Grating intelligent design method and device applied to terahertz wave band
By employing intelligent design methods based on artificial neural networks and deep neural networks, combined with femtosecond laser processing technology, the problems of high cost and low efficiency in grating design and fabrication have been solved, enabling rapid and low-cost fabrication of gratings.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- BEIJING UNIV OF TECH
- Filing Date
- 2023-09-08
- Publication Date
- 2026-06-26
Smart Images
Figure CN117270093B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the fields of optical design and laser precision machining technology, and in particular to an intelligent design method and device for gratings applied in the terahertz band. Background Technology
[0002] Terahertz waves lie between infrared and microwave radiation in the electromagnetic spectrum, with wavelengths ranging from 30 μm to 3000 μm. They are a type of electromagnetic wave that is harmless and non-invasive to the human body. Terahertz waves have applications in optical imaging, data transmission, medical diagnosis, and materials characterization. Therefore, the manipulation of terahertz waves is of great significance and has broad application value. In some studies, gratings are used to manipulate terahertz waves. A grating is a commonly used optical manipulation device, consisting of a large number of parallel slits of equal width and spacing, fabricated on a dielectric substrate through etching and other processes.
[0003] With technological advancements, higher demands are being placed on grating manufacturing. Femtosecond laser processing technology, as a high-precision machining method, is increasingly being applied to the fabrication of terahertz devices due to its extremely high peak power and low thermal effects. Especially in the processing of materials at the micrometer scale, femtosecond laser processing technology offers exceptional controllability, better meeting the requirements for terahertz wave manipulation.
[0004] In existing technologies, grating fabrication faces two major challenges: grating structural design and grating fabrication parameter design. Currently, the most common method for grating structural design is simulation, which involves simulating possible structures and continuously adjusting the design to match the required optical properties. Similarly, grating fabrication parameter design also requires constant adjustment of parameters such as laser flux, number of scans, and scanning speed to fabricate specific grating structures. Both of these steps require significant time for trial and error, resulting in high costs and low efficiency. Summary of the Invention
[0005] This invention provides an intelligent design method and apparatus for gratings in the terahertz band, which solves the problems of high cost and low efficiency in the design and fabrication of gratings in the prior art, and realizes intelligent, efficient and low cost grating design.
[0006] This invention provides a method for intelligent grating design applied in the terahertz band, comprising:
[0007] Obtain the input data and design goals;
[0008] If the input data is laser processing parameters and the design target is optical properties, the laser processing parameters are input into a pre-built forward prediction model to obtain the grating structure prediction result; the grating structure prediction result is input into a pre-built forward design model to obtain the optical property design result.
[0009] If the input data is optical property data and the design target is laser processing parameters, the optical property data is input into a pre-built reverse design model to obtain the grating structure prediction result; the grating structure prediction result is input into a pre-built reverse prediction model to obtain the laser processing parameter design result.
[0010] The forward prediction model and the reverse prediction model are obtained by training an artificial neural network using a database of laser processing parameters and processing material grating structures; the forward design model and the reverse design model are obtained by training a deep neural network using a database of grating structures and optical properties.
[0011] According to the present invention, an intelligent grating design method for the terahertz band is provided, which acquires the input data and design objectives, and then further includes:
[0012] If the input data is grating structure data and the design goal is optical properties, the grating structure data is input into the forward design model to obtain the optical property prediction results.
[0013] According to the present invention, an intelligent grating design method for the terahertz band is provided, which acquires the input data and design objectives, and then further includes:
[0014] If the input data is grating structure data and the design target is laser processing parameters, the grating structure data is input into the inverse prediction model to obtain the laser processing parameter prediction results.
[0015] The present invention provides an intelligent grating design method for the terahertz band, which, based on an artificial neural network, uses a database of laser processing parameters and processing materials to train a forward prediction model and a backward prediction model, specifically including:
[0016] A massive database of laser processing parameters and grating structure samples of processed materials is constructed by acquiring these parameters. The database includes laser processing parameter samples for processing grating structure samples, processing material samples for the grating structure samples, and grating structure samples.
[0017] Using the laser processing parameter-processing material grating structure database, a forward prediction model from laser processing parameters to grating structure is obtained based on a pre-built artificial neural network; using the same database, a reverse prediction model from grating structure to laser processing parameters is obtained based on an artificial neural network.
[0018] The grating structure sample includes at least the grating shape and size, and the processing material sample includes at least one of Si, SiC, GaSb, GaAs, GaN, InSb, InAsSb, HgCdTe, MnCoNi, and VOx.
