CN112001036A - An accelerometer based on artificial intelligence design and distributed manufacturing - Google Patents

Abstract

The invention discloses an accelerometer based on artificial intelligence design and distributed manufacturing. First, a large data set of one-to-one correspondence between the geometric parameters of the device and the response results is generated by a simulation method, and then fed into a two-way artificial intelligence network for Training, and then build a 3D model according to the geometric parameters obtained by the above reverse retrieval, and use a 3D printer controlled and managed by a cloud computing platform that can provide multiple materials at the same time. The electrodes and leads of the device are interconnected, and the temporary support structure of the three-dimensional structure is made of the sacrificial material that can be selectively removed. After the integrated manufacturing is completed, the support structure is selectively removed to obtain a complete device, which reduces the difficulty and cycle of design and manufacture. Personalized and intelligent manufacturing is realized.

Description

An accelerometer based on artificial intelligence design and distributed manufacturing

technical field

The invention relates to the field of electronic technology, in particular to an artificial intelligence design method and a distributed manufacturing method of an accelerometer.

Background technique

Electronic sensors such as accelerometers, pressure sensors, and gyroscopes are widely used in industrial production and people's daily life. The design of these sensors is usually done by technicians with specialized knowledge, usually with isolated modeling and simulation, once the geometry is modified, it must be remodeled, which usually takes several weeks, and it is difficult to write analytical reverse design problems. , that is, determining the geometric parameters based on the response of the device is often difficult to solve effectively. The fabrication methods for these sensors are typically based on silicon-based microelectronics processes such as deposition, lithography, etching, and metallization, which typically take months. The entire design and manufacturing process is completed in a centralized place, with long production cycles, few product models, and high transportation costs... It is difficult to meet users' growing demand for personalized customization.

SUMMARY OF THE INVENTION

The purpose of the present invention is to provide an artificial intelligence-based design method and a distributed manufacturing method, which are applied in the field of design and manufacture of typical device accelerometers. Including artificial intelligence design part and distributed manufacturing part.

The artificial intelligence design part first uses the simulation method to generate a large data set of one-to-one correspondence between the geometric parameters of the device and the response results, and then feeds the two-way artificial intelligence network respectively for training. Once the training is completed, the forward network can be used. In order to quickly predict the response results according to the geometric parameters of the device, the inverse network can be used to quickly retrieve the geometric parameters of the device according to the result requirements.

The distributed manufacturing part is first constructed into a three-dimensional model according to the geometric parameters obtained by the reverse retrieval, and then integrated with a three-dimensional printer controlled and managed by a cloud computing platform that can provide multiple materials at the same time, and the printing materials are at least At the same time, it includes insulating materials, conductive materials, and selectively removable supporting materials. In the printing process, insulating materials are used to make the mechanical structure of the device, and conductive materials are used to make the electrodes and lead interconnects of the device. The temporary support structure of the three-dimensional structure, after the integrated manufacturing is completed, the support structure is selectively removed to obtain a complete device.

The accelerometer has a differential capacitive structure, the artificial intelligence design method is used to establish a bidirectional network of the accelerometer, a three-dimensional model is constructed using the geometric parameters obtained by reverse retrieval, and the insulating material of the multi-nozzle three-dimensional printer is used to print the part of the device. The mechanical structure is obtained by printing the electrodes and leads of the device with the conductive material of the 3D printer, printing the supporting structure of the device with the removable supporting material of the 3D printer, and selectively removing the supporting structure after the manufacturing is completed.

Compared with the prior art, the present invention has the following advantages:

First, the design part of the present invention uses electrical simulation or finite element simulation to obtain a large data set of one-to-one correspondence between the geometric parameters of the device and the response results, and the data is sufficient and accurate. Once the data set is fed into the bidirectional AI network and the training is completed, the designer can use the trained forward network to quickly predict the response result based on the geometric parameters of the device without remodeling even if the geometric parameters are modified; It is difficult to write the analytical formula of the inverse problem with relevant design experience, and the trained inverse network can be used to quickly retrieve the geometric parameters of the device according to the required results.

Second, the manufacturing part of the present invention uses a distributed integrated three-dimensional printing method, and simultaneously prints the mechanical structure, conductive electrodes, and suspension supports of the device, which not only saves materials compared to traditional subtractive manufacturing, but also does not require any alignment during the manufacturing process. and assembly processes, in addition to the ability to fabricate complex geometries that are difficult to achieve with conventional microelectronics processes.

Third, using the artificial intelligence design method and distributed manufacturing method of the present invention in combination, even if the general user does not have the professional knowledge and hands-on manufacturing ability of the relevant device, he only needs to input the personalized demand result into the trained reverse network to quickly Corresponding geometric parameters are retrieved and a 3D model is constructed, which is integrally manufactured by a distributed multi-nozzle 3D printer, which reduces the difficulty and cycle of design and manufacturing, and realizes personalized and intelligent manufacturing.

