Deep learning based nanoscale structure processing prediction and monitoring method and apparatus

Abstract

The application provides a kind of nanoscale structure processing prediction and monitoring method and device based on deep learning, comprising: obtaining laser processing parameters, inputting the laser processing parameters into a laser module to obtain a laser beam;The laser beam is split into a monitoring laser beam and a processing laser beam;Real-time monitoring of the monitoring laser beam is carried out using a pre-constructed online monitoring module to obtain a monitoring result;The monitoring result is input into a pre-trained nanoscale structure topography type and quality prediction model to obtain a nanoscale structure topography type and quality prediction result.The application uses a deep learning algorithm to predict the structure topography and processing quality of the input laser processing parameters, and monitors the state of the system, including spot quality, laser power, etc., in real time while predicting.The application combines offline prediction and online monitoring to achieve low-cost, high-efficiency laser processing parameter selection.

Description

Technical Field

[0001] This invention relates to the field of laser precision machining technology, and in particular to a method and device for predicting and monitoring the machining of nanoscale structures based on deep learning. Background Technology

[0002] By fabricating micro- and nano-structures on material surfaces, numerous material properties can be enhanced, including antibacterial properties, abrasion resistance, and cell compatibility. Furthermore, the optical and wettability properties of the material surface can be controlled, leading to wide applications in both production and daily life. Commonly used fabrication methods include femtosecond laser processing, ion beam etching, and two-photon polymerization.

[0003] Femtosecond lasers possess extremely high peak power and minimal thermal effects, making them a high-precision machining method. Compared to methods such as ion beam etching and two-photon polymerization, they are more efficient and can achieve rapid processing of large-area nanoscale structures using a galvanometer system. However, nanoscale structures with specific morphologies and sizes can only be produced within a certain range of laser processing parameters. Combinations of other parameters can lead to irregularities or structural damage in the resulting nanoscale structures. Furthermore, in laser processing, especially for precision structures, even subtle changes in conditions can affect the processing results. In other words, the morphology and size of nanostructures are greatly influenced by the laser processing parameters and the stability of the system.

[0004] To determine the laser processing parameters required for fabricating target nanoscale structures, current methods primarily rely on trial and error. This involves directly using possible laser processing parameters to fabricate nanoscale structures, adjusting these parameters repeatedly until the target nanoscale structure is obtained. This method of determining the range of processing parameters for a specific nanoscale structure is time-consuming and costly. Furthermore, system instability during the fabrication process can easily lead to deviations in the processing results.

[0005] In summary, existing technologies suffer from high costs and low efficiency when selecting laser processing parameters. Summary of the Invention

[0006] This invention provides a method and apparatus for predicting and monitoring the processing of nanoscale structures based on deep learning, which solves the problems of high cost and low efficiency in the selection of laser processing parameters in the prior art, and realizes the selection of laser processing parameters at low cost and high efficiency.

[0007] This invention provides a deep learning-based method for predicting and monitoring the fabrication of nanoscale structures, comprising:

[0008] The laser processing parameters are obtained and input into the laser module to obtain a laser beam;

[0009] The laser beam is split into a monitoring laser beam and a processing laser beam;

[0010] The monitoring laser beam is monitored in real time using a pre-built online monitoring module to obtain monitoring results, including spot morphology data and laser energy density data.

[0011] The monitoring results are input into a pre-trained nanoscale structure morphology type and quality prediction model to obtain nanoscale structure morphology type and quality prediction results.

[0012] The nanoscale structure morphology type and quality prediction model is based on a deep neural network trained using a training sample database, which includes at least a large number of processing data samples.

[0013] According to the present invention, a method for predicting and monitoring the fabrication of nanoscale structures based on deep learning is provided, which utilizes a pre-built online monitoring module to monitor the monitoring laser beam in real time to obtain monitoring results, specifically including:

[0014] The monitoring laser beam is split into a spot morphology monitoring laser beam and a laser energy density monitoring laser beam;

[0015] The online monitoring module is used to monitor the laser beam for spot morphology monitoring in real time to obtain spot morphology data. The online monitoring module is also used to monitor the laser beam for laser energy density monitoring in real time to obtain laser energy density data.

[0016] The spot morphology data and laser energy density data constitute the monitoring results.