[0019] According to the present invention, an intelligent grating design method for the terahertz band is provided, which obtains a forward design model and a reverse design model by training a deep neural network using a grating structure-optical property database. Specifically, it includes:
[0020] Obtain a massive number of grating structure samples;
[0021] The modulation characteristics of the terahertz band light of the grating structure sample were simulated sequentially using the finite element analysis method to obtain a large number of optical characteristic samples.
[0022] A grating structure-optical property database was constructed using all the grating structure samples and their corresponding optical property samples.
[0023] A forward design model from grating structure to optical properties is obtained by training a pre-built deep neural network using the grating structure-optical property database; a reverse design model from optical properties to grating structure is obtained by training a deep neural network using the grating structure-optical property database.
[0024] The optical property samples include at least reflection, transmission, absorption, polarization, phase, and amplitude, and the grating structure samples include at least the grating shape and size.
[0025] According to the present invention, a method for intelligent design of gratings applied in the terahertz band is provided, wherein the number of hidden layers in the deep neural network is 5-10, and the number of neurons in each hidden layer is 20-256.
[0026] The present invention also provides an intelligent grating design device for the terahertz band, comprising:
[0027] The acquisition unit is used to acquire the input data and design goals.
[0028] The first prediction unit is configured to, if the input data is laser processing parameters and the design target is optical properties, input the laser processing parameters into a pre-built forward prediction model to obtain a grating structure prediction result; and input the grating structure prediction result into a pre-built forward design model to obtain an optical property design result.
[0029] The second prediction unit is used to input the optical characteristic data into a pre-built reverse design model to obtain the grating structure prediction result if the input data is optical characteristic data and the design target is laser processing parameters; and to input the grating structure prediction result into a pre-built reverse prediction model to obtain the laser processing parameter design result.
[0030] The forward prediction model and the reverse prediction model are obtained by training an artificial neural network using a database of laser processing parameters and processing material grating structures; the forward design model and the reverse design model are obtained by training a deep neural network using a database of grating structures and optical properties.
[0031] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the intelligent grating design method applied to the terahertz band as described above.
[0032] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the intelligent grating design method applied to the terahertz band as described above.
[0033] The present invention also provides a computer program product, including a computer program that, when executed by a processor, implements the intelligent grating design method applied to the terahertz band as described above.
[0034] This invention provides an intelligent grating design method and apparatus applied to the terahertz band. The method involves acquiring input data and a design target. If the input data is laser processing parameters and the design target is optical properties, the laser processing parameters are input into a pre-built forward prediction model to obtain a grating structure prediction result. The grating structure prediction result is then input into a pre-built forward design model to obtain an optical property design result. If the input data is optical property data and the design target is laser processing parameters, the optical property data is input into a pre-built reverse design model to obtain a grating structure prediction result. The grating structure prediction result is then input into a pre-built reverse prediction model to obtain a laser processing parameter design result. The forward and reverse prediction models are obtained by training an artificial neural network using a database of laser processing parameters and processing materials grating structures. The forward and reverse design models are obtained by training a deep neural network using a database of grating structures and optical properties. This invention utilizes four pre-built sub-models to combine deep learning with the manufacturing and design of gratings, enabling intelligent, efficient, and low-cost forward rapid prediction and reverse intelligent design across the entire process from process to device to performance. Attached Figure Description
[0035] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.
[0036] Figure 1 This is one of the flowcharts illustrating the intelligent grating design method for the terahertz band provided by this invention;
[0037] Figure 2 This is the second flowchart of the intelligent grating design method for the terahertz band provided by the present invention;
[0038] Figure 3 This is a schematic diagram of a grating structure according to an embodiment of the intelligent grating design method for the terahertz band provided by the present invention;
[0039] Figure 4 This is a schematic diagram of the intelligent grating design device for the terahertz band provided by the present invention;
[0040] Figure 5 This is a schematic diagram of the structure of the electronic device provided by the present invention.
[0041] Figure label:
[0042] 410: Acquisition unit; 420: First prediction unit; 430: Second prediction unit;
[0043] 510: Processor; 520: Communication interface; 530: Memory; 540: Communication bus. Detailed Implementation
[0044] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.
[0045] The following is combined with Figures 1-3 The present invention describes an intelligent grating design method applied to the terahertz band, such as... Figure 1 As shown, Figure 1 This is one of the flowcharts illustrating the intelligent grating design method for the terahertz band provided by the present invention, which includes the following steps:
[0046] Step 110: Obtain the input data and design goals.