Description of drawings

1 is a schematic diagram of an artificial intelligence design method and a distributed manufacturing method in the present invention;

2(a)-(c) are schematic diagrams of the structure of the accelerometer in the present invention, wherein (a) is a three-dimensional structural diagram, (b) is a top view of the beam-mass block structure, and (c) is a cross-sectional view;

3(a)～(b) are schematic diagrams of the bidirectional artificial intelligence network of the present invention, wherein (a) is a forward prediction network, and (b) is a reverse retrieval network;

4(a)-(1) are schematic diagrams of the integrated three-dimensional printing process of the present invention;

Fig. 5 is the real photo of the accelerometer of the present invention;

FIG. 6 is the output voltage curve of the accelerometer circuit system of the present invention flipping test in the gravitational field;

FIG. 7 is a dynamic response curve of the accelerometer circuit system of the present invention.

Detailed ways

In order to make the above objects, features and advantages of the present invention more clearly understood, the present invention will be described in further detail below with reference to the accompanying drawings and specific embodiments.

FIG. 1 is a schematic diagram of an artificial intelligence design method and a distributed manufacturing method of the present invention. The artificial intelligence design part is to first use a computer with powerful computing power managed by the cloud platform to perform electrical simulation or finite element simulation on the electronic device model to obtain a large data set of one-to-one correspondence between the geometric parameters of the device and the response results. Then, the large data sets are fed into the two-way artificial intelligence network for training. After the training is completed, the data and network parameters are saved in the cloud computing platform. When in use, the user can input the geometric parameters into the forward network at the user end to quickly get the response result of the device; the user can input the desired result into the reverse network at the user end to quickly obtain the geometric parameters of the device. The distributed manufacturing part is that the user inputs the required results into the reverse network at the user terminal to retrieve the geometric parameters of the device and builds a three-dimensional model, and then uses the distributed three-dimensional printer controlled and managed by the cloud computing platform to integrate manufacturing. Data transmission can use high-speed communication methods (such as 5G) to improve efficiency and reduce latency.

FIG. 2 is a schematic structural diagram of an accelerometer in the present invention. The artificial intelligence design method and distributed manufacturing method of the present invention are applied to the manufacture of the accelerometer, and FIG. 2( a ) shows a three-dimensional structure diagram of the accelerometer. The accelerometer is based on the differential capacitive working principle of the "sandwich" structure, that is, the upper and lower variable parallel plate capacitors form a differential capacitor, wherein the lower capacitor consists of the lower plate 1, the lower plate electrode 2, the lower capacitance gap 3, the beam-mass The block lower electrode 5 and the beam-mass block 6 are composed. The lower support 4 is a temporary structure when manufacturing the lower capacitor, and the whole device is removed after the manufacturing is completed. The upper capacitor is composed of the upper plate 11, the upper plate electrode 10, the upper capacitance gap 8, the beam-mass block upper electrode 7, and the beam-mass block 6. The upper support 9 is a temporary structure when the upper capacitor is manufactured, and the entire device is removed after manufacturing. . In order to achieve the above functions, insulating materials such as polylactic acid (PLA) can be used to manufacture the lower plate 1, upper plate 11, lower gap 3, upper gap 8, beam-mass 6 of the capacitor; conductive materials such as carbon black, The lower plate electrode 2, the upper plate electrode 10, the beam-mass lower electrode 5, the beam-mass upper electrode 7 are made of PLA made of graphene, carbon nanotubes, etc.; soluble high-impact polystyrene (HIPS) or polystyrene can be used. The lower support 4 and the upper support 9 are manufactured from vinyl alcohol (PVA) or the like.

FIG. 2( b ) shows a top view of the beam-mass structures 5 , 6 , and 7 . The mass 12 is suspended by four symmetrically distributed beam springs 13 to form the beam-mass structure of the accelerometer, each beam spring is composed of three L-shaped beams, and each beam has the same length and width. The mass and all spring beams have the same thickness.

Figure 2(c) shows a cross-sectional view of the accelerometer. In order to simplify the model, the thickness of the side wall 14 and the thickness of the plate 15 of the accelerometer are preferably fixed at 2000 μm.

