 to generate a sound potential trap depth distribution matrix.
The invention also provides an ultrasonic suspension three-dimensional control system based on the neural network, which comprises an artificial intelligence system used as an upper computer and a driving system used as a lower computer; the artificial intelligence system comprises a memory, a processor and a computer program stored on the memory and operable on the processor, the processor implementing the method steps when executing the program; the driving system comprises a control module and an ultrasonic phased array module driven by the control module; the processor outputs a phase control matrix to the control module; the control module generates a drive signal to drive the ultrasonic phased array module to produce a sound field.
Furthermore, the control module is a control module based on a digital register signal or a control module based on an analog oscillation signal; wherein:
the control module based on the digital register signal comprises a control board and a drive board A, the processor outputs a control matrix to the control board, and the control board generates a square wave signal to the drive board A; the driving board A supplies power to the control board, amplifies the power of the square wave signal, and outputs the voltage peak value to the ultrasonic phased array module after the voltage peak value reaches the optimal working value of the ultrasonic vibrator;
the control module based on the analog oscillation signal comprises a master control board, a control circuit unit and a drive board B; the control circuit unit comprises a gate, a delay timer and a signal generator; the processor outputs a control matrix to the master control board; the master control board selects the delay timer through the gate, writes delay time into a data register in the delay timer, sends a signal to start the delay timer, and sends a signal to start the signal generator when the delay time is up; the signal generator starts to generate square wave signals and outputs the square wave signals to the driving plate B, and the driving plate B amplifies the power of the received square wave signals and outputs the square wave signals to the ultrasonic phased array module.
Further, the ultrasonic phased array module is a biconcave ultrasonic phased array module or a planar ultrasonic phased array module.
Further, the double-concave ultrasonic phased array module is composed of 36-256 ultrasonic vibrators, and the diameter of each ultrasonic vibrator is 6-15 mm.
The invention has the advantages and positive effects that: the ultrasonic suspension three-dimensional control method provides a reverse path calculation and control method based on an artificial intelligence neural network, namely, oscillator phase distribution is directly obtained by only knowing sound field distribution corresponding to a moving path to be controlled through the artificial intelligence neural network calculation, the practical application value of ultrasonic control is improved, and a complicated one-to-one specific modeling calculation process is avoided. When the ultrasonic suspension three-dimensional manipulation control system adopts the double-concave ultrasonic phased array module, the transverse and longitudinal controllable stable movement of particles can be realized by means of a double-concave structure under the conditions of simple system element composition and low cost; when the planar ultrasonic phased array module is adopted, three-dimensional movement of particles along a specific path in three directions can be realized under the conditions that the ultrasonic phased array module is simple in structure and has no specific curved surface structure, and ultrasonic suspension and manipulation are realized in a three-dimensional mode from the angle of the device.
Drawings
FIG. 1 is a schematic workflow diagram of the present invention;
FIG. 2 is a schematic flow chart of the present invention when using a biconcave ultrasonic phased array module;
FIG. 3 is a wiring diagram of the UNO type IC and LS98N IC of the Arduino module of the present invention;
FIG. 4 is a schematic diagram of the ultrasonic transducer arrangement and wiring of a biconcave ultrasonic phased array module of the present invention;
FIG. 5 is a schematic flow chart of the present invention when using a planar ultrasonic phased array module;
FIG. 6 is a schematic diagram of the operation of a control circuit unit of the present invention;
FIG. 7 is a wiring diagram of a strobe circuit of the present invention;
FIG. 8 is a wiring diagram of the operation of a delay timer of the present invention;
FIG. 9 is a schematic diagram of a single-sided structure in an ultrasound transducer support structure of a biconcave ultrasonic phased array module of the present invention;
FIG. 10 is a top view of an ultrasonic vibrator support structure of a planar ultrasonic phased array module of the present invention;
FIG. 11 is a half cross-sectional front view of an ultrasonic vibrator support structure of a planar ultrasonic phased array module of the present invention;
FIG. 12 is a graph showing the lateral acceleration of a working bead over time with a biconcave ultrasonic phased array module in accordance with the present invention;
FIG. 13 is a graph of the longitudinal acceleration of a working bead over time using a biconcave ultrasonic phased array module in accordance with the present invention.
Detailed Description
For further understanding of the contents, features and effects of the present invention, the following embodiments are enumerated in conjunction with the accompanying drawings, and the following detailed description is given:
referring to fig. 1 to 13, an ultrasonic levitation three-dimensional steering control method based on a neural network includes: generating a plurality of groups of random phase distribution data by using MATLAB software, generating sound potential trap depth distribution data corresponding to the phase distribution data one by one, collecting the phase distribution data and the sound potential trap depth distribution data and establishing a training sample set; constructing a neural network model and training; and training and verifying the neural network model by using the phase distribution data in the training sample set and the corresponding sound potential trap depth distribution data to optimize parameters of the neural network model, and obtaining the trained neural network model by training for multiple times and finding the best model parameter fitted by the mapping relation between the sound field and the phase.
Setting a pre-manipulated path and decomposing the pre-manipulated path into unit paths in the x direction, the y direction and the z direction, and constructing a sound potential well depth distribution matrix for realizing the movement of the unit paths so that the sound potential well depth distribution matrix meets the following requirements: generating potential well gradient near the point to be controlled and maximizing the depth of the potential well of the point to be controlled; inputting the depth distribution matrix of the sound potential well into the three trained neural network models simultaneously; and the three neural network models respectively output a phase control matrix corresponding to the phase control of the ultrasonic phased array in the x direction, the y direction and the z direction. For example, the phase control matrix output by the first neural network model corresponds to the phase control of the ultrasonic phased array in the x direction; the phase control matrix output by the second neural network model corresponds to the phase control of the ultrasonic phased array in the y direction; the phase control matrix output by the third neural network model corresponds to the phase control of the ultrasonic phased array in the z direction; wherein the elements in the phase control matrix represent the phase information of a single element.
Preferably, the neural network model is constructed based on a U-net neural network model, and the neural network model includes two parts: the first half part comprises a convolution layer and a pooling layer and is used for carrying out feature extraction on input information and reducing the dimensionality of an input matrix; the second half includes an inverse convolutional layer to unwind the data features and increase the dimensionality of the output matrix.