[0047] The acquired data includes the design goals of the grating and the corresponding input data. Existing technologies only consider the design of one stage of the manufacturing process, and generally employ a trial-and-error approach for grating design. Therefore, this invention proposes an intelligent design method based on a neural network algorithm, such as... Figure 2 As shown, this invention aims to achieve both rapid simulation of grating structures and optical properties under specific processing parameters, and rapid prediction of grating structures with specific optical properties and design of processing parameters. Finally, based on the design results, femtosecond lasers are used for rapid large-area fabrication. This invention integrates artificial intelligence technology into all aspects of process, device, and performance, proposing an integrated intelligent solution from process to device to performance.
[0048] Step 120: If the input data is laser processing parameters and the design target is optical properties, input the laser processing parameters into a pre-built forward prediction model to obtain the grating structure prediction result; input the grating structure prediction result into a pre-built forward design model to obtain the optical property design result.
[0049] If the input data is laser processing parameters and the design objective is optical properties, then the selected laser processing parameters are used to predict the grating structure and optical properties processed using those parameters. In practice, the acquired laser processing parameters are first input into the forward prediction model to obtain the grating structure prediction result, and then the grating structure prediction result is input into the forward design model to obtain the optical property design result.
[0050] It should be noted that in some embodiments, the laser processing parameters include at least laser flux, repetition rate, scanning speed, and number of scans. Furthermore, in some embodiments, a femtosecond laser is used to process the grating, and the range of laser processing parameters is as follows: femtosecond laser flux is 2 J / cm². 2 -10J / cm 2 The scanning speed is 100mm / s-1000mm / s, the number of scans is 10-100, and the repetition rate is 10kHz-100kHz.
[0051] Furthermore, the grating structure prediction results mentioned in this invention include shape structure prediction results and size structure prediction results. In some embodiments, the shape structure prediction result is a sawtooth grating, such as... Figure 3 As shown, the dimensional structure prediction results include: width, height, repeating spacing, etc. It should be noted that in one embodiment, the range of the dimensional structure prediction results is as follows: width 10μm-100μm, height 0.5μm-100μm, and the ratio of structural repeating spacing to width is 1-2.
[0052] It should be understood that the materials used to process the grating described in this invention include at least: Si, SiC, GaSb, GaAs, GaN, InSb, InAsSb, HgCdTe, MnCoNi, and Vox. In one embodiment, the optical properties in the optical property design result include: reflection, transmission, absorption, polarization, phase, and amplitude.
[0053] Among them, the forward prediction model is obtained by training an artificial neural network using a database of laser processing parameters and processing material grating structures; the forward design model is obtained by training a deep neural network using a database of grating structures and optical properties.
[0054] Furthermore, after obtaining the optical characteristic design results, the process includes: comparing the optical characteristic design results with the target optical characteristics; if they are the same as the target optical characteristics, then using a femtosecond laser processing method to rapidly and extensively fabricate the grating based on the laser processing parameters included in the input data; if they are the same as the target optical characteristics, then the laser processing parameters are re-acquired for prediction and design.
[0055] Step 130: If the input data is optical characteristic data and the design target is laser processing parameters, input the optical characteristic data into the pre-built reverse design model to obtain the grating structure prediction result; input the grating structure prediction result into the pre-built reverse prediction model to obtain the laser processing parameter design result.
[0056] If the input data is optical property data, and the design target is laser processing parameters, then the required grating structure and its laser processing parameters are designed based on specific optical property data. In the specific implementation process, the acquired optical property data is first input into the reverse design model to obtain the grating structure prediction result; then the grating structure prediction result is input into the reverse prediction model to obtain the laser processing parameter design result.
[0057] It should be noted that in some embodiments, the optical properties in the optical property data include: reflection, transmission, absorption, polarization, phase, amplitude, etc. Furthermore, the grating structure prediction results mentioned in this invention include shape structure prediction results and size structure prediction results. In some embodiments, the shape structure prediction result is a sawtooth grating, such as... Figure 3 As shown, the dimensional structure prediction results include: width, height, repeating spacing, etc. It should be noted that in one embodiment, the range of the dimensional structure prediction results is as follows: width 10μm-100μm, height 0.5μm-100μm, and the ratio of structural repeating spacing to width is 1-2.
[0058] It should be understood that the materials used to process the grating described in this invention include at least: Si, SiC, GaSb, GaAs, GaN, InSb, InAsSb, HgCdTe, MnCoNi, and VOx. The laser processing parameter design results include at least laser flux, repetition rate, scanning speed, and number of scans. Further, in some embodiments, the grating is processed using a femtosecond laser based on the laser processing parameter design results. The range of the laser processing parameter design results is as follows: femtosecond laser flux is 2 J / cm². 2 -10J / cm 2 The scanning speed is 100mm / s-1000mm / s, the number of scans is 10-100, and the repetition rate is 10kHz-100kHz.
[0059] Among them, the reverse prediction model is obtained by training an artificial neural network using a database of laser processing parameters and processing material grating structures, while the reverse design model is obtained by training a deep neural network using a database of grating structures and optical properties.