FIG. 3 is a schematic diagram of a bidirectional artificial intelligence network according to the present invention. The geometric parameters of the accelerometer include the length of the spring beam (BL), the width of the spring beam (BW), the thickness of the spring beam-mass block (MT), the side length of the mass block (MW), and the distance between the mass block and the side frame (SD) , plate capacitance gap (CD); the results of the accelerometer are capacitance difference (DC), first-order vibration frequency (F), initial capacitance (C ₀ ), plate displacement (x), overall width (OW), overall Height (OH). The preferred spring beam length (BL) is in the range of 7000-19000 μm, the preferred spring beam width (BW) is in the range of 740-900 μm, the preferred spring beam-mass thickness (MT) is in the range of 1000-3000 μm, and the preferred mass side length (MW) The range of 10000-22000 μm, the preferred range of the distance (SD) between the mass block and the side frame is 5400-7000 μm, and the preferred range of the plate capacitance gap (CD) is 700-900 μm. The combination of 11875 sets of geometric parameters is preferred, the acceleration load of 10m/s ² is applied, and the finite element simulation is carried out to obtain the corresponding results. The data set composed of the normalized geometric parameters and corresponding results is fed into the bidirectional artificial intelligence network described in Figure 3 for training and testing. Figure 3(a) shows the forward prediction network, which predicts the results by geometric parameters. Figure 3(b) shows the reverse retrieval network, that is, the geometric parameters are retrieved from the results. By optimizing the parameters of the bidirectional network, the mean square error of the forward network after training is 4.4×10 ^-7 , and the mean square error of the reverse network after training is 3.1×10 ^-5 .

It is preferable for the user to use the inverse network example. The user needs the accelerometer in the gravitational field to have a capacitance difference (DC) of 0.05pF, a first-order vibration frequency (F) of 220Hz, an initial capacitance (C ₀ ) of 0.05pF, and a mass displacement of 5.4 μm, the overall width (OW) is 38mm, and the overall height (OH) is 7.9mm. Inputting the result value into the trained inverse network can quickly obtain the geometric parameters as the spring beam length (BL) is 17600μm, the spring beam width is 17600μm (BW) is 800 μm, the spring beam-mass thickness (MT) is 2100 μm, the mass side length (MW) is 21000 μm, the distance (SD) between the mass and the side frame is 6500 μm, and the plate capacitance gap (CD) is 900 μm. A three-dimensional digital model of the device is constructed according to the retrieved geometric parameters, and input to the distributed multi-nozzle three-dimensional printer for integrated manufacturing.

FIG. 4 is a schematic diagram of the integrated three-dimensional printing process of the present invention. Figure 4(a) shows the lower plate 1 for printing the accelerometer, preferably material PLA, preferably the printing temperature is 190° C., and the printing speed is preferably 40 mm/s.

Figure 4(b) shows the lower plate electrode 2 for printing the accelerometer, preferably conductive PLA doped with graphene, the preferred printing temperature is 240°C, and the preferred printing speed is 40 mm/s.

Figure 4(c) shows the lower capacitance gap 3 for printing the accelerometer, preferably the material is PLA, the preferred printing temperature is 190°C, and the preferred printing speed is 40 mm/s.

Figure 4(d) shows the lower support 4 for printing the accelerometer, preferably a soluble material HIPS, the preferred printing temperature is 230°C, and the preferred printing speed is 40 mm/s.

Figure 4(e) shows the lower electrode 5 of the beam-mass block for printing the accelerometer, preferably conductive PLA doped with graphene, the preferred printing temperature is 240°C, and the preferred printing speed is 40 mm/s.

Figure 4(f) shows the beam-mass block 6 for printing the accelerometer, preferably the material PLA, the preferred printing temperature is 190°C, and the preferred printing speed is 40 mm/s.

Figure 4(g) shows the upper electrode 7 of the beam-mass block for printing the accelerometer, preferably conductive PLA doped with graphene, the preferred printing temperature is 240°C, and the preferred printing speed is 40 mm/s.

Figure 4(h) shows the upper capacitive gap 8 of the accelerometer being printed, preferably material PLA, preferably the printing temperature is 190°C, and the printing speed is preferably 40mm/s.

Figure 4(i) shows the upper support 9 for printing the accelerometer, preferably a soluble material HIPS, the preferred printing temperature is 230°C, and the preferred printing speed is 40 mm/s.

FIG. 4(j) shows the upper plate electrode 10 of the accelerometer printed, preferably conductive PLA doped with graphene, preferably the printing temperature is 240° C., and the preferred printing speed is 40 mm/s.

Figure 4(k) shows the upper plate 11 for printing the accelerometer, preferably material PLA, preferably the printing temperature is 190°C, and the printing speed is preferably 40 mm/s.

Figure 4(1) shows the accelerometer obtained by selectively dissolving the HIPS of the lower support 4 and the upper support 9, preferably the dissolving solution is D-limonene.

FIG. 5 is a real photo of the accelerometer of the present invention, which realizes the requirement of the user's personalized customization.