Preferably, the training method of the neural network model includes: using MATLAB software to generate a plurality of groups of random phase distribution data and sound potential trap depth distribution data corresponding to the random phase distribution data one by one, collecting two kinds of associated data and establishing a training sample set; taking a part of data in the training sample set as a training set, and taking a part of data in the training sample set as a verification set; inputting data in a training set into the neural network model for training, verifying the data in a verification set after each round of training and calculating loss; and taking the trained neural network model with the lowest loss in verification as the trained neural network model.
Preferably, the method for generating random phase distribution data and corresponding sound potential trap depth distribution data by using MATLAB software comprises the following steps: and generating a random phase distribution matrix by using a rand function, inputting the phase distribution matrix into MATLAB software to calculate a sound pressure field, and calculating a sound potential well depth distribution matrix by using the sound pressure field so as to obtain random phase distribution data and corresponding sound potential well depth distribution data.
Preferably, when the neural network model is trained, the loss function adopts a mean square error loss function, and a gradient descent momentum method is adopted to optimize parameters of the neural network model.
Preferably, the method for constructing the sound potential well depth distribution matrix for realizing the unit path movement comprises the following steps: the wave path of the sound wave sent by each vibrator at the point is consistent, namely sound focusing is carried out, so that the corresponding sound potential at the point to be suspended reaches the peak value, and a sound potential matrix is obtained; multiplying the sound potential matrix by-1 to maximize the depth of the sound potential trap at the point to be suspended to obtain a target matrix, calculating the sound potential matrix at different heights during sound focusing, and multiplying the sound potential matrix by-1 to generate a sound potential trap depth distribution matrix.
Referring to fig. 1, fig. 1 is a schematic diagram of a work flow of the present invention, in which s100-s1400 represent each work substep, respectively, and the work flow of the present invention is described in detail below with reference to fig. 1:
s100, collecting a random phase sequence; s200, obtaining random sound field distribution; s300 represents training a neural network; s400, judging whether the loss function of the neural network is extremely small; s500, obtaining a trained neural network; s600 represents a phase distribution of the output target; s700 represents a hardware drive output signal; s800, the ultrasonic array sends out a specific signal to realize operation; s900 represents the judgment of whether or not this is the end point of the manipulation object; s1000 denotes the end of acoustic steering; s1100, selecting an acoustic control position; s1200, converting into a sound field slice image; s1300, sound field information is input into a neural network; s1400 represents performing the next acoustic manipulation step.
The method comprises the steps of training a neural network by using a computer, building a U-net neural network, dividing the U-net neural network into two parts, extracting features of input information by a convolution layer and a pooling layer in the first half part, and reducing dimensionality of an input matrix. And the data characteristics are expanded in the second half part through reverse convolution, the dimensionality of the processed matrix is improved, and finally the mapping from the matrix to the matrix is formed.
The random phase sequence in S100 refers to: and (3) carrying out phase calculation on the oscillators by using MATLAB software, establishing a model of 6X 6-16X 16 oscillator arrays in the MATLAB, specifically 6X 6-16X 16 matrixes, and describing the positions of sound sources by using matrix elements.
The random sound field distribution in S200 means: in the MATLAB program programming, an initial phase distribution matrix is input and a sound pressure field is calculated, the depth of a sound potential well is given according to the sound pressure field, two adjacent action surfaces are selected and respective sound pressure distribution is calculated when the gradient of sound pressure along the z direction is calculated, and finally the derivative of the sound pressure along the z direction is calculated by using the value of a gradient () function. Thus, the calculation relation from the phase distribution to the depth distribution of the sound potential well is established.
And generating a random phase distribution matrix by using a rand () function, calculating corresponding potential well distribution, and repeatedly obtaining a phase distribution-potential well distribution data pair which is used as a learning sample for the following neural network training.
The training of the neural network in S300 refers to: and obtaining a plurality of groups of data pairs of sound fields and phases in one-to-one correspondence, wherein the size of a sound field distribution matrix is determined by required precision, and the phase distribution is represented by a matrix with the size of 6 x 1-16 x 1. These data are used as the basic data set for U-net neural network training. The U-net neural network selects a mean square error loss function, the optimizer selects a Momentum optimizer, a gradient descending Momentum system is introduced, and the situation that the U-net neural network falls into a local minimum value in the training process can be avoided.
During training, calculated data enter from an input layer, a predicted output is finally obtained at an output layer through the processing of the U-net neural network, parameters of the whole network are randomly generated at the beginning of training, the difference between the predicted result and the standard output is large, the loss function is a mode for quantifying the difference, and after the loss function at the moment is calculated, the parameters of the whole network are adjusted through a gradient descending mode, so that the fitting effect of the network on input-output mapping is better.
Whether the loss function in S400 is extremely small means: as training progresses, the loss function gradually decreases and eventually approaches a value, indicating that the network can better fit on all the data involved in training, and ending the data training process (S500).
In order to prevent the phenomenon of 'overfitting', namely, the phenomenon that the fitting effect on training data is too good and the fitting effect on a non-training set is not good, a group of verification sets which are not used for training is reserved, a loss function on the verification sets is calculated after each round of training, the loss function value of the verification sets can show a trend of decreasing and then increasing along with the increase of the training rounds, the training rounds with the minimum loss function on the verification sets are found, and the model at the moment is the state with the best fitting on the mapping relation between the sound field and the phase and can be used as an output model for training.
After the training is finished, any pre-realized sound field distribution can be used as input data and input into the trained U-net neural network model input layer, and a corresponding phase distribution can be obtained in an output result.
The acoustic manipulation position in S1100 means: in the practical use process, after the pre-steering path is selected, the pre-steering path is divided into paths in the x direction, the y direction and the z direction, the paths of the x component, the y component and the z component are all divided into the same tiny units, the size of each tiny unit is 0.1mm-1mm, and the tiny units are divided into the same number of unit paths.
The sound field slice in S1200 means: constructing a numerical matrix of the depth distribution of the sound potential trap capable of realizing unit path movement: a large gradient is generated near the point to be manipulated and the depth of the potential well of the point to be manipulated is maximized. A slice of the acoustic field is formed from a matrix of values of the acoustic well depth profile.