[0060] Furthermore, after obtaining the laser processing parameter design results, the process also includes: rapidly and extensively fabricating the grating using a femtosecond laser processing method based on the laser processing parameter design results. Because femtosecond laser processing is faster than other methods such as ion beam etching, it is more suitable for large-area fabrication in industry, and therefore this invention has wide industrial applications.
[0061] Furthermore, in some embodiments, the gratings involved in this invention can be used to modulate terahertz waves in the terahertz band of 1THz-10THz.
[0062] Based on the above embodiments, the method, after obtaining the input data and design goals, further includes:
[0063] If the input data is grating structure data and the design goal is optical properties, the grating structure data is input into the forward design model to obtain the optical property prediction results.
[0064] Specifically, if the input data is grating structure data and the design objective is optical properties, that is, to predict the possible optical properties of the grating structure based on specific grating structure data, then in the implementation process, the acquired grating structure data is input into the forward design model to obtain the predicted optical properties.
[0065] It should be noted that in some embodiments, the grating structure data includes shape structure data and size structure data. In some embodiments, the shape structure data is a sawtooth grating, such as... Figure 3 As shown, the dimensional structure data includes: width, height, repeating spacing, etc. It should be noted that in one embodiment, the dimensional structure data ranges as follows: width 10μm-100μm, height 0.5μm-100μm, and the ratio of repeating spacing to width is 1-2. The optical properties predicted in the optical property prediction results include: reflection, transmission, absorption, polarization, phase, amplitude, etc.
[0066] The forward design model is based on a deep neural network trained using a grating structure-optical property database.
[0067] Based on the above embodiments, the method, after obtaining the input data and design goals, further includes:
[0068] If the input data is grating structure data and the design target is laser processing parameters, the grating structure data is input into the inverse prediction model to obtain the laser processing parameter prediction results.
[0069] Specifically, if the input data is grating structure data, and the design objective is laser processing parameters, then the laser processing parameters are designed based on the specific grating structure data. In practice, the acquired grating structure data is input into the inverse prediction model to obtain the predicted laser processing parameters.
[0070] It should be noted that in some embodiments, the grating structure data includes shape structure data and size structure data. In some embodiments, the shape structure data is a sawtooth grating, such as... Figure 3 As shown, the dimensional structure data includes width, height, and repeat interval. It should be noted that in one embodiment, the range of the dimensional structure data is as follows: width 10μm-100μm, height 0.5μm-100μm, and the ratio of the repeat interval to the width is 1-2. The laser processing parameter prediction results include at least laser flux, repetition rate, scanning speed, and number of scans. Further, in some embodiments, the grating is processed using a femtosecond laser based on the laser processing parameter prediction results. The range of the laser processing parameter prediction results is as follows: femtosecond laser flux of 2J / cm². 2 -10J / cm 2 The scanning speed is 100mm / s-1000mm / s, the number of scans is 10-100, and the repetition rate is 10kHz-100kHz.
[0071] The reverse prediction model is based on an artificial neural network trained using a database of laser processing parameters and processing material grating structures.
[0072] Based on the above embodiments, in this method, a forward prediction model and a backward prediction model are obtained by training an artificial neural network using a database of laser processing parameters and processing material grating structures, specifically including:
[0073] A massive database of laser processing parameters and grating structure samples of processed materials is constructed by acquiring these parameters. The database includes laser processing parameter samples for processing grating structure samples, processing material samples for the grating structure samples, and grating structure samples.
[0074] Using the laser processing parameter-processing material grating structure database, a forward prediction model from laser processing parameters to grating structure is obtained based on a pre-built artificial neural network; using the same database, a reverse prediction model from grating structure to laser processing parameters is obtained based on an artificial neural network.
[0075] The grating structure sample includes at least the grating shape and size, and the processing material sample includes at least one of Si, SiC, GaSb, GaAs, GaN, InSb, InAsSb, HgCdTe, MnCoNi, and VOx.
[0076] Specifically, during the training process of the forward prediction model and the backward prediction model, a laser processing parameter-processing material grating structure database is first constructed. In one embodiment, the laser processing parameter-processing material grating structure database is a femtosecond laser processing parameter-processing material grating structure database.
[0077] In practical operation, the laser processing parameter-processing material grating structure database is composed of a massive number of laser processing parameter-processing material grating structure samples. These samples include laser processing parameter samples for processing grating structure samples, processing material samples for the grating structure samples, and grating structure samples. It is important to note that the laser processing parameter samples include: laser flux, repetition rate, scanning speed, and number of scans; the grating structure samples are sawtooth gratings, and the grating structure samples include: width, height, repetition spacing, etc.; the processing material samples include at least one of: Si, SiC, GaSb, GaAs, GaN, InSb, InAsSb, HgCdTe, MnCoNi, and VOx.