FIG. 6 is an output voltage curve of the accelerometer circuit system of the present invention in a flip test in a gravitational field. Connect the differential capacitive accelerometer to the optimal capacitance amplifying readout circuit MS3110, convert and amplify the tiny capacitance change under the action of acceleration into a voltage signal, and use the indexing head to flip in the gravitational field to test positive and negative gravitational acceleration The output voltage of the accelerometer circuitry under load. The obtained accelerometer system has a low nonlinear error of 4.3% and a sensitivity of 75mV/g.

FIG. 7 is a dynamic response curve of the accelerometer circuit system of the present invention. Preferably an adult wears the accelerometer circuitry and runs to test dynamic performance. Figure 7 shows a real-time recording of the output voltage change over two seconds. The dynamic response curve has 6 distinct peaks indicating that the adult has run 6 steps in the past two seconds. It is proved that the described accelerometer has good dynamic performance.

An accelerometer based on artificial intelligence design and distributed manufacturing provided by the present invention has been introduced in detail above. Specific examples are used in this paper to illustrate the principles and implementations of the present invention. The descriptions of the above embodiments are only used for Help to understand the method of the present invention and its core idea; at the same time, for those skilled in the art, according to the idea of the present invention, there will be changes in the specific implementation and application scope. In summary, the content of this specification It should not be construed as a limitation of the present invention.

Claims (4)

1. an accelerometer based on artificial intelligence design and distributed manufacturing, is characterized in that, comprises artificial intelligence design part and distributed manufacturing part, The above artificial intelligence design part is to first use the simulation method to generate a large data set of one-to-one correspondence between the geometric parameters of the device and the response results, and then feed them into the two-way artificial intelligence network for training. Once the training is completed, the forward network can be used to The response results are quickly predicted according to the geometric parameters of the device, and the inverse network can be used to quickly retrieve the geometric parameters of the device according to the result requirements, The above-mentioned distributed manufacturing part is first constructed into a three-dimensional model according to the geometric parameters obtained by the above reverse retrieval, and then the integrated manufacturing is carried out using a three-dimensional printer controlled and managed by a cloud computing platform that can simultaneously provide printing of various materials, and the printing materials are at least At the same time, it includes insulating materials, conductive materials, and selectively removable support materials, and uses insulating materials to manufacture the mechanical structure of the device during integrated printing. The temporary support structure of the three-dimensional structure is manufactured, and the support structure is selectively removed after the integrated manufacturing is completed to obtain a complete device, The above-mentioned accelerometer is firstly trained on a bidirectional network according to the above artificial intelligence design method and retrieved to obtain geometric parameters on demand, then a three-dimensional model is constructed and integrated according to the above-mentioned distributed manufacturing method, and finally the supporting structure is selectively removed to obtain a complete accelerometer. device. 2. artificial intelligence design method according to claim 1, is characterized in that, comprises the following steps: Data simulation step: use a computer with powerful computing power managed by the cloud platform to perform electrical simulation or finite element simulation on the electronic device model to obtain a large data set that corresponds one-to-one between the geometric parameters of the device and the response results; Two-way network training step: the two-way artificial neural network includes a forward network that predicts the response results from the geometric parameters of the device and a reverse network that retrieves the geometric parameters from the response results of the device, and the big data obtained in the above data simulation steps is used when training the network. The set is respectively fed into the forward network and the reverse network for training, and the data and network parameters are saved in the cloud computing platform after the training is completed; Two-way network calling steps: Input the geometric parameters into the trained forward network on the user side to quickly get the response results of the device, input the desired results into the reverse network to quickly obtain the geometric parameters of the device, and use high-speed communication for data transmission to improve efficiency. Reduce latency. 3. The distributed manufacturing method according to claim 1, characterized in that, comprising the steps of: 3D model construction step: input the required results into the above-mentioned reverse network retrieval on the user side to obtain the geometric parameters of the device, and build a 3D model for 3D printing; The integrated 3D printing step: the 3D model constructed by the above method is printed and manufactured in an integrated manner using a distributed 3D printer controlled and managed by a cloud computing platform. The multi-nozzle 3D printer can at least simultaneously print insulating materials and conductive materials, which can be selectively removed. support material. 4. The accelerometer according to claim 1, characterized in that the accelerometer has a differential capacitive structure, preferably a spring beam-mass block structure as a sensitive unit and two upper and lower cover plates respectively constitute two upper and lower variable flat plates capacitor, use the artificial intelligence design method to build the bidirectional network of the accelerometer, build a three-dimensional model with the geometric parameters obtained by reverse retrieval, use the insulating material of the three-dimensional printer to print the mechanical structure of the device, use the electrical conductivity of the three-dimensional printer The electrodes and leads of the material-printed device are interconnected, and the support structure of the device is printed with the selectively removable support material of the three-dimensional printer, and the support structure is selectively removed after the manufacturing is completed without affecting the mechanical structure and the interconnection of electrodes and leads, obtaining final device.

Images