The method for constructing the sound potential trap depth distribution matrix for realizing the unit path movement specifically comprises the following steps:
1. and (3) enabling the sound potential corresponding to the point to be suspended to have a peak value, and adopting the scheme: the wave path of the sound wave emitted by each vibrator at the point is consistent, namely, the sound is focused. This is easily achieved in a forward design. A matrix of the sound potentials is obtained.
2. Taking a negative sign of the whole sound potential matrix, namely multiplying the matrix by-1 to obtain a target matrix, wherein the matrix is characterized in that a large gradient is arranged near a point to be controlled, and the depth of a potential well of the point is maximum.
3. When the target matrixes with different heights are required to be obtained, a sound field slice can be formed and led into the neural network by sequentially calculating the matrixes at the different heights during sound focusing and taking the negative sign.
Inputting the sound field into the neural network in S1300 means: and three potential well matrixes exist corresponding to the three components of x, y and z, the three potential well matrixes are simultaneously input into three trained neural networks, the networks output corresponding three phase distributions, and the three phase distributions are linearly superposed into one phase distribution, so that the phase distribution corresponding to the unit path in any direction can be obtained. The network randomly outputs the corresponding phase by inputting different potential well matrices at time intervals of 1ns-1 mus microseconds (other reasonable ranges are all related by the response accuracy of specific hardware, i.e. the set time interval is larger than the standard time for executing the program command by the hardware), so that the particles can move along the planned path in practice.
In one preferred embodiment, the parameters in the above steps are as follows:
the S100 random phase sequence refers to: and (3) carrying out phase calculation on the oscillators by using MATLAB software, and establishing a model of an 8 x 8 oscillator array in the MATLAB, specifically an 8 x 8 matrix, wherein matrix elements describe the position of a sound source.
S300 training the neural network refers to: a number of sets of data pairs of sound fields and phases are obtained, the sound field distribution being represented by a matrix of size 256 x 1 and the phase distribution by a matrix of size 8 x 1. These data are used as the basic data set for U-net training. The U-net neural network selects a mean square error loss function, the optimizer selects Momentum, a gradient descending Momentum system is introduced, and the situation that the U-net neural network falls into a local minimum value in the training process can be avoided.
S1100 acoustic manipulation position: in practical use, after the pre-steering path is selected, the pre-steering path is divided into paths in the x direction, the y direction and the z direction, and the paths of the x component, the y component and the z component are divided into the same number of unit paths by the same tiny units, wherein the size of each tiny unit is 1 mm.
S1300 inputs the sound field into a neural network: and three potential well matrixes exist corresponding to the three components of x, y and z, the three potential well matrixes are simultaneously input into three trained neural networks, the networks output corresponding three phase distributions, and the three phase distributions are linearly superposed into one phase distribution, so that the phase distribution corresponding to the unit path in any direction can be obtained. The network randomly outputs the corresponding phase by inputting different well matrices at 1 mus intervals (other reasonable ranges are all dependent on the response accuracy of the specific hardware, i.e. the set time interval is larger than the standard time for the hardware to execute the program command), so that in practice the particles can move along the planned path.
The invention also provides an embodiment of an ultrasonic suspension three-dimensional control system based on the neural network, which comprises an artificial intelligence system used as an upper computer and a driving system used as a lower computer; the artificial intelligence system comprises a memory, a processor and a computer program stored on the memory and operable on the processor, the processor implementing the method steps when executing the program; the driving system comprises a control module and an ultrasonic phased array module driven by the control module; the processor outputs a phase control matrix to the control module; the control module generates a drive signal to drive the ultrasonic phased array module to produce a sound field. The processor outputs three phase control matrixes to the control module; the three phase control matrixes respectively correspond to the phase control of the ultrasonic phased array in the x direction, the y direction and the z direction, elements in the phase control matrixes represent phase information of a single oscillator, the processor serially inputs the elements in the phase control matrixes into the control module, the control module performs superposition processing on input signals, the control module outputs driving signals to the ultrasonic phased array module after processing, and the ultrasonic phased array module sends out ultrasonic waves to generate a sound field under the action of the driving signals.
The ultrasonic phased array module is an array unit consisting of a plurality of ultrasonic vibrators. And the artificial intelligence system serving as the upper computer obtains three phase control matrixes corresponding to the array phase information in the x direction, the y direction and the z direction respectively according to the size and the position of a preset suspension target. The elements in the phase control matrix correspond to phase information representing a single element, and each element is serially input into the control module. The control module outputs a square wave driving signal to drive the ultrasonic phased array module to send out ultrasonic waves so that the particles are suspended at a specific position.
Preferably, the control module is a control module based on a digital register signal or a control module based on an analog oscillation signal; wherein:
the control module based on the digital register signal comprises a control board and a drive board A, the processor serially outputs the phase control matrix to the control board, and the control board generates a square wave signal to the drive board A; the drive board A supplies power to the control board, amplifies the power of the square wave signal, and outputs the voltage peak value to the ultrasonic phased array module after the voltage peak value reaches the optimal working value of the ultrasonic vibrator.
The control module based on the analog oscillation signal comprises a master control board, a control circuit unit and a drive board B; the control circuit unit comprises a gate, a delay timer and a signal generator; the processor serially outputs the control matrix to the master control board; the master control board selects the delay timer through the gate, writes delay time into a data register in the delay timer, sends a signal to start the delay timer, and sends a signal to start the signal generator when the delay time is up; the signal generator starts to generate square wave signals and outputs the square wave signals to the driving plate B, and the driving plate B amplifies the power of the received square wave signals and outputs the square wave signals to the ultrasonic phased array module.
Preferably, the ultrasonic phased array module is a biconcave ultrasonic phased array module or a planar ultrasonic phased array module.
One of the concave ultrasonic phased array structures in the double-concave ultrasonic phased array module can be seen in fig. 9, and the ultrasonic vibrators are uniformly distributed on the concave surface and are sequentially arranged into a plurality of rings from inside to outside.
Preferably, the double-concave ultrasonic phased array module consists of 36-256 ultrasonic vibrators, and the diameter of each ultrasonic vibrator is 6-15 mm.