[0017] According to the present invention, a deep learning-based method for predicting and monitoring the fabrication of nanoscale structures includes inputting the monitoring results into a pre-trained model for predicting the morphology type and quality of nanoscale structures to obtain prediction results for the morphology type and quality of nanoscale structures. The method further includes:

[0018] The monitoring results and the predicted morphology type and quality of the nanoscale structure were tested, and erroneous prediction data were extracted.

[0019] The erroneous prediction data is added to the training sample database to update the prediction model for the morphology type and quality of the nanoscale structure.

[0020] According to the present invention, a method for predicting and monitoring the fabrication of nanoscale structures based on deep learning is provided. This method uses a deep neural network trained on a training sample database to obtain a prediction model for the morphology type and quality of nanoscale structures. Specifically, it includes:

[0021] Acquire a large amount of processing data samples of nanoscale structure samples processed under different laser processing conditions. The processing data samples include at least the surface nanomorphology type samples, nanostructure quality samples, laser spot morphology samples and laser processing parameter samples of the nanoscale structure samples.

[0022] The training sample database is constructed using a massive amount of the aforementioned processing data samples;

[0023] The pre-constructed deep neural network is trained using the training sample database to obtain a prediction model for the morphology type and quality of nanoscale structures.

[0024] According to the present invention, a method for predicting and monitoring the processing of nanoscale structures based on deep learning is provided. The deep neural network includes a convolutional neural network and a feedforward neural network. The convolutional neural network is used to extract laser spot morphology samples from the processing data samples, and the feedforward neural network is used to extract laser processing parameter samples from the processing data samples. The convolutional neural network and the feedforward neural network are fused together through a fully connected layer.

[0025] According to the present invention, a method for predicting and monitoring the fabrication of nanoscale structures based on deep learning is provided, wherein the number of convolutional layers in the convolutional neural network is 10-20 and the number of pooling layers is 5-10.

[0026] This invention also provides a deep learning-based device for predicting and monitoring the fabrication of nanoscale structures, comprising:

[0027] The laser unit is used to acquire laser processing parameters and input the laser processing parameters into the laser module to obtain a laser beam;

[0028] The beam splitting unit is used to split the laser beam into a monitoring laser beam and a processing laser beam.

[0029] The monitoring unit is used to monitor the monitoring laser beam in real time using a pre-built online monitoring module to obtain monitoring results, including spot morphology data and laser energy density data.

[0030] The prediction unit is used to input the monitoring results into a pre-trained nanoscale structure morphology type and quality prediction model to obtain nanoscale structure morphology type and quality prediction results.

[0031] The nanoscale structure morphology type and quality prediction model is based on a deep neural network trained using a training sample database, which includes at least a large number of processing data samples.

[0032] 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 deep learning-based nanoscale structure fabrication prediction and monitoring method as described above.

[0033] 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 deep learning-based nanoscale structure fabrication prediction and monitoring method as described above.

[0034] The present invention also provides a computer program product, including a computer program that, when executed by a processor, implements the deep learning-based nanoscale structure fabrication prediction and monitoring method as described above.

[0035] This invention provides a method and apparatus for predicting and monitoring the processing of nanoscale structures based on deep learning. The method involves acquiring laser processing parameters and inputting these parameters into a laser module to obtain a laser beam. The laser beam is then split into a monitoring laser beam and a processing laser beam. A pre-built online monitoring module monitors the monitoring laser beam in real time to obtain monitoring results, including spot morphology data and laser energy density data. These monitoring results are then input into a pre-trained nanoscale structure morphology type and quality prediction model to obtain prediction results for the nanoscale structure morphology type and quality. The nanoscale structure morphology type and quality prediction model is trained using a deep neural network with a training sample database, which includes at least a large number of processing data samples. This invention utilizes a deep learning algorithm to predict the morphology and quality of the processed structure based on the input laser processing parameters. Simultaneously, it monitors the system state, including spot quality and laser power, in real time. This invention combines offline prediction and online monitoring to achieve low-cost and high-efficiency selection of laser processing parameters. Attached Figure Description

[0036] 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.