[0078] Secondly, a pre-built artificial neural network is trained using a laser processing parameter-processing material grating structure database. In one specific embodiment, the laser processing parameter-processing material grating structure database contains 5000 sets of data. The pre-built artificial neural network is a feedforward neural network with 5-10 layers, each containing 20-256 neurons. The input is the laser processing parameters, the output is the grating structure, and the loss function is the mean squared error. This training yields a forward prediction model from the laser processing parameters to the grating structure.
[0079] The pre-built artificial neural network is then trained using a database of laser processing parameters and grating structures of processing materials. In one specific embodiment, the database contains 5000 sets of data. The pre-built artificial neural network is a feedforward neural network with 5-10 layers, each containing 20-256 neurons. A genetic algorithm is used to optimize the initial values of the neural network through population iteration. The network input is the grating structure, the output is the laser processing parameters, and the loss function is the mean squared error. This training yields a backpropagation model from the grating structure to the laser processing parameters.
[0080] It should be noted that the training of the forward prediction model and the backward prediction model does not take place in any order. In one embodiment, the forward prediction model and the backward prediction model are trained simultaneously.
[0081] This invention enables rapid prediction of the dimensions of the grating under different laser processing parameters—depth, width, and repeatability—through a forward prediction model; and enables rapid prediction of the laser processing parameters—laser flux, number of scans, and scanning speed—through a backward prediction model.
[0082] Based on the above embodiments, in this method, a forward design model and a reverse design model are obtained by training a deep neural network using a grating structure-optical property database, specifically including:
[0083] Obtain a massive number of grating structure samples;
[0084] The modulation characteristics of the terahertz band light of the grating structure sample were simulated sequentially using the finite element analysis method to obtain a large number of optical characteristic samples.
[0085] A grating structure-optical property database was constructed using all the grating structure samples and their corresponding optical property samples.
[0086] A forward design model from grating structure to optical properties is obtained by training a pre-built deep neural network using the grating structure-optical property database; a reverse design model from optical properties to grating structure is obtained by training a deep neural network using the grating structure-optical property database.
[0087] The optical property samples include at least reflection, transmission, absorption, polarization, phase, and amplitude, and the grating structure samples include at least the grating shape and size.
[0088] Specifically, during the training process of the forward and reverse design models, a grating structure-optical property database is first constructed.
[0089] In practice, the grating structure-optical property database consists of a massive number of grating structure samples and their corresponding optical property samples. The optical property samples include at least reflection, transmission, absorption, polarization, phase, and amplitude, while the grating structure samples include at least the grating shape and size. In some embodiments, the grating shape is a sawtooth grating, and the dimensions include width, height, and repeating spacing.
[0090] Furthermore, the method for obtaining the grating structure sample and its corresponding optical property sample includes the following steps:
[0091] 1) Obtain a massive number of grating structure samples;
[0092] 2) Extract an optical property sample and use the finite element method to simulate the modulation characteristics of the terahertz band light of the modified grating structure sample to obtain the optical property sample.
[0093] 3) Repeat steps 1)-2) until all grating structure samples are traversed to obtain a large number of optical property samples.
[0094] Secondly, a pre-built deep neural network is trained using a grating structure-optical property database. In one specific embodiment, the grating structure-optical property database contains 10,000 sets of data. The pre-built deep neural network is a feedforward neural network with 5-10 layers, each containing 128-256 neurons. The network input is the grating structure, the output is the optical property, and the loss function is the mean squared error. The training yields a forward design model from grating structure to optical property.
[0095] Subsequently, a pre-built deep neural network is trained using a grating structure-optical property database. In one specific embodiment, the grating structure-optical property database contains 10,000 data sets, and the pre-built deep neural network is a feedforward neural network with 5-10 layers, each containing 128-256 neurons. The network input is the optical property, the output is the grating structure, and the loss function is the mean squared error. This training yields a reverse design model from optical performance to grating structure. It is important to note that the training of the forward and reverse design models is not sequential; in one embodiment, both models are trained simultaneously.
[0096] This invention enables rapid prediction of the optical performance of different grating structures through a forward design model, including reflection, transmission, absorption, polarization, amplitude, and phase. Furthermore, this invention enables the design of grating structures with specific optical performance through a reverse design model, including depth, width, and repeatability.
[0097] Based on the above embodiments, in this method, the number of hidden layers in the deep neural network is 5-10, and the number of neurons in each hidden layer is 20-256.