The digital register signal based control module can be used in a biconcave ultrasonic phased array module. The control module based on the analog oscillation signal can be used for a planar ultrasonic phased array module.
The working principle of the present invention is illustrated below by taking a biconcave ultrasonic phased array module and a planar ultrasonic phased array module as examples respectively:
the invention relates to an ultrasonic suspension three-dimensional control system based on a neural network, wherein the working principle of an ultrasonic phased array module of the ultrasonic suspension three-dimensional control system when a biconcave ultrasonic phased array module is adopted is described as follows:
referring to fig. 2, an ultrasonic levitation three-dimensional control system based on a neural network includes an artificial intelligence system as an upper computer, a power supply, a control board, a driving board a, a dual-concave ultrasonic phased array module, a dual-concave ultrasonic phased array supporting device, a control board and a driving board a, which form a control module for driving the ultrasonic phased array module.
The control panel receives a signal from an artificial intelligence system serving as an upper computer, processes the signal and outputs a square wave signal; wherein the control panel can include Arduino module, and Arduino module can regard as signal generator. The driving board A comprises an amplifying circuit, the amplifying circuit amplifies signals sent by the Arduino module and outputs the signals to the double-concave-surface ultrasonic phased array module.
The artificial intelligence system is used as an upper computer and comprises a memory, a processor and a computer program which is stored on the memory and can run on the processor, and when the processor executes the program, the ultrasonic levitation three-dimensional control method based on the neural network is realized. The processor inputs a pre-control path, the pre-control path is processed through the neural network, then signals are output to the control panel, the control panel is further processed, corresponding driving signals are output, the signals are amplified through the driving panel A and then output to the double-concave ultrasonic phased array module, and therefore the double-concave ultrasonic phased array module works.
The power supply can be 9V lithium ion rechargeable battery or stabilized voltage power supply (the embodiment can be respectively 9V lithium ion battery, power supply with 550mAh/MPS-3303 type stabilized voltage power supply, voltage regulation range of 0-30V, and output current of 0-3A).
The Arduino module in the control panel is used for generating an electric signal and controlling the phase of the signal. The UNO type integrated circuit of the Arduino module can be selected.
The driving board A can comprise an LS98N integrated circuit, and an LS98N integrated circuit can be selected to have an operating voltage of 5-35V and a maximum power of 25W. And while providing an output voltage of 5v for the control board, the power amplification is carried out on the square wave signal sent by the UNO pin, and the square wave signal is output to the biconcave ultrasonic array.
Referring to fig. 4 and 9, the biconcave ultrasonic phased array support device has a spherical concave shape, which is manufactured by PLA polylactic acid using 3D printing, and has a size range of a height of 120mm to 130mm, a width of 90mm to 95mm, preferably, a height of 125mm and a width of 95 mm; the center of the concave surface is positioned at the position 40-50mm away from the bottom surface of the device, 36 grooves with the diameter of 10.5mm are arranged on the concave surface, and two small holes with the size of 1.2mm are arranged in each groove according to the pin positions of the ultrasonic vibrators and used for inserting the ultrasonic vibrators.
The double-concave ultrasonic phased array module consists of 36-256 ultrasonic vibrators, the diameter of each ultrasonic vibrator can be 6-15mm, and the characteristic parameters of the ultrasonic phased array module can be as follows: the operating temperature is-40 to 85 degrees centigrade, the maximum input voltage is 20Vpp, the capacitance 2550pF, the frequency 40KHz, and the sound pressure level 120 dB. The working current of each oscillator is 0.017 mA. After the groove of the 5-double-concave-surface ultrasonic array supporting device is coated with AB glue, the ultrasonic vibrator is inserted into the groove, the positive pole of the vibrator is uniformly arranged on one side far away from the spherical center of the concave surface, and the pin of the vibrator is externally connected into a circuit through a small hole. All the ultrasonic vibrators in the upper array and the lower array are respectively connected in parallel.
See fig. 3 for wiring: the power supply supplies power to the LS98N integrated circuit in driver board a, which may provide a 5V operating voltage to the Arduino module UNO board in the control board. The analog output ports A0-A3 of the Arduino module UNO board in the control board are connected with LS98N integrated circuits DN1-DN4 in the drive board A, and the LS98N integrated circuits in the drive board A amplify square wave signals output by the Arduino module UNO board in the control board. The digital pin 10 and the pin 11 of the Arduino module UNO board in the control board are communicated for generating square wave signals. And after the GND pin lead is communicated with the pins 2, 3 and 4, a null command can be executed to further regulate and control the oscillator phase, wherein the pin 3 can enable the suspended matter to rise, the pin 4 enables the suspended matter to fall, and the pin 2 is a reset command. And then, two paths of outputs of the LS98N integrated circuit in the driving board A are respectively connected with the upper array and the lower array of the double-concave ultrasonic phased array module.
The double-concave ultrasonic array is used for generating ultrasonic waves with variable phases to form a stable sound field in the space between the upper array and the lower array, and the suspension target is suspended at a specific position with the minimum acoustic radiation force on the suspension point.
When the double-concave ultrasonic phased array module is adopted, the working steps of the ultrasonic suspension three-dimensional control system based on the neural network are as follows:
(1) the regulated power supply is activated to set the voltage at the appropriate value so that the vpp of the signals received by the ultrasound array is at the optimum location.
(2) The Arduino module in the control board transmits signals to the drive board a and can adjust the phase of the output signals to vary the sound field in space.
(3) The driving board A amplifies the power of a phase-adjustable signal output by an Arduino module in the control board, enables the voltage peak value to reach the optimal working value of the ultrasonic vibrator and outputs the signal to the hyperboloid ultrasonic array;
(4) the hyperboloid ultrasonic array emits ultrasonic waves with the same phase but adjustable, a stable sound field is formed in the middle of the array, and the sound suspension force of suspended matters on suspension points is minimum, so that the suspended objects are suspended at specific positions;
(5) the position of the suspended matter is changed in real time by changing the phase of the signal generated by the Arduino module, and the movement of the suspended matter is realized.
The delay time of exciting signals to the double-concave ultrasonic phased array module can be changed by declaring and executing an empty reading instruction in one period of oscillator signals through the programming of an artificial intelligence system of the upper computer, so that the initial phase of the sounding signals of the ultrasonic oscillators in the double-concave ultrasonic phased array module is changed.