[0037] Figure 1 This is one of the flowcharts of the deep learning-based nanoscale structure fabrication prediction and monitoring method provided by the present invention;

[0038] Figure 2This is a schematic diagram of the device structure of an embodiment of the deep learning-based nanoscale structure fabrication prediction and monitoring method provided by the present invention;

[0039] Figure 3 This is a schematic diagram of the structure of the deep learning-based nanoscale structure fabrication prediction and monitoring device provided by the present invention.

[0040] Figure 4 This is a schematic diagram of the structure of the electronic device provided by the present invention.

[0041] Figure label:

[0042] 310: Laser unit; 320: Beam splitter unit; 330: Monitoring unit; 340: Prediction unit;

[0043] 410: Processor; 420: Communication interface; 430: Memory; 440: 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 Figures 1-2 The present invention describes a deep learning-based method for predicting and monitoring the fabrication of nanoscale structures, such as... Figure 1 , Figure 1 This is one of the flowcharts of the deep learning-based nanoscale structure fabrication prediction and monitoring method provided by the present invention, including the following steps:

[0046] Step 110: Obtain laser processing parameters and input the laser processing parameters into the laser module to obtain a laser beam.

[0047] This retrieves the laser processing parameters that the user wishes to use for fabricating nanoscale structures. These parameters can be obtained through direct user input. The laser processing parameters include laser flux, scanning speed, number of scans, and scanning spacing.

[0048] The acquired laser processing parameters are input to the laser module, and a laser beam is modulated according to the laser processing parameters. The laser module is used to modulate a femtosecond laser beam according to the laser processing parameters. In some embodiments, the laser module includes a laser. Further, the laser module may be a femtosecond laser.

[0049] Step 120: Split the laser beam into a monitoring laser beam and a processing laser beam.

[0050] After obtaining the laser beam, the laser beam is split into a processing laser beam for laser processing and a monitoring laser beam for monitoring using a beam splitting method.

[0051] Furthermore, in some embodiments, the laser beam is directly split into three beams: a processing laser beam, a spot morphology monitoring laser beam, and a laser energy density monitoring laser beam. The processing laser beam is used for laser processing, the spot morphology monitoring laser beam is used for spot quality monitoring, and the laser energy density monitoring laser beam is used for power density monitoring. In actual operation, a beam expander, a first beam splitter, and a second beam splitter are used to split the laser beam.

[0052] Furthermore, before splitting the laser beam, the process also includes amplifying the laser beam using an amplifier.

[0053] Step 130: Use a pre-built online monitoring module to monitor the monitoring laser beam in real time to obtain monitoring results, including spot morphology data and laser energy density data.

[0054] Real-time monitoring of the system's state is crucial during the fabrication of nanoscale structures. After obtaining the monitoring laser beam, a pre-built online monitoring module is used to perform real-time steady-state monitoring of the system, yielding monitoring results. These results can be used not only for real-time prediction of the fabricated nanoscale structures but also for monitoring the state of the fabrication system, including spot quality and laser power.

[0055] The online monitoring module includes at least a spot analysis device and a power metering device. The spot analysis device is used to analyze the spot of the monitored laser beam to obtain spot morphology data, and the power metering device is used to test the laser energy density of the monitored laser beam to obtain laser energy density data.

[0056] Step 140: Input the monitoring results into the pre-trained nanoscale structure morphology type and quality prediction model to obtain the nanoscale structure morphology type and quality prediction results;

[0057] The nanoscale structure morphology type and quality prediction model is based on a deep neural network trained using a training sample database, which includes at least a large number of processing data samples.

[0058] After obtaining the monitoring results, the real-time monitoring results are input into the nanoscale structure morphology type and quality prediction model to obtain the nanoscale structure morphology type and quality prediction results.

[0059] The nanoscale structure morphology type and quality prediction model is obtained by training a deep neural network using a training sample database, which includes at least a massive amount of processing data samples. The trained nanoscale structure morphology type and quality prediction model is a multimodal offline prediction model that reflects the relationship between laser processing conditions, nanoscale structure morphology type, and processing quality. This invention, by introducing a neural network algorithm model, can predict the nanoscale structure morphology type and processing quality in real time under different combinations of laser processing parameters.