[0098] The intelligent grating design method for the terahertz band provided by this invention includes four sub-models: a forward and reverse design model and a forward and reverse fabrication model. These models enable rapid digital manufacturing and optical performance simulation of grating structures under specific fabrication parameters. Furthermore, they allow for the rapid design of fabrication parameters for grating structures with specific optical performance characteristics. Based on the final design results of the fabrication parameters, gratings can be rapidly and extensively fabricated, achieving rapid forward prediction and intelligent reverse design and manufacturing from process to device to performance. Compared to traditional trial-and-error methods, the method provided by this invention significantly reduces fabrication and design time. Traditional fabrication and design times can range from several hours to several days. In contrast, the method provided by this invention can provide the corresponding structural and manufacturing parameters within minutes or even seconds.
[0099] This invention provides an intelligent grating design method applied to the terahertz band. The method involves acquiring input data and a design objective. If the input data is laser processing parameters and the design objective is optical properties, the laser processing parameters are input into a pre-built forward prediction model to obtain a grating structure prediction result. The grating structure prediction result is then input into a pre-built forward design model to obtain an optical property design result. If the input data is optical property data and the design objective is laser processing parameters, the optical property data is input into a pre-built reverse design model to obtain a grating structure prediction result. The grating structure prediction result is then input into a pre-built reverse prediction model to obtain a laser processing parameter design result. The forward and reverse prediction models are obtained by training an artificial neural network using a database of laser processing parameters and processing materials grating structures. The forward and reverse design models are obtained by training a deep neural network using a database of grating structures and optical properties. This invention utilizes four pre-built sub-models to combine deep learning with the manufacturing and design of gratings, enabling intelligent, efficient, and low-cost forward rapid prediction and reverse intelligent design across the entire process from process to device to performance.
[0100] The intelligent grating design device for the terahertz band provided by the present invention is described below. The intelligent grating design device for the terahertz band described below and the intelligent grating design method for the terahertz band described above can be referred to in correspondence with each other. Figure 4 yes Figure 4 This is a schematic diagram of the intelligent grating design device for the terahertz band provided by the present invention, as shown below. Figure 4 As shown, it includes:
[0101] Acquisition unit 410 is used to acquire the input data and design objectives;
[0102] The first prediction unit 420 is configured to, if the input data is laser processing parameters and the design target is optical properties, input the laser processing parameters into a pre-built forward prediction model to obtain grating structure prediction results; and input the grating structure prediction results into a pre-built forward design model to obtain optical property design results.
[0103] The second prediction unit 430 is used to input the optical characteristic data into a pre-built reverse design model to obtain the grating structure prediction result if the input data is optical characteristic data and the design target is laser processing parameters; and to input the grating structure prediction result into a pre-built reverse prediction model to obtain the laser processing parameter design result.
[0104] The forward prediction model and the reverse prediction model are obtained by training an artificial neural network using a database of laser processing parameters and processing material grating structures; the forward design model and the reverse design model are obtained by training a deep neural network using a database of grating structures and optical properties.
[0105] Based on the above embodiments, the device, after acquiring the input data and design target, further includes:
[0106] If the input data is grating structure data and the design goal is optical properties, the grating structure data is input into the forward design model to obtain the optical property prediction results.
[0107] Based on the above embodiments, the device, after acquiring the input data and design target, further includes:
[0108] If the input data is grating structure data and the design target is laser processing parameters, the grating structure data is input into the inverse prediction model to obtain the laser processing parameter prediction results.
[0109] Based on the above embodiments, in this device, a forward prediction model and a backward prediction model are obtained by training an artificial neural network using a database of laser processing parameters and processing material grating structures, specifically including:
[0110] A massive database of laser processing parameters and grating structure samples of processed materials is constructed by acquiring these parameters. The database includes laser processing parameter samples for processing grating structure samples, processing material samples for the grating structure samples, and grating structure samples.
[0111] Using the laser processing parameter-processing material grating structure database, a forward prediction model from laser processing parameters to grating structure is obtained based on a pre-built artificial neural network; using the same database, a reverse prediction model from grating structure to laser processing parameters is obtained based on an artificial neural network.
[0112] The grating structure sample includes at least the grating shape and size, and the processing material sample includes at least one of Si, SiC, GaSb, GaAs, GaN, InSb, InAsSb, HgCdTe, MnCoNi, and VOx.
[0113] Based on the above embodiments, in this device, a forward design model and a reverse design model are obtained by training a deep neural network using a grating structure-optical property database, specifically including:
[0114] Obtain a massive number of grating structure samples;
[0115] The modulation characteristics of the terahertz band light of the grating structure sample were simulated sequentially using the finite element analysis method to obtain a large number of optical characteristic samples.