In a preferred embodiment of the ultrasonic suspension three-dimensional control system based on the neural network and adopting the biconcave ultrasonic phased array module, components or modules in the ultrasonic suspension three-dimensional control system are selected from the following components or modules in model specifications:
power supply: selecting 9V lithium ion battery, power 550mAh or MPS-3303 model regulated power supply, voltage regulating range 0-30V, output current 0-3A
The double-concave ultrasonic phased array supporting device is manufactured by PLA polylactic acid through 3D printing, the size range of the double-concave ultrasonic phased array supporting device is that the height is 125mm, the width is 95mm, the center of a sphere of a concave surface is positioned at the height position of 45mm from the bottom surface of the device, 36 grooves with the diameter of 10.5mm are arranged on the concave surface, and two small holes with the size of 1.2mm are arranged in each groove according to the pin position of 61 ultrasonic vibrators for splicing the ultrasonic vibrators.
The double-concave ultrasonic phased array module consists of 72 ultrasonic oscillators.
The preferred embodiment can achieve the various parameters of the pellet in the sound field as shown in table 1 below:
TABLE 1 various parameters of the pellets in the Sound field
Referring to fig. 5, fig. 6, fig. 10, and fig. 11, the operation principle of the ultrasonic phased array module of the ultrasonic levitation three-dimensional steering control system based on neural network according to the present invention when the planar ultrasonic phased array module is adopted is described as follows:
an ultrasonic suspension three-dimensional control system based on a neural network comprises an artificial intelligence system serving as an upper computer, a power supply, a master control board, a control circuit unit, a drive board B, a planar ultrasonic phased array module supporting device, a planar ultrasonic phased array module and the like. The master control board, the control circuit unit and the drive board B form a control module for driving the ultrasonic phased array module. The plane ultrasonic phased array module supporting device can adopt a structure with a square cross section, and ultrasonic vibrators are uniformly distributed on the plane ultrasonic phased array module supporting device.
The artificial intelligence system is used as an upper computer and comprises a memory, a processor and a computer program which is stored on the memory and can run on the processor, and when the processor executes the program, the ultrasonic levitation three-dimensional control method based on the neural network is realized. The processor inputs a pre-control path, the pre-control path is processed by a neural network, then a signal is output to obtain a control matrix, the control matrix signals are changed into serial signals according to the position sequence of matrix elements and transmitted to the master control board, a decoder in the master control board carries out gating according to the relation between the input signal sequence and the corresponding control circuit, the control signals are written into the corresponding control circuit units, and after all the control circuit units write the control signals, the control circuit units are electrified to start delay counting. And the control circuit unit outputs a square wave signal with specified frequency after finishing delaying, outputs a corresponding driving signal to the driving board B, amplifies the signal by the driving board B, and outputs the signal to the planar ultrasonic phased array module to enable the planar ultrasonic phased array module to work.
The power supply supplies power to the whole circuit, the power supply can be a 9V lithium ion rechargeable battery or a stabilized voltage power supply, preferably a 9V lithium ion battery with the quantity being twice that of the power supply, a 550mAh/MPS-3303 type stabilized voltage power supply, the voltage regulation range is 0-30V, and the output current is 0-3A.
The control circuit unit may include a gate, a delay timer, and a signal generator. The gate can be 74HC238 and 74HC138 integrated circuits, the delay timer can be intel8253 chip, and the signal generator can be 40-80khz active crystal oscillator or 2Mhz active crystal oscillator. The function of the gate is: a specific delay timer is selected, and after the specific delay timer generates a phase delay signal, corresponding phase information is input. The intel8253 chip in the delay timer implements phase delay by counting, using the same delayed clock signal.
Referring to fig. 6, the control circuit unit can generate identical independent delay signals of 40-80kHz, and the working steps are as follows:
1) and the master control board accesses the data register part in the delay timer through the delay timer selected by the gate and writes the delay time.
2) The counter in the delay timer starts working, and the number in the data register is counted down to 0 according to the clock signal. And starting to output high level after 0, and starting the signal generator.
3) The signal generator signal generation module begins generating a signal.
Referring to fig. 7, the gate circuit: in order to realize the control of a wired pin pair 60-100 independent delay 40-80kHz signal generation modules, a decoder is expanded to a 7-line 128-bit decoder, and 60-100 generation units are subjected to time-sharing control through a chip selection terminal CS.
The realization characteristics are as follows: the writing of the delay time and the delay of the signal are separated, and although the writing of the delay time is separated, the delay starting of the signal is based on the same starting signal, so that the accuracy of the phase difference is ensured.
The driving board B can select an L298N electrode driving module, is used for outputting a phase delay signal to the planar ultrasonic phased array module through a driving circuit according to a time phase distribution signal output by the artificial intelligence system, and has the advantages of stable performance, high signal bearing frequency and the like, and short time delay effect and the like.
The planar ultrasonic phased array module is used for emitting ultrasonic waves with different phases, forming a specific sound field above the single-sided ultrasonic vibrator array, and realizing suspension of a suspended target at a specific position due to the fact that the acoustic radiation force of the suspended target on a suspension point is minimum. The planar ultrasonic phased array module can be composed of 60-100 ultrasonic vibrators with diameters of 6-15 mm.
Referring to fig. 8, the total control board is a raspberry pi 3B + control board, and an 8253 module is disposed in the total control board and is configured to receive phase information of the artificial intelligence system and distribute the phase information to the control circuit unit.