[0060] The prediction results for the morphology type and quality of nanoscale structures include predictions for the morphology type and the fabrication quality. It's important to understand that since the morphology types of nanoscale structures include three types—periodic stripe structures, peak structures, and groove structures—the prediction results for the morphology type include four outcomes: periodic stripe structures, peak structures, groove structures, and poor nanoscale structures. The fabrication quality of nanoscale structures includes both good and bad cases, and the prediction results for the fabrication quality of nanoscale structures also include both good and bad cases. In some embodiments, the prediction results for the morphology type of nanoscale structures also include the size of the nanoscale structures; it should be noted that the size range of the nanoscale structures is within 100 nm–1200 nm.

[0061] Furthermore, after obtaining the predicted morphology and quality of the nanoscale structure, the process further includes: comparing the predicted morphology and quality with a specific nanoscale structure to determine whether the obtained laser processing parameters meet the requirements. If they are the same, the processing laser beam obtained in step 120 is used to process the nanoscale structure. Specifically, the processing laser beam is input into a galvanometer system for rapid, large-area fabrication of nanoscale structures. Simultaneously, the system status, including spot quality and laser power, can be monitored in real time through an online monitoring module. If they differ, the laser processing parameters are re-inputted, and monitoring and prediction are repeated. Based on this, the present invention combines deep learning algorithms with a laser processing system and an online monitoring system to achieve intelligent processing, large-area fabrication, and real-time online quality monitoring of different nanoscale structures.

[0062] Through the above steps, this invention can determine the optimized nanoscale structure processing window, enabling users to quickly find the best laser processing parameters for the desired nanoscale morphology. It also allows for real-time monitoring of the processing system's stability and provides feedback and upgrades. This invention enables intelligent and low-cost rapid, large-area fabrication of various types of femtosecond laser nanoscale structures.

[0063] Based on the above embodiments, in this method, a pre-built online monitoring module is used to monitor the monitoring laser beam in real time to obtain monitoring results, specifically including:

[0064] The monitoring laser beam is split into a spot morphology monitoring laser beam and a laser energy density monitoring laser beam;

[0065] The online monitoring module is used to monitor the laser beam for spot morphology monitoring in real time to obtain spot morphology data. The online monitoring module is also used to monitor the laser beam for laser energy density monitoring in real time to obtain laser energy density data.

[0066] The spot morphology data and laser energy density data constitute the monitoring results.

[0067] Specifically, after splitting the laser beam into a monitoring laser beam and a processing laser beam, in some embodiments, the online monitoring module includes a spot analyzer and a power meter for monitoring. That is, the spot morphology and laser energy density during processing are monitored in real time using the spot analyzer and power meter. At this point, it is necessary to split the monitoring laser beam. Specifically, the monitoring laser beam is extracted, and then split into a spot morphology monitoring laser beam and a laser energy density monitoring laser beam using a beam expander and a beam splitter. The spot morphology monitoring laser beam is input into the spot analyzer in the online monitoring module to obtain spot morphology data, and the laser energy density monitoring laser beam is input into the power meter in the online monitoring module to obtain laser energy density data.

[0068] Based on the above embodiments, in this method, the monitoring results are input into a pre-trained nanoscale structure morphology type and quality prediction model to obtain nanoscale structure morphology type and quality prediction results, and then the method further includes:

[0069] The monitoring results and the predicted morphology type and quality of the nanoscale structure were tested, and erroneous prediction data were extracted.

[0070] The erroneous prediction data is added to the training sample database to update the prediction model for the morphology type and quality of the nanoscale structure.

[0071] Specifically, after obtaining the prediction results of the morphology type and quality of the nanoscale structure, the monitoring results and the prediction results of the morphology type and quality of the nanoscale structure can be tested. The morphology and quality of the nanoscale structure are measured by SEM. The actual data in the monitoring results are compared with the predicted data in the prediction results of the morphology type and quality of the nanoscale structure. The prediction results of the morphology type and quality of the nanoscale structure that do not match the monitoring results are extracted as erroneous prediction data. The erroneous prediction data are put into the training sample database for updating the prediction model of the morphology type and quality of the nanoscale structure.

[0072] In actual operation, laser processing parameters are continuously acquired, monitored, and predicted. Incorrect prediction data are put into the training sample database, the offline model is retrained, and the new model is used in online monitoring to continuously improve the model and increase its accuracy.