[0116] A grating structure-optical property database was constructed using all the grating structure samples and their corresponding optical property samples.
[0117] A forward design model from grating structure to optical properties is obtained by training a pre-built deep neural network using the grating structure-optical property database; a reverse design model from optical properties to grating structure is obtained by training a deep neural network using the grating structure-optical property database.
[0118] The optical property samples include at least reflection, transmission, absorption, polarization, phase, and amplitude, and the grating structure samples include at least the grating shape and size.
[0119] Based on the above embodiments, in this device, the number of hidden layers in the deep neural network is 5-10, and the number of neurons in each hidden layer is 20-256.
[0120] This invention provides an intelligent grating design device applied to the terahertz band. It acquires input data and a design target. If the input data is laser processing parameters and the design target is optical properties, the laser processing parameters are input into a pre-built forward prediction model to obtain a grating structure prediction result. The grating structure prediction result is then input into a pre-built forward design model to obtain an optical property design result. If the input data is optical property data and the design target is laser processing parameters, the optical property data is input into a pre-built reverse design model to obtain a grating structure prediction result. The grating structure prediction result is then input into a pre-built reverse prediction model to obtain a laser processing parameter design result. The forward and reverse prediction models are obtained by training an artificial neural network using a database of laser processing parameters and processing materials grating structures. The forward and reverse design models are obtained by training a deep neural network using a database of grating structures and optical properties. This invention utilizes four pre-built sub-models to combine deep learning with the manufacturing and design of gratings, enabling intelligent, efficient, and low-cost forward rapid prediction and reverse intelligent design across the entire process from process to device to performance.
[0121] Figure 5 An example is a schematic diagram of the physical structure of an electronic device, such as... Figure 5As shown, the electronic device may include: a processor 510, a communication interface 520, a memory 530, and a communication bus 540, wherein the processor 510, the communication interface 520, and the memory 530 communicate with each other through the communication bus 540. The processor 510 can call logic instructions in the memory 530 to execute a grating intelligent design method applied to the terahertz band. This method includes: acquiring input data and a design target; if the input data is laser processing parameters and the design target is optical characteristics, inputting the laser processing parameters into a pre-built forward prediction model to obtain a grating structure prediction result; inputting the grating structure prediction result into a pre-built forward design model to obtain an optical characteristic design result; if the input data is optical characteristic data and the design target is laser processing parameters, inputting the optical characteristic data into a pre-built reverse design model to obtain a grating structure prediction result; inputting the grating structure prediction result into a pre-built reverse prediction model to obtain a laser processing parameter design result; wherein the forward prediction model and the reverse prediction model are obtained by training an artificial neural network using a laser processing parameter-processing material grating structure database; and the forward design model and the reverse design model are obtained by training a deep neural network using a grating structure-optical characteristic database.
[0122] Furthermore, the logical instructions in the aforementioned memory 530 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, essentially, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.
[0123] On the other hand, the present invention also provides a computer program product, which includes a computer program that can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the intelligent grating design method for the terahertz band provided by the above methods. This method includes: acquiring input data and a design target; if the input data is laser processing parameters and the design target is optical properties, inputting the laser processing parameters into a pre-built forward prediction model to obtain a grating structure prediction result; and inputting the grating structure prediction result into a pre-built forward design model to obtain a grating structure prediction result. Obtain the optical property design result; if the input data is optical property data and the design target is laser processing parameters, input the optical property data into a pre-built reverse design model to obtain the grating structure prediction result; input the grating structure prediction result into a pre-built reverse prediction model to obtain the laser processing parameter design result; wherein, the forward prediction model and the reverse prediction model are obtained by training an artificial neural network using a laser processing parameter-processing material grating structure database; the forward design model and the reverse design model are obtained by training a deep neural network using a grating structure-optical property database.
[0124] In another aspect, the present invention also provides a non-transitory computer-readable storage medium storing a computer program thereon. When executed by a processor, the computer program performs the intelligent grating design method for the terahertz band provided by the methods described above. This method includes: acquiring input data and a design target; if the input data is laser processing parameters and the design target is optical properties, inputting the laser processing parameters into a pre-built forward prediction model to obtain a grating structure prediction result; inputting the grating structure prediction result into a pre-built forward design model to obtain an optical property design result; if the input data is optical property data and the design target is laser processing parameters, inputting the optical property data into a pre-built reverse design model to obtain a grating structure prediction result; inputting the grating structure prediction result into a pre-built reverse prediction model to obtain a laser processing parameter design result; wherein the forward prediction model and the reverse prediction model are obtained by training an artificial neural network using a laser processing parameter-processing material grating structure database; and the forward design model and the reverse design model are obtained by training a deep neural network using a grating structure-optical property database.