When a plane ultrasonic phased array module is adopted, the working steps of the ultrasonic suspension three-dimensional control system based on the neural network are as follows:
(1) starting an artificial intelligence system serving as an upper computer, calculating phase information required by the planar ultrasonic phased array module through the upper computer according to the size and the position of a preset suspended target, and sending the phase information to a master control board through serial port communication;
(2) the master control board distributes the phase information to the control circuit unit;
(3) the control circuit unit selects a specific delay timer according to the received phase information and a gate in the control circuit unit according to the input information, and sends a delay signal corresponding to the delay timer, the delay timer delays according to the received delay signal, the signal generator is started after the delay is finished, and finally square wave signals with different phases are sent from different signal generator pins, wherein the frequency of the square wave signals is 40-80KHz (other ultrasonic frequency ranges can be adopted, and the control circuit unit is not limited to the examples in the patent specification);
(4) the driving board B amplifies the power of the square wave signal sent out from the pin of the signal generator and outputs the square wave signal to the planar ultrasonic phased array module;
(5) the plane ultrasonic phased array module sends out ultrasonic waves with different phases, a specific sound field is formed above the plane ultrasonic phased array module, and the acoustic radiation force of a suspended target on a suspension point is minimum, so that the suspended target is suspended at a specific position;
the artificial intelligence system is used for calculating phase information required by the planar ultrasonic phased array module according to the size and the position of a preset suspended target and sending the phase information to the master control board; the master control board is used for receiving the phase information and distributing the phase information to the control circuit unit; the control circuit unit is used for sending square wave signals with different phases from each pin according to the received phase information; the drive board B is used for amplifying the power of the square wave signal sent by the pin of the control circuit unit and outputting the square wave signal to the planar ultrasonic phased array module; the planar ultrasonic phased array module is used for emitting ultrasonic waves with different phases, a specific sound field is formed above the array, the acoustic radiation force of a suspension target on a suspension point is minimum, and suspension at a specific position is achieved.
In a preferred embodiment of an ultrasonic suspension three-dimensional control system based on a neural network and adopting a planar ultrasonic phased array module, components or modules in the ultrasonic suspension three-dimensional control system are selected from the following types and specifications:
a gating device: in order to realize the control of a wired pin on 64 independent delay 40kHz signal generation modules, a 38 decoder is expanded to a 7-line 128-bit decoder, and 64 generation units are subjected to time-sharing control through a chip selection terminal CS.
The planar ultrasonic phased array module consists of 64 ultrasonic vibrators with the diameter of 10 mm.
(3) The control circuit unit selects a specific delay timer according to the received phase information and a gating device in the control circuit unit according to the input information, and sends a delay signal corresponding to the delay timer, the delay timer delays according to the received delay signal, the signal generator is started after the delay is finished, and finally square wave signals with different phases are sent from different signal generator pins, wherein the frequency of the square wave signals is 40 KHz.
The above-mentioned embodiments are only for illustrating the technical ideas and features of the present invention, and the purpose thereof is to enable those skilled in the art to understand the contents of the present invention and to carry out the same, and the present invention shall not be limited to the embodiments, i.e. the equivalent changes or modifications made within the spirit of the present invention shall fall within the scope of the present invention.
Claims (10)
1. An ultrasonic levitation three-dimensional manipulation control method based on a neural network is characterized by comprising the following steps: generating a plurality of groups of random phase distribution data by using MATLAB software, generating sound potential trap depth distribution data corresponding to the phase distribution data one by one, collecting the phase distribution data and the sound potential trap depth distribution data and establishing a training sample set; constructing a neural network model and training; setting a pre-manipulated path and decomposing the pre-manipulated path into unit paths in the x direction, the y direction and the z direction, and constructing a sound potential well depth distribution matrix for realizing the movement of the unit paths so that the sound potential well depth distribution matrix meets the following requirements: generating gradient near the point to be controlled and maximizing the depth of the potential well of the point to be controlled; inputting the depth distribution matrix of the sound potential well into the three trained neural network models simultaneously; and the three neural network models respectively output a phase control matrix corresponding to the phase control of the ultrasonic phased array in the x direction, the y direction and the z direction.
2. The ultrasonic levitation three-dimensional manipulation control method based on the neural network as claimed in claim 1, wherein the neural network model is constructed based on a U-net neural network model, and the neural network model comprises two parts: the first half part comprises a convolution layer and a pooling layer and is used for carrying out feature extraction on input information and reducing the dimensionality of an input matrix; the second half includes an inverse convolutional layer to unwind the data features and increase the dimensionality of the output matrix.
3. The ultrasonic levitation three-dimensional manipulation control method based on the neural network as claimed in claim 1, wherein the training method of the neural network model comprises: taking a part of data in the training sample set as a training set, and taking a part of data in the training sample set as a verification set; inputting data in a training set into the neural network model for training, verifying the data in a verification set after each round of training and calculating loss; and taking the trained neural network model with the lowest loss in verification as the trained neural network model.
4. The ultrasonic levitation three-dimensional manipulation control method based on the neural network as claimed in claim 1, wherein the method for generating random phase distribution data and corresponding sound potential well depth distribution data by using MATLAB software comprises: and generating a random phase distribution matrix by using a rand function, inputting the phase distribution matrix into MATLAB software to calculate a sound pressure field, and calculating a sound potential well depth distribution matrix by using the sound pressure field so as to obtain random phase distribution data and corresponding sound potential well depth distribution data.
5. The ultrasonic levitation three-dimensional manipulation control method based on the neural network as claimed in claim 1, wherein when the neural network model is trained, a loss function is a mean square error loss function, and a gradient descent momentum method is adopted to optimize parameters of the neural network model.
6. The ultrasonic levitation three-dimensional manipulation control method based on the neural network as claimed in claim 1, wherein the method for constructing the sound potential trap depth distribution matrix for realizing the unit path movement comprises: the wave path of the sound wave sent by each vibrator at the point is consistent, namely sound focusing is carried out, so that the corresponding sound potential at the point to be suspended reaches the peak value, and a sound potential matrix is obtained; multiplying the sound potential matrix by-1 to maximize the depth of the sound potential trap at the point to be suspended to obtain a target matrix, calculating the sound potential matrix at different heights during sound focusing, and multiplying the sound potential matrix by-1 to generate a sound potential trap depth distribution matrix.
7. An ultrasonic levitation three-dimensional manipulation control system based on a neural network is characterized by comprising an artificial intelligence system serving as an upper computer and a driving system serving as a lower computer; the artificial intelligence system comprising a memory, a processor and a computer program stored on the memory and executable on the processor, the processor implementing the method steps of any of claims 1 to 6 when executing the program; the driving system comprises a control module and an ultrasonic phased array module driven by the control module; the processor outputs a phase control matrix to the control module; the control module generates a drive signal to drive the ultrasonic phased array module to produce a sound field.