[0073] Based on the above embodiments, in this method, a prediction model for the morphology type and quality of nanoscale structures is obtained by training a deep neural network using a training sample database, specifically including:

[0074] Acquire a large amount of processing data samples of nanoscale structure samples processed under different laser processing conditions. The processing data samples include at least the surface nanomorphology type samples, nanostructure quality samples, laser spot morphology samples and laser processing parameter samples of the nanoscale structure samples.

[0075] The training sample database is constructed using a massive amount of the aforementioned processing data samples;

[0076] The pre-constructed deep neural network is trained using the training sample database to obtain a prediction model for the morphology type and quality of nanoscale structures.

[0077] Specifically, a training sample database is first constructed. This database includes a database of nanoscale structure morphology types and processing quality under different laser processing conditions. It consists of a massive amount of processing data samples of nanoscale structure samples processed under various laser processing conditions. Each processing data sample includes at least the following: surface nanoscale morphology type samples, nanoscale structure quality samples, laser spot morphology samples used to process the nanoscale structure samples, and laser processing parameter samples. The laser processing parameter samples include laser flux samples, scanning speed samples, number of scans samples, and scanning interval samples.

[0078] The pre-built deep neural network is trained using a training sample database. For example, in one specific embodiment, the training sample database contains 5000 sets of processing data samples. Each set includes samples of the surface nanomorphology type of the nanoscale structure sample, samples of the nanostructure quality, samples of the laser spot morphology used to process the nanoscale structure sample, and samples of laser processing parameters. The deep neural network includes a convolutional neural network and a feedforward neural network. The convolutional neural network has 10-20 convolutional layers and 3*3 convolutional kernels. The feedforward neural network has 5-10 layers. The convolutional neural network and the feedforward neural network are connected by fully connected layers, which have 1-3 layers and 20-100 neurons per layer. The loss function used is mean squared error. The input consists of the spot distribution image in the spot morphology sample and the laser processing parameters in the laser processing parameter sample. The output consists of the structural morphology in the surface nanomorphology type sample of the nanoscale structure sample and the processing quality in the nanostructure quality sample. The nanoscale structure morphology type and quality prediction model is trained to reveal the changes in sample processing morphology type and quality under laser processing conditions.

[0079] Based on the above embodiments, in this method, the deep neural network includes a convolutional neural network and a feedforward neural network. The convolutional neural network is used to extract laser spot morphology samples from the processing data samples, and the feedforward neural network is used to extract laser processing parameter samples from the processing data samples. The convolutional neural network and the feedforward neural network are fused together through a fully connected layer.

[0080] Specifically, the pre-built deep neural network includes two sub-networks: a convolutional neural network and a feedforward neural network, which are used to extract light spot morphology samples and laser processing parameter samples from the processing data samples, respectively. The two sub-networks are fused together through a fully connected layer.

[0081] Based on the above embodiments, in this method, the number of convolutional layers in the convolutional neural network is 10-20, and the number of pooling layers is 5-10.

[0082] Specifically, in some embodiments, the deep neural network includes a convolutional neural network with 10-20 convolutional layers and 3*3 convolutional kernels, and a feedforward neural network with 5-10 layers. The convolutional neural network and the feedforward neural network are connected by fully connected layers, with 1-3 fully connected layers and 20-100 neurons in each layer.

[0083] Based on the above embodiments, the present invention also provides a specific embodiment of a single processing using the above-described deep learning-based nanoscale structure processing prediction and monitoring method, comprising the following steps:

[0084] S1: Obtain laser processing parameters, input the laser processing parameters into a femtosecond laser to obtain a laser beam;

[0085] S2: Input the laser beam into the beam splitter to obtain the monitoring laser beam and the processing laser beam;

[0086] S3: Input the monitoring laser beam into the beam splitter to obtain the laser beam for monitoring the spot morphology and the laser energy density;

[0087] S4: Input the laser beam for spot morphology monitoring into the spot analyzer to obtain spot morphology data, and input the laser beam for laser energy density monitoring into the power meter to obtain laser energy density data. The spot analyzer and the power meter together constitute the online monitoring module. The output of the monitoring module is the monitoring result, which includes spot morphology data and laser energy density data.