[0125] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.
[0126] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.
[0127] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.
Claims
1. A method for intelligent grating design applied in the terahertz band, characterized in that, include: Obtain the input data and design goals; If the input data is laser processing parameters and the design target is optical properties, the laser processing parameters are input into a pre-built forward prediction model to obtain the grating structure prediction result; The predicted results of the grating structure are input into a pre-built forward design model to obtain the optical property design results; If the input data is optical property data and the design target is laser processing parameters, the optical property data is input into a pre-built reverse design model to obtain the grating structure prediction result; the grating structure prediction result is input into a pre-built reverse prediction model to obtain the laser processing parameter design result. The forward prediction model and the reverse prediction model are obtained by training an artificial neural network using a database of laser processing parameters and processing material grating structures; the forward design model and the reverse design model are obtained by training a deep neural network using a database of grating structures and optical properties. After obtaining the input data and design goals, the following steps are also included: If the input data is grating structure data and the design goal is optical properties, the grating structure data is input into the forward design model to obtain the optical property prediction result; After obtaining the input data and design goals, the following steps are also included: If the input data is grating structure data and the design target is laser processing parameters, the grating structure data is input into the inverse prediction model to obtain the laser processing parameter prediction results.
2. The intelligent grating design method applied to the terahertz band according to claim 1, characterized in that, Based on an artificial neural network, a forward prediction model and a backward prediction model are obtained by training using a database of laser processing parameters and processing material grating structures. Specifically, these include: A massive database of laser processing parameters and grating structure samples of processed materials is constructed by acquiring these parameters. The database includes laser processing parameter samples for processing grating structure samples, processing material samples for the grating structure samples, and grating structure samples. Using the laser processing parameter-processing material grating structure database, a forward prediction model from laser processing parameters to grating structure is obtained based on a pre-built artificial neural network; using the same database, a reverse prediction model from grating structure to laser processing parameters is obtained based on an artificial neural network. The grating structure sample includes at least the grating shape and size, and the processing material sample includes at least one of Si, SiC, GaSb, GaAs, GaN, InSb, InAsSb, HgCdTe, MnCoNi, and VOx.
3. The intelligent grating design method applied to the terahertz band according to claim 1, characterized in that, The forward and reverse design models are obtained by training a deep neural network using a grating structure-optical property database, specifically including: Obtain a massive number of grating structure samples; The modulation characteristics of the terahertz band light of the grating structure sample were simulated sequentially using the finite element analysis method to obtain a large number of optical characteristic samples. A grating structure-optical property database was constructed using all the grating structure samples and their corresponding optical property samples. A forward design model from grating structure to optical properties is obtained by training a pre-built deep neural network using the grating structure-optical property database; a reverse design model from optical properties to grating structure is obtained by training a deep neural network using the grating structure-optical property database. The optical property samples include at least reflection, transmission, absorption, polarization, phase, and amplitude, and the grating structure samples include at least the grating shape and size.
4. The intelligent grating design method applied to the terahertz band according to claim 1 or 3, characterized in that, The deep neural network has 5-10 hidden layers, and each hidden layer has 20-256 neurons.
5. A grating intelligent design device applied in the terahertz band, characterized in that, include: The acquisition unit is used to acquire the input data and design goals. The first prediction unit is used to input the laser processing parameters into a pre-built forward prediction model if the input data is laser processing parameters and the design target is optical properties, so as to obtain the grating structure prediction result. The predicted results of the grating structure are input into a pre-built forward design model to obtain the optical property design results; The second prediction unit is used to input the optical characteristic data into a pre-built reverse design model to obtain the grating structure prediction result if the input data is optical characteristic data and the design target is laser processing parameters; and to input the grating structure prediction result into a pre-built reverse prediction model to obtain the laser processing parameter design result. The forward prediction model and the reverse prediction model are obtained by training an artificial neural network using a database of laser processing parameters and processing material grating structures; the forward design model and the reverse design model are obtained by training a deep neural network using a database of grating structures and optical properties. After obtaining the input data and design goals, the following steps are also included: If the input data is grating structure data and the design goal is optical properties, the grating structure data is input into the forward design model to obtain the optical property prediction result; After obtaining the input data and design goals, the following steps are also included: If the input data is grating structure data and the design target is laser processing parameters, the grating structure data is input into the inverse prediction model to obtain the laser processing parameter prediction results.
6. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the intelligent grating design method for the terahertz band as described in any one of claims 1 to 4.
7. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the intelligent grating design method for the terahertz band as described in any one of claims 1 to 4.
8. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements the intelligent grating design method for the terahertz band as described in any one of claims 1 to 4.