8. The ultrasonic levitation three-dimensional manipulation control system based on the neural network as claimed in claim 7, wherein the control module is a control module based on a digital register signal or a control module based on an analog oscillation signal; wherein:
the control module based on the digital register signal comprises a control board and a drive board A, the processor outputs a control matrix to the control board, and the control board generates a square wave signal to the drive board A; the driving board A supplies power to the control board, amplifies the power of the square wave signal, and outputs the voltage peak value to the ultrasonic phased array module after the voltage peak value reaches the optimal working value of the ultrasonic vibrator;
the control module based on the analog oscillation signal comprises a master control board, a control circuit unit and a drive board B; the control circuit unit comprises a gate, a delay timer and a signal generator; the processor outputs a control matrix to the master control board; the master control board selects the delay timer through the gate, writes delay time into a data register in the delay timer, sends a signal to start the delay timer, and sends a signal to start the signal generator when the delay time is up; the signal generator starts to generate square wave signals and outputs the square wave signals to the driving plate B, and the driving plate B amplifies the power of the received square wave signals and outputs the square wave signals to the ultrasonic phased array module.
9. The neural network-based ultrasonically suspended three-dimensional steering control system according to claim 7, wherein the ultrasonic phased array module is a biconcave ultrasonic phased array module or a planar ultrasonic phased array module.
10. The ultrasonic levitation three-dimensional manipulation control system based on the neural network is characterized in that the biconcave ultrasonic phased array module is composed of 36-256 ultrasonic vibrators, and the diameter of each ultrasonic vibrator is 6-15 mm.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911008064.5A CN110806260B (en) | 2019-10-22 | 2019-10-22 | Ultrasonic levitation three-dimensional manipulation control method and system based on neural network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911008064.5A CN110806260B (en) | 2019-10-22 | 2019-10-22 | Ultrasonic levitation three-dimensional manipulation control method and system based on neural network |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110806260A true CN110806260A (en) | 2020-02-18 |
| CN110806260B CN110806260B (en) | 2020-10-23 |
Family
ID=69488976
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201911008064.5A Active CN110806260B (en) | 2019-10-22 | 2019-10-22 | Ultrasonic levitation three-dimensional manipulation control method and system based on neural network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110806260B (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111541048A (en) * | 2020-04-29 | 2020-08-14 | 西南电子技术研究所（中国电子科技集团公司第十研究所） | Terahertz active phased array antenna |
| CN111547515A (en) * | 2020-04-10 | 2020-08-18 | 天津大学 | Broadband ultrasonic suspension device and manufacturing method thereof |
| CN112108263A (en) * | 2020-09-09 | 2020-12-22 | 海略(连云港)科技有限公司 | Fluid suspended matter classification system and method based on vortex filter and 3D U-Net network |
| WO2022104529A1 (en) * | 2020-11-17 | 2022-05-27 | 深圳先进技术研究院 | Generating method and generating system for acoustic tweezers |
Citations (16)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US3153304A (en) * | 1960-09-30 | 1964-10-20 | Owens Corning Fiberglass Corp | Acoustical suspended ceiling |
| GB1254037A (en) * | 1968-10-30 | 1971-11-17 | Euphonios Corp | Ultrasonic transducer employing suspended piezoelectric plate |
| US4218921A (en) * | 1979-07-13 | 1980-08-26 | The United States Of America As Represented By The Administrator Of The National Aeronautics And Space Administration | Method and apparatus for shaping and enhancing acoustical levitation forces |
| US4284403A (en) * | 1978-10-06 | 1981-08-18 | Intersonics Incorporated | Acoustic levitation and methods for manipulating levitated objects |
| US4402221A (en) * | 1981-06-12 | 1983-09-06 | The United States Of America As Represented By The Administrator Of The National Aeronautics And Space Administration | Acoustic suspension system |
| EP0256238A2 (en) * | 1986-08-16 | 1988-02-24 | Battelle-Institut e.V. | Process and apparatus for producing single crystals |
| US4962330A (en) * | 1989-03-21 | 1990-10-09 | The United States Of America As Represented By The Administrator Of The National Aeronautics And Space Administration | Acoustic transducer apparatus with reduced thermal conduction |
| US5500493A (en) * | 1994-11-02 | 1996-03-19 | Guigne International Ltd. | Acoustic beam levitation |
| CN1998043A (en) * | 2004-05-27 | 2007-07-11 | 布雷克成像有限公司 | Method and apparatus for producing stationary intense wave fields of arbitrary shape |
| EP2096628A1 (en) * | 2008-02-29 | 2009-09-02 | ETH Zurich | Acoustic levitation system |
| US20110314914A1 (en) * | 2009-01-16 | 2011-12-29 | John Francis Gregg | Acoustic oscillator |
| CN104870077A (en) * | 2012-01-31 | 2015-08-26 | 宾夕法尼亚州立大学研究基金会 | Microfluidic manipulation and sorting of particles using tunable standing surface acoustic wave |
| CN109508787A (en) * | 2018-10-16 | 2019-03-22 | 深圳大学 | Neural network model training method and system for ultrasound displacement estimation |
| CN109669485A (en) * | 2019-01-21 | 2019-04-23 | 哈尔滨工业大学（深圳） | Acoustic levitation system and its control method based on ultrasonic array |
| CN110161509A (en) * | 2019-05-08 | 2019-08-23 | 华南理工大学 | A kind of more sound field interference sound suspending devices and sound field switching method |
| CN110651322A (en) * | 2016-12-13 | 2020-01-03 | 超级触觉资讯处理有限公司 | Driving techniques for phased array systems |
-
2019
- 2019-10-22 CN CN201911008064.5A patent/CN110806260B/en active Active
Patent Citations (16)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US3153304A (en) * | 1960-09-30 | 1964-10-20 | Owens Corning Fiberglass Corp | Acoustical suspended ceiling |