[0088] S5: Input the monitoring results into the nanoscale structure morphology type and quality prediction model (intelligent processing module) to obtain the nanoscale structure morphology type and quality prediction results;

[0089] S6: Input the monitoring results and the nanoscale structure morphology type and quality prediction results into the SEM test module, compare the prediction results with the monitoring results. If they match, proceed to step S7; otherwise, proceed to step S8.

[0090] S7: Input the laser beam obtained in step S2 into the galvanometer system, process it using the processing platform, and input the processing result into the SEM test module;

[0091] S8: Input the erroneous prediction data into the training sample database of the nanoscale structure morphology type and quality prediction model for feedback and upgrading. Then, input the upgraded nanoscale structure morphology type and quality prediction model into the intelligent processing module for online monitoring.

[0092] like Figure 2 As shown, Figure 2 This is a schematic diagram of the device corresponding to this embodiment.

[0093] This invention provides a deep learning-based method for predicting and monitoring the processing of nanoscale structures. The method involves acquiring laser processing parameters and inputting these parameters into a laser module to obtain a laser beam. The laser beam is then split into a monitoring laser beam and a processing laser beam. A pre-built online monitoring module monitors the monitoring laser beam in real time to obtain monitoring results, including spot morphology data and laser energy density data. These monitoring results are then input into a pre-trained nanoscale structure morphology type and quality prediction model to obtain prediction results for the nanoscale structure morphology type and quality. The nanoscale structure morphology type and quality prediction model is trained using a deep neural network with a training sample database, which includes at least a large number of processing data samples. This invention utilizes a deep learning algorithm to predict the morphology and quality of the processed structure based on the input laser processing parameters. Simultaneously, it monitors the system state, including spot quality and laser power, in real time. This invention combines offline prediction and online monitoring to achieve low-cost and high-efficiency selection of laser processing parameters.

[0094] The following describes the deep learning-based nanoscale structure fabrication prediction and monitoring device provided by this invention. The deep learning-based nanoscale structure fabrication prediction and monitoring device described below can be referred to in correspondence with the deep learning-based nanoscale structure fabrication prediction and monitoring method described above. Figure 3 , Figure 3 This is a schematic diagram of the structure of the deep learning-based nanoscale structure fabrication prediction and monitoring device provided by the present invention, including:

[0095] The laser unit 310 is used to acquire laser processing parameters and input the laser processing parameters into the laser module to obtain a laser beam;

[0096] The beam splitting unit 320 is used to split the laser beam into a monitoring laser beam and a processing laser beam.

[0097] The monitoring unit 330 is used to monitor the monitoring laser beam in real time using a pre-built online monitoring module to obtain monitoring results, including spot morphology data and laser energy density data.

[0098] The prediction unit 340 is used to input the monitoring results into a pre-trained nanoscale structure morphology type and quality prediction model to obtain nanoscale structure morphology type and quality prediction results.

[0099] The nanoscale structure morphology type and quality prediction model is based on a deep neural network trained using a training sample database, which includes at least a large number of processing data samples.

[0100] Based on the above embodiments, in this device, a pre-built online monitoring module is used to monitor the monitoring laser beam in real time to obtain monitoring results, specifically including:

[0101] The monitoring laser beam is split into a spot morphology monitoring laser beam and a laser energy density monitoring laser beam;

[0102] The online monitoring module is used to monitor the laser beam for spot morphology monitoring in real time to obtain spot morphology data. The online monitoring module is also used to monitor the laser beam for laser energy density monitoring in real time to obtain laser energy density data.

[0103] The spot morphology data and laser energy density data constitute the monitoring results.

[0104] Based on the above embodiments, in this device, the monitoring results are input into a pre-trained nanoscale structure morphology type and quality prediction model to obtain nanoscale structure morphology type and quality prediction results, and then the device further includes:

[0105] The monitoring results and the predicted morphology type and quality of the nanoscale structure were tested, and erroneous prediction data were extracted.

[0106] The erroneous prediction data is added to the training sample database to update the prediction model for the morphology type and quality of the nanoscale structure.

[0107] Based on the above embodiments, in this device, a nanoscale structure morphology type and quality prediction model is obtained by training a deep neural network using a training sample database, specifically including:

[0108] Acquire a large amount of processing data samples of nanoscale structure samples processed under different laser processing conditions. The processing data samples include at least the surface nanomorphology type samples, nanostructure quality samples, laser spot morphology samples and laser processing parameter samples of the nanoscale structure samples.

[0109] The training sample database is constructed using a massive amount of the aforementioned processing data samples;

[0110] The pre-constructed deep neural network is trained using the training sample database to obtain a prediction model for the morphology type and quality of nanoscale structures.

[0111] Based on the above embodiments, in this device, the deep neural network includes a convolutional neural network and a feedforward neural network. The convolutional neural network is used to extract laser spot morphology samples from the processing data samples, and the feedforward neural network is used to extract laser processing parameter samples from the processing data samples. The convolutional neural network and the feedforward neural network are fused together through a fully connected layer.

[0112] Based on the above embodiments, in this device, the number of convolutional layers in the convolutional neural network is 10-20, and the number of pooling layers is 5-10.

[0113] This invention provides a deep learning-based device for predicting and monitoring the processing of nanoscale structures. It acquires laser processing parameters and inputs these parameters into a laser module to obtain a laser beam. The laser beam is then split into a monitoring laser beam and a processing laser beam. A pre-built online monitoring module monitors the monitoring laser beam in real time to obtain monitoring results, including spot morphology data and laser energy density data. These monitoring results are then input into a pre-trained nanoscale structure morphology type and quality prediction model to obtain prediction results for the nanoscale structure morphology type and quality. The nanoscale structure morphology type and quality prediction model is trained using a deep neural network with a training sample database, which includes at least a large number of processing data samples. This invention utilizes a deep learning algorithm to predict the morphology and quality of the processed structure based on the input laser processing parameters. Simultaneously, it monitors the system state, including spot quality and laser power, in real time. This invention combines offline prediction and online monitoring to achieve low-cost and high-efficiency selection of laser processing parameters.

[0114] Figure 4 An example is a schematic diagram of the physical structure of an electronic device, such as... Figure 4As shown, the electronic device may include a processor 410, a communication interface 420, a memory 430, and a communication bus 440, wherein the processor 410, the communication interface 420, and the memory 430 communicate with each other via the communication bus 440. The processor 410 can call logical instructions in the memory 430 to execute a deep learning-based method for predicting and monitoring the processing of nanoscale structures. This method includes: acquiring laser processing parameters; inputting the laser processing parameters into a laser module to obtain a laser beam; splitting the laser beam into a monitoring laser beam and a processing laser beam; using a pre-built online monitoring module to monitor the monitoring laser beam in real time to obtain monitoring results, the monitoring results including spot morphology data and laser energy density data; inputting the monitoring results into a pre-trained nanoscale structure morphology type and quality prediction model to obtain nanoscale structure morphology type and quality prediction results; wherein the nanoscale structure morphology type and quality prediction model is obtained by training a deep neural network using a training sample database, the training sample database including at least a large number of processing data samples.

[0115] Furthermore, the logical instructions in the aforementioned memory 430 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.

[0116] 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 deep learning-based nanoscale structure processing prediction and monitoring method provided by the above methods. The method includes: acquiring laser processing parameters; inputting the laser processing parameters into a laser module to obtain a laser beam; splitting the laser beam into a monitoring laser beam and a processing laser beam; using a pre-built online monitoring module to monitor the monitoring laser beam in real time to obtain monitoring results, the monitoring results including spot morphology data and laser energy density data; inputting the monitoring results into a pre-trained nanoscale structure morphology type and quality prediction model to obtain nanoscale structure morphology type and quality prediction results; wherein the nanoscale structure morphology type and quality prediction model is obtained by training a deep neural network using a training sample database, the training sample database including at least a large number of processing data samples.

[0117] 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 deep learning-based method for predicting and monitoring the fabrication of nanoscale structures provided by the methods described above. This method includes: acquiring laser fabrication parameters; inputting the laser fabrication parameters into a laser module to obtain a laser beam; splitting the laser beam into a monitoring laser beam and a processing laser beam; using a pre-built online monitoring module to monitor the monitoring laser beam in real time to obtain monitoring results, the monitoring results including spot morphology data and laser energy density data; inputting the monitoring results into a pre-trained nanoscale structure morphology type and quality prediction model to obtain nanoscale structure morphology type and quality prediction results; wherein the nanoscale structure morphology type and quality prediction model is obtained by training a deep neural network using a training sample database, the training sample database including at least a large number of fabrication data samples.

[0118] 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.

[0119] 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.

[0120] 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 predicting and monitoring the fabrication of nanoscale structures based on deep learning, characterized in that, include: The laser processing parameters are obtained and input into the laser module to obtain a laser beam; The laser beam is split into a monitoring laser beam and a processing laser beam; The monitoring laser beam is monitored in real time using a pre-built online monitoring module to obtain monitoring results, including spot morphology data and laser energy density data. The monitoring results are input into a pre-trained nanoscale structure morphology type and quality prediction model to obtain nanoscale structure morphology type and quality prediction results. The nanoscale structure morphology type and quality prediction model is based on a deep neural network trained using a training sample database, which includes at least a large number of processing data samples.

2. The method for predicting and monitoring the fabrication of nanoscale structures based on deep learning according to claim 1, characterized in that, The monitoring laser beam is monitored in real time using a pre-built online monitoring module to obtain monitoring results, specifically including: The monitoring laser beam is split into a spot morphology monitoring laser beam and a laser energy density monitoring laser beam; The online monitoring module is used to monitor the laser beam for spot morphology monitoring in real time to obtain spot morphology data. The online monitoring module is also used to monitor the laser beam for laser energy density monitoring in real time to obtain laser energy density data. The spot morphology data and laser energy density data constitute the monitoring results.

3. The method for predicting and monitoring the fabrication of nanoscale structures based on deep learning according to claim 1, characterized in that, The monitoring results are input into a pre-trained nanoscale structure morphology type and quality prediction model to obtain nanoscale structure morphology type and quality prediction results, and then the process further includes: The monitoring results and the predicted morphology type and quality of the nanoscale structure were tested, and erroneous prediction data were extracted. The erroneous prediction data is added to the training sample database to update the prediction model for the morphology type and quality of the nanoscale structure.

4. The method for predicting and monitoring the fabrication of nanoscale structures based on deep learning according to claim 1, characterized in that, A model for predicting the morphology type and quality of nanoscale structures is obtained by training a deep neural network using a training sample database. Specifically, it includes: Acquire a large amount of processing data samples of nanoscale structure samples processed under different laser processing conditions. The processing data samples include at least the surface nanomorphology type samples, nanostructure quality samples, laser spot morphology samples and laser processing parameter samples of the nanoscale structure samples. The training sample database is constructed using a massive amount of the aforementioned processing data samples; The pre-constructed deep neural network is trained using the training sample database to obtain a prediction model for the morphology type and quality of nanoscale structures.

5. The method for predicting and monitoring the fabrication of nanoscale structures based on deep learning according to claim 1 or 4, characterized in that, The deep neural network includes a convolutional neural network and a feedforward neural network. The convolutional neural network is used to extract laser spot morphology samples from the processing data samples, and the feedforward neural network is used to extract laser processing parameter samples from the processing data samples. The convolutional neural network and the feedforward neural network are fused together through a fully connected layer.

6. The method for predicting and monitoring the fabrication of nanoscale structures based on deep learning according to claim 5, characterized in that, The convolutional neural network has 10-20 convolutional layers and 5-10 pooling layers.

7. A device for predicting and monitoring the fabrication of nanoscale structures based on deep learning, characterized in that, include: The laser unit is used to acquire laser processing parameters and input the laser processing parameters into the laser module to obtain a laser beam; The beam splitting unit is used to split the laser beam into a monitoring laser beam and a processing laser beam. The monitoring unit is used to monitor the monitoring laser beam in real time using a pre-built online monitoring module to obtain monitoring results, including spot morphology data and laser energy density data. The prediction unit is used to input the monitoring results into a pre-trained nanoscale structure morphology type and quality prediction model to obtain nanoscale structure morphology type and quality prediction results. The nanoscale structure morphology type and quality prediction model is based on a deep neural network trained using a training sample database, which includes at least a large number of processing data samples.

8. 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 deep learning-based nanoscale structure fabrication prediction and monitoring method as described in any one of claims 1 to 6.

9. 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 deep learning-based nanoscale structure fabrication prediction and monitoring method as described in any one of claims 1 to 6.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements the deep learning-based nanoscale structure fabrication prediction and monitoring method as described in any one of claims 1 to 6.

Images