| GB1254037A (en) * | 1968-10-30 | 1971-11-17 | Euphonios Corp | Ultrasonic transducer employing suspended piezoelectric plate |
| US4284403A (en) * | 1978-10-06 | 1981-08-18 | Intersonics Incorporated | Acoustic levitation and methods for manipulating levitated objects |
| US4218921A (en) * | 1979-07-13 | 1980-08-26 | The United States Of America As Represented By The Administrator Of The National Aeronautics And Space Administration | Method and apparatus for shaping and enhancing acoustical levitation forces |
| US4402221A (en) * | 1981-06-12 | 1983-09-06 | The United States Of America As Represented By The Administrator Of The National Aeronautics And Space Administration | Acoustic suspension system |
| EP0256238A2 (en) * | 1986-08-16 | 1988-02-24 | Battelle-Institut e.V. | Process and apparatus for producing single crystals |
| US4962330A (en) * | 1989-03-21 | 1990-10-09 | The United States Of America As Represented By The Administrator Of The National Aeronautics And Space Administration | Acoustic transducer apparatus with reduced thermal conduction |
| US5500493A (en) * | 1994-11-02 | 1996-03-19 | Guigne International Ltd. | Acoustic beam levitation |
| CN1998043A (en) * | 2004-05-27 | 2007-07-11 | 布雷克成像有限公司 | Method and apparatus for producing stationary intense wave fields of arbitrary shape |
| EP2096628A1 (en) * | 2008-02-29 | 2009-09-02 | ETH Zurich | Acoustic levitation system |
| US20110314914A1 (en) * | 2009-01-16 | 2011-12-29 | John Francis Gregg | Acoustic oscillator |
| CN104870077A (en) * | 2012-01-31 | 2015-08-26 | 宾夕法尼亚州立大学研究基金会 | Microfluidic manipulation and sorting of particles using tunable standing surface acoustic wave |
| CN110651322A (en) * | 2016-12-13 | 2020-01-03 | 超级触觉资讯处理有限公司 | Driving techniques for phased array systems |
| CN109508787A (en) * | 2018-10-16 | 2019-03-22 | 深圳大学 | Neural network model training method and system for ultrasound displacement estimation |
| CN109669485A (en) * | 2019-01-21 | 2019-04-23 | 哈尔滨工业大学（深圳） | Acoustic levitation system and its control method based on ultrasonic array |
| CN110161509A (en) * | 2019-05-08 | 2019-08-23 | 华南理工大学 | A kind of more sound field interference sound suspending devices and sound field switching method |
Non-Patent Citations (4)
| Title |
|---|
| E H BRANDT: ""Suspended by sound"", 《NATURE》 * |
| 张文兵: ""超声悬浮及转子振动抑制的理论与实验研究"", 《中国优秀硕士学位论文》 * |
| 沈昌乐 等: ""声悬浮技术的发展及应用"", 《现代物理》 * |
| 解文军: ""声悬浮优化设计理论及其应用研究"", 《中国优秀博士学位论文》 * |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111547515A (en) * | 2020-04-10 | 2020-08-18 | 天津大学 | Broadband ultrasonic suspension device and manufacturing method thereof |
| CN111547515B (en) * | 2020-04-10 | 2021-04-02 | 天津大学 | Broadband ultrasonic suspension device and manufacturing method thereof |
| CN111541048A (en) * | 2020-04-29 | 2020-08-14 | 西南电子技术研究所（中国电子科技集团公司第十研究所） | Terahertz active phased array antenna |
| CN111541048B (en) * | 2020-04-29 | 2021-06-15 | 西南电子技术研究所（中国电子科技集团公司第十研究所） | Terahertz active phased array antenna |
| CN112108263A (en) * | 2020-09-09 | 2020-12-22 | 海略(连云港)科技有限公司 | Fluid suspended matter classification system and method based on vortex filter and 3D U-Net network |
| WO2022104529A1 (en) * | 2020-11-17 | 2022-05-27 | 深圳先进技术研究院 | Generating method and generating system for acoustic tweezers |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110806260B (en) | 2020-10-23 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110806260B (en) | Ultrasonic levitation three-dimensional manipulation control method and system based on neural network | |
| Rappel et al. | Mechanisms of cell polarization | |
| Marzo et al. | Ultraino: An open phased-array system for narrowband airborne ultrasound transmission | |
| CN110405343B (en) | Laser welding process parameter optimization method based on Bagging integrated prediction model and particle swarm optimization algorithm | |
| Sontag | Some new directions in control theory inspired by systems biology | |
| White et al. | Three dimensional stochastic reconfiguration of modular robots. | |
| CN109669485B (en) | Acoustic levitation system based on ultrasonic array and control method thereof | |
| CN100573228C (en) | A kind of device that utilizes distorting lens to realize light beam automatic shaping | |
| CN109195717A (en) | The launching beam of ultrasonic transducer two-dimensional array shapes | |
| CN111111585B (en) | Non-contact type multi-focus ultrasonic phased array suspension liquid transfer method and system | |
| US20200061904A1 (en) | Method for 3d printing and 3d printer using ultrasound | |
| TWI550530B (en) | Method, apparatus, computer readable medium, and computer program product for generating compact representations of spike timing-dependent plasticity curves | |
| Borau et al. | Probabilistic Voxel-Fe model for single cell motility in 3D | |
| Reinhart et al. | A constrained regularization approach for input-driven recurrent neural networks | |
| CN114637196A (en) | System and method for micro-object density distribution control by machine learning via a digital computer | |
| US11054910B1 (en) | System and method for producing mid-air tactile stimulation | |
| CN114505105A (en) | Micro-fluidic chip based on memory calculation | |
| CN104985822A (en) | 3d cell printing system and printing method thereof | |
| CN103754820B (en) | Based on sound field synthesis and the parallel operation device of ultrasonic transducer annular array | |
| CN115505524B (en) | Temperature self-adaptive regulation method and device for culture dish, host and storage medium | |
| Asgari et al. | 3D particle-based cell modelling for haptic microrobotic cell injection | |
| CN109917647A (en) | One kind optimizing sliding-mode control based on the improved learning aid algorithm of instructional strategies and filled Spacecraft | |
| CN114624997A (en) | System and method for enabling machine-learned micropackage control with a digital computer | |
| Schakenraad et al. | Stress fibers orient traction forces on micropatterns: A hybrid cellular potts model study | |
| Beasley et al. | Ultrasonic levitation with software-defined FPGAs and electronically phased arrays |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |