CN116595339A - Intelligent processing method, device and equipment for space data - Google Patents
Intelligent processing method, device and equipment for space data Download PDFInfo
- Publication number
- CN116595339A CN116595339A CN202310882558.6A CN202310882558A CN116595339A CN 116595339 A CN116595339 A CN 116595339A CN 202310882558 A CN202310882558 A CN 202310882558A CN 116595339 A CN116595339 A CN 116595339A
- Authority
- CN
- China
- Prior art keywords
- data
- processing model
- processed
- processing
- encoder
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000003672 processing method Methods 0.000 title claims abstract description 15
- 238000012545 processing Methods 0.000 claims abstract description 245
- 238000000034 method Methods 0.000 claims abstract description 50
- 238000007781 pre-processing Methods 0.000 claims abstract description 43
- 238000012549 training Methods 0.000 claims abstract description 33
- 238000007405 data analysis Methods 0.000 claims abstract description 8
- 230000006870 function Effects 0.000 claims description 64
- 238000010606 normalization Methods 0.000 claims description 23
- 230000007246 mechanism Effects 0.000 claims description 22
- 238000012360 testing method Methods 0.000 claims description 14
- 238000005457 optimization Methods 0.000 claims description 11
- 238000003860 storage Methods 0.000 claims description 11
- 230000009466 transformation Effects 0.000 claims description 6
- 238000004590 computer program Methods 0.000 claims description 5
- 230000000873 masking effect Effects 0.000 claims description 5
- 230000003042 antagnostic effect Effects 0.000 claims description 3
- 238000013528 artificial neural network Methods 0.000 claims description 3
- 230000008569 process Effects 0.000 abstract description 20
- 238000013461 design Methods 0.000 abstract description 10
- 238000011161 development Methods 0.000 abstract description 7
- 230000004044 response Effects 0.000 abstract description 5
- 238000013473 artificial intelligence Methods 0.000 abstract description 2
- 238000011156 evaluation Methods 0.000 description 16
- 238000010276 construction Methods 0.000 description 13
- 230000018109 developmental process Effects 0.000 description 5
- 238000004422 calculation algorithm Methods 0.000 description 3
- 230000008878 coupling Effects 0.000 description 3
- 238000010168 coupling process Methods 0.000 description 3
- 238000005859 coupling reaction Methods 0.000 description 3
- 230000000694 effects Effects 0.000 description 3
- 230000006978 adaptation Effects 0.000 description 2
- 238000004458 analytical method Methods 0.000 description 2
- 238000004364 calculation method Methods 0.000 description 2
- 238000004891 communication Methods 0.000 description 2
- 238000000354 decomposition reaction Methods 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 230000000977 initiatory effect Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000012544 monitoring process Methods 0.000 description 2
- 230000006798 recombination Effects 0.000 description 2
- 238000005215 recombination Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000002372 labelling Methods 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 238000003058 natural language processing Methods 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 230000002787 reinforcement Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000012216 screening Methods 0.000 description 1
- 230000011218 segmentation Effects 0.000 description 1
- 238000004088 simulation Methods 0.000 description 1
- 238000012795 verification Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/33—Querying
- G06F16/332—Query formulation
- G06F16/3329—Natural language query formulation or dialogue systems
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/0464—Convolutional networks [CNN, ConvNet]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/0499—Feedforward networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
- Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
- Y02P90/30—Computing systems specially adapted for manufacturing
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Artificial Intelligence (AREA)
- Mathematical Physics (AREA)
- General Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Computational Linguistics (AREA)
- Evolutionary Computation (AREA)
- Molecular Biology (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computing Systems (AREA)
- General Health & Medical Sciences (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Evolutionary Biology (AREA)
- Human Computer Interaction (AREA)
- Databases & Information Systems (AREA)
- Machine Translation (AREA)
Abstract
The invention provides an intelligent processing method, device and equipment for space data, and relates to the field of artificial intelligence. The intelligent processing method of the aerospace data is applied to a spacecraft and comprises the following steps: acquiring data to be processed of a spacecraft; performing data preprocessing on the data to be processed to obtain text data of the data to be processed; inputting the text data into an intelligent processing model for data analysis processing to obtain a processing result of the data to be processed, and outputting the processing result; the intelligent processing model is obtained by training a first preset processing model according to a historical processing data set of spacecraft data to be processed, and the first preset processing model is obtained by constructing the first preset processing model according to the data to be processed. The scheme of the invention realizes the rapid response of the spacecraft problem in the development process of the spacecraft, and reduces the design difficulty and development cost of the spacecraft.
Description
Technical Field
The invention relates to the field of artificial intelligence, in particular to an intelligent processing method, device and equipment for space data.
Background
In the existing spacecraft construction process, a technician is required to design each part of the spacecraft according to the use requirement of the spacecraft and the design experience of the technician, and the designed spacecraft is required to be subjected to procedures such as verification, manufacture, assembly and test according to experience, the process requires a great deal of time and effort of scientific researchers, the development cost is high, and after the spacecraft is successfully developed, the technician is required to analyze the faults of the spacecraft according to the parameters and experience of the spacecraft in the operation working process, and an optimal solution is provided when the spacecraft is successfully developed, and the process not only requires a great deal of technicians to stand by in real time, but also requires a great deal of time to research a reasonable solution, so that the test initiation end cannot quickly respond when the spacecraft encounters a problem, and the reliability of the spacecraft in the operation process is reduced.
Disclosure of Invention
The invention provides an intelligent processing method, device and equipment for space data. The method solves the problems that the development time of the spacecraft is long, the cost is high, and an optimal response scheme cannot be rapidly given when the spacecraft is in a running process.
In order to solve the technical problems, the technical scheme of the invention is as follows:
the embodiment of the invention provides an intelligent processing method of space data, which is applied to a spacecraft, and comprises the following steps:
acquiring data to be processed of a spacecraft;
performing data preprocessing on the data to be processed to obtain text data of the data to be processed;
inputting the text data into an intelligent processing model for data analysis processing to obtain a processing result of the data to be processed, and outputting the processing result;
the intelligent processing model is obtained by training a first preset processing model according to a historical processing data set of spacecraft data to be processed, and the first preset processing model is obtained by constructing the first preset processing model according to the data to be processed.
Optionally, constructing a first preset processing model according to the data to be processed includes:
constructing a data preprocessing function of a first preset processing model according to the data to be processed;
constructing an encoder of a first preset processing model based on the data preprocessing function;
a decoder for constructing a first preset processing model based on the encoder;
constructing a linear transformation function and a connection function of a first preset processing model based on the decoder;
constructing a loss function of a first preset processing model based on the connection function;
and constructing a back propagation function and a parameter updating function of a first preset processing model based on the loss function, and acquiring the first preset processing model.
Optionally, constructing a data preprocessing function of a first preset processing model according to the data to be processed, including:
acquiring a data sequence x to be processed according to the data to be processed;
and carrying out data preprocessing on the data sequence x to be processed to obtain a data preprocessing function of a first preset processing model.
Optionally, based on the data preprocessing function, an encoder for constructing a first preset processing model includes:
determining the number of encoder layers N_encoder;
constructing an initial encoder, and performing N_encoder iterations on the initial encoder based on the data preprocessing function to obtain an encoder of a first preset processing model;
wherein each iteration performs multi-head attention mechanism, layer normalization and position feed-forward network operation on the current encoder output result.
Optionally, based on the encoder, a decoder for constructing a first preset processing model includes:
determining the number of decoder layers N_decoder;
constructing an initial decoder according to the encoder of the first preset processing model;
performing N_decoder iterations on the initial decoder based on the output result of the encoder of the first preset processing model to obtain a decoder of the first preset processing model;
wherein each iteration performs masking multi-head attention mechanisms, layer normalization, encoding-decoding multi-head attention mechanisms, layer normalization, and position feed-forward network operations on the current decoder output result. Optionally, training the first preset processing model according to a historical processing data set of the spacecraft data to be processed to obtain the intelligent processing model, including:
acquiring a historical processing data set of the spacecraft data to be processed;
constructing an autonomous decision engine, integrating the autonomous decision engine with the first preset processing model, and acquiring a second preset processing model;
and training the second preset processing model through the historical processing data set to obtain an intelligent processing model.
Optionally, training the second preset processing model through the historical processing data set to obtain an intelligent processing model, including:
dividing the historical processing data set into a historical processing data training set and a historical processing data testing set;
training the second preset processing model through an antagonistic neural network based on the historical processing data training set to obtain an optimized processing model;
and repeatedly evaluating and iterating the optimization processing model based on the historical processing data test set to obtain an intelligent processing model.
The embodiment of the invention also provides an intelligent processing device for the space data, which comprises the following steps:
the acquisition module is used for acquiring data to be processed of the spacecraft, carrying out data preprocessing on the data to be processed, and acquiring text data of the data to be processed;
the processing module is used for inputting the text data into an intelligent processing model for data analysis and processing, obtaining a processing result of the data to be processed and outputting the processing result; the intelligent processing model is obtained by training a first preset processing model according to a historical processing data set of spacecraft data to be processed, and the first preset processing model is obtained by constructing the first preset processing model according to the data to be processed.
Embodiments of the present invention also provide a computing device comprising: the intelligent processing method for the aerospace data comprises a processor and a memory storing a computer program, wherein the computer program is executed by the processor.
Embodiments of the present invention also provide a computer-readable storage medium including: and storing instructions, and enabling the computer to execute the intelligent processing method of the aerospace data when the instructions run on the computer.
The scheme of the invention at least comprises the following beneficial effects:
according to the scheme, the data to be processed of the spacecraft is obtained, and the data to be processed is subjected to data preprocessing, so that text data of the data to be processed are obtained; the text data is input into the intelligent processing model for data analysis processing, a processing result of the data to be processed is obtained and output, the quick response of the problems of the spacecraft in the development process of the spacecraft is realized, the design difficulty and development cost of the spacecraft are reduced, meanwhile, through the design of an autonomous decision engine, the real-time monitoring and control of the full life cycle in the flight process of the spacecraft and the quick response of the problems in the operation of the spacecraft are realized, the efficiency of a testing initiation control end is improved, and the reliability of a rocket is ensured.
Drawings
FIG. 1 is a flow diagram of an intelligent processing method of aerospace data of the present invention;
fig. 2 is a schematic block diagram of an intelligent processing device for space data according to the present invention.
Detailed Description
Exemplary embodiments of the present invention will be described in more detail below with reference to the accompanying drawings. While exemplary embodiments of the present invention are shown in the drawings, it should be understood that the present invention may be embodied in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the invention to those skilled in the art.
As shown in fig. 1, an embodiment of the present invention provides an intelligent processing method of aerospace data, applied to a spacecraft, where the method includes:
step 11, obtaining data to be processed of a spacecraft;
step 12, carrying out data preprocessing on the data to be processed to obtain text data of the data to be processed;
step 13, inputting the text data into an intelligent processing model for data analysis processing, obtaining a processing result of the data to be processed, and outputting the processing result;
the intelligent processing model is obtained by training a first preset processing model according to a historical processing data set of spacecraft data to be processed, and the first preset processing model is obtained by constructing the first preset processing model according to the data to be processed.
In this embodiment, the data to be processed includes: fault problem data in the running process of the spacecraft and construction problem data in the construction process of various types of spacecraft; the data preprocessing is carried out on the data to be processed, and the method comprises the steps of word segmentation, labeling, input sample construction, text data acquisition and the like of the data to be processed; in addition, the data can be filtered and cleaned according to the data screening requirements in different scenes, so that the high quality and consistency of the data are ensured; in the embodiment, through the design of the intelligent processing model, the quick response to the construction problem in the spacecraft construction process is realized, and the design difficulty and development cost of the spacecraft are reduced; meanwhile, the method can also quickly respond to the fault problem encountered in the running process of the spacecraft, so that the efficiency of testing the launch control terminal is improved, and the reliability of the rocket is ensured.
In an optional embodiment of the invention, constructing a first preset processing model according to the data to be processed includes:
step 14, constructing a data preprocessing function of a first preset processing model according to the data to be processed;
step 15, constructing an encoder of a first preset processing model based on the data preprocessing function;
step 16, constructing a decoder of a first preset processing model based on the encoder;
step 17, constructing a linear transformation function and a connection function of a first preset processing model based on the decoder;
step 18, constructing a loss function of a first preset processing model based on the connection function;
and step 19, constructing a back propagation function and a parameter updating function of a first preset processing model based on the loss function, and acquiring the first preset processing model.
In this embodiment, when a first preset processing model is constructed, firstly, an infrastructure of the first preset processing model needs to be obtained according to the type of the data to be processed; in this embodiment, a GPT-3.5 model based on a transducer architecture is selected as the infrastructure.
In an alternative embodiment of the present invention, step 14 may include:
acquiring a data sequence x to be processed according to the data to be processed;
the data preprocessing method comprises the steps of carrying out data preprocessing on a data sequence x to be processed to obtain a data preprocessing function of a first preset processing model, wherein the specific process is as follows: x_preprocessed=preprocess (x);
in an alternative embodiment of the present invention, step 15 may include:
determining the number of encoder layers N_encoder;
constructing an initial encoder, and performing N_encoder iterations on the initial encoder based on the data preprocessing function to obtain an encoder of a first preset processing model;
wherein each iteration performs multi-head attention mechanism, layer normalization and position feed-forward network operation on the current encoder output result.
In this embodiment, in order to determine the number of layers n_encoder of the encoder, the data sequence x to be processed needs to be preprocessed according to the data preprocessing function first, so as to obtain a data preprocessing function x_preprocessed of the first preset processing model. And then, carrying out N_encoder iterations in the loop, and carrying out operations such as a multi-head attention mechanism, layer normalization, a position feedforward network and the like on the output of the encoder in each iteration to finally obtain the encoder of the first preset processing model.
The method comprises the following specific steps:
preprocessing a data sequence x to be processed to obtain a preprocessed sequence x_preprocessed;
constructing an initial encoder, enabling the output of the initial encoder to be x_preprocessed, and carrying out N_encoder iterations on the initial encoder based on the data preprocessing function to obtain an encoder of a first preset processing model;
in each cycle, the current encoder output is:
applying a multi-head attention mechanism, and acquiring output data of the current multi-head attention mechanism, namely first output data; wherein the number of multi-head attention heads in the encoder is set by the case according to the model architecture;
performing layer normalization operation on the first output data to obtain second output data, wherein the second output data is output data of the current layer normalization operation;
inputting the second output data into a position feedforward network of an initial encoder for processing, and obtaining third output data, wherein the third output data is output data of the position feedforward network of the current initial encoder;
updating the output of the initial encoder into third output data, and continuing to circulate;
and after the circulation is finished, obtaining the encoder of the first preset processing model.
In this embodiment, through such a calculation step, an encoder having an n_encoder layer can be obtained, which is capable of performing operations such as multi-head attention, layer normalization, and position feed forward network according to an input data sequence to capture context information and features of the input sequence.
In an alternative embodiment of the present invention, step 16 may include:
determining the number of decoder layers N_decoder;
constructing an initial decoder according to the encoder of the first preset processing model;
performing N_decoder iterations on the initial decoder based on the output result of the encoder of the first preset processing model to obtain a decoder of the first preset processing model;
wherein each iteration performs masking multi-head attention mechanisms, layer normalization, encoding-decoding multi-head attention mechanisms, layer normalization, and position feed-forward network operations on the current decoder output result.
In this embodiment, the input of the initial decoder is the output of the encoder of the first preset processing model, i.e., decoder_output=decoder_output; in the n_decoder number of loop iterations for the initial decoder, each loop, the following is performed on the current decoder output:
applying a mask multi-head attention mechanism to acquire output data of the current mask multi-head attention mechanism, namely fourth output data; wherein the number of masking multi-headed attention heads in the initial decoder can be set according to the situation of the model architecture;
performing layer normalization operation on the fourth output data to obtain fifth output data, wherein the fifth output data is output data of the current layer normalization operation;
applying a encode-decode multi-headed attention mechanism, wherein the encoder output is the query vector and the decoder output is the key value pair;
performing layer normalization operation on the fifth output data to obtain sixth output data, wherein the sixth output data is output data of the current layer normalization operation;
inputting the sixth output data into a position feedforward network of the initial decoder for processing to obtain seventh output data, wherein the seventh output data is output data of the position feedforward network of the current initial decoder;
updating the output of the initial decoder into seventh output data, and continuing to circulate;
and after the circulation is finished, acquiring a decoder of the first preset processing model.
In this embodiment, the specific construction process of the first preset processing model is as follows:
the GPT-3.5 model of a transducer architecture is first selected as the basic architecture of a first preset processing model before construction; acquiring a data sequence x to be processed according to the data to be processed;
1. input sequence length (L): assuming that the input sequence is x and the length is L;
2. input embedding dimension (d_emb): embedding each word in the input sequence into a vector space in d_emb dimension, denoted E, e=embedding (x, d_emb);
3. set the number of attention headers (n_headers): dividing the attention mechanism in the transducer model into n_heads, denoted as H;
4. set hidden layer dimension (d_hidden): the dimension of the hidden layer is d_hidden;
5. setting the number of encoder layers (n_encoder_layers): denoted n_encoder;
6. setting the decoder layer number (n_decoder_layers): denoted n_decoder;
7. setting a learning rate (learning_rate): updating the learning rate of the parameters in the training process of the model;
8. batch size (batch_size) is set: represented as B;
9. setting a loss function (loss_function): denoted as Loss;
10. setting a maximum training step number (max_train_steps): represented as M;
11. setting an input data preprocessing mode (data_preprocessing): denoted as Preprocess;
12. setting a regularization method (regularization_method): expressed as regular;
13. preprocessing input data: x_preprocessed=preprocess (x);
14. construction of an encoder:
an encoder_output =x_processed
for i in range(N_encoder):
The encoder_output=multi-head attention (encoder_output) # multi-head attention mechanism
Layer # normalization of encoder_output=layernormalization (encoder_output)
An encoder_output=locationwisetfeed forward (encoder_output) # location feed forward network;
15. construction of the decoder:
decoder_output = encoder_output；
for i in range(N_decoder):
decoder_output=maskedmultiteadattention (decoder_output) # mask multi-head attention mechanism;
decoder_output=layernormalization (decoder_output) # layer normalization;
decoder_output=multi-head attention (decoder_output) # encoding-decoding multi-head attention mechanism;
decoder_output=layernormalization (decoder_output) # layer normalization;
decoder_output=locationwisetfeed forward (decoder_output) # location feed forward network;
16. linear transformation function and join function (Softmax) construction:
output=linear (decoder_output) # performs Linear transformation;
output= Softmax (output) # applies Softmax operation;
17. construction of a loss function:
loss = Loss (output, y) # calculation Loss;
18. construction of back propagation functions and parameter update functions:
backward(loss)；
update_parameters(learning_rate)。
in an optional embodiment of the present invention, training a first preset processing model according to a historical processing data set of data to be processed of a spacecraft to obtain the intelligent processing model, including:
step 21, acquiring a historical processing data set of the spacecraft data to be processed;
step 22, constructing an autonomous decision engine, integrating the autonomous decision engine with the first preset processing model, and obtaining a second preset processing model;
and step 23, training the second preset processing model through the historical processing data set to obtain an intelligent processing model.
In this embodiment, the design of the autonomous decision engine enables the model to make an autonomous decision based on the input problem or task, and the embodiment is based on a strategy gradient algorithm of deep reinforcement learning: proximal Policy Optimization (PPO) algorithms train the autonomous decision engines of the model and improve the performance of the model by optimizing the policy network; the engine can determine whether to rely on answers generated by the model only or to perform additional reasoning, search or call external resources according to factors such as complexity of the questions, field-specific requirements, confidence of the model and the like; for example, in answering specific technical details about an aerospace task, the model may autonomously decide whether further searching for relevant documents or expert opinions is required; meanwhile, the design of the autonomous decision engine enables the intelligent processing method of the aerospace data to be directly carried on the spacecraft, and real-time monitoring and control of the flight process of the spacecraft and collection of full life cycle information can be achieved.
In an alternative embodiment of the present invention, step 23 may include:
step 231, dividing the historical processing data set into a historical processing data training set and a historical processing data testing set;
step 232, training the second preset processing model through a reward function based on the historical processing data training set to obtain an optimized processing model;
and 233, repeatedly evaluating and iterating the optimized processing model based on the historical processing data test set to obtain an intelligent processing model.
In this embodiment, the historical processing dataset includes technical literature, expert knowledge, historical questions and corresponding target answers in the aerospace field; the bonus function may be defined as: scsreward=similarity_score (model_answer, experert_answer) ×weight; where similarity score is a function used to calculate the similarity between the result output by the model and the target answer. weight is an adjustable weight used to balance the size of the reward; when the model is used, the model is assumed to answer a question related to space mission planning, text similarity measurement (BLEU similarity) is used for comparing the model output result with a target answer after the completion of the task answer, and when the model output result is consistent or similar to the target answer, forward rewards are given, so that the training of the model is achieved;
in this embodiment, based on the historical processing data test set, the optimization processing model is repeatedly evaluated and iterated, and the specific process of obtaining the intelligent processing model is as follows:
defining an evaluation index: firstly, defining a model performance index to be evaluated; in the question-and-answer task in the aerospace field, the present embodiment uses standard natural language processing evaluation index (BLEU) and field-specific evaluation index, such as expert knowledge accuracy, technical detail integrity, etc.
Constructing an evaluation data set: the evaluation dataset contains questions related to the aerospace field and corresponding target answers. The data set covers a wide range of aerospace problems to evaluate the performance of the model in different fields and difficulty levels;
model evaluation was performed: the evaluation dataset is used for evaluating the model, the questions in the dataset are input into the model, answers generated by the model are obtained, then the answers are compared with target answers, and the scores of evaluation indexes are calculated, wherein the scores of the indexes can be evaluated manually by field experts or by an automatic evaluation method.
Analysis and evaluation results: analysis of the evaluation results determines the performance advantages and disadvantages of the model in different respects, for example, the model may perform well in terms of accuracy but is problematic in terms of coverage of technical details.
Collecting expert feedback: communicating with experts in the aerospace field to obtain their opinion and advice on model answers, the feedback of the experts can provide valuable information about model performance, helping to find the shortcomings of the model and the direction of improvement.
Iterative optimization: according to the evaluation result and expert feedback, carrying out iterative optimization on the optimized processing model, and obtaining the processing model after iterative optimization;
and (5) repeatedly evaluating and iterating the optimization processing model to obtain the intelligent processing model.
Repeating the evaluation and iteration: and carrying out model evaluation again on the processing model after iterative optimization, testing the processing model after iterative optimization, analyzing the evaluation result, continuously carrying out evaluation and iteration until the model reaches the expected performance level on the aerospace task, and taking the final model as an intelligent processing model.
As shown in fig. 2, the intelligent processing device 30 for space data provided by the present invention includes:
an acquiring module 31, configured to acquire data to be processed of the spacecraft;
the processing module 32 is configured to perform data preprocessing on the data to be processed, and obtain text data of the data to be processed; inputting the text data into an intelligent processing model for data analysis processing to obtain a processing result of the data to be processed, and outputting the processing result; the intelligent processing model is obtained by training a first preset processing model according to a historical processing data set of spacecraft data to be processed, and the first preset processing model is obtained by constructing the first preset processing model according to the data to be processed.
Optionally, constructing a first preset processing model according to the data to be processed includes:
constructing a data preprocessing function of a first preset processing model according to the data to be processed;
constructing an encoder of a first preset processing model based on the data preprocessing function;
a decoder for constructing a first preset processing model based on the encoder;
constructing a linear transformation function and a connection function of a first preset processing model based on the decoder;
constructing a loss function of a first preset processing model based on the connection function;
and constructing a back propagation function and a parameter updating function of a first preset processing model based on the loss function, and acquiring the first preset processing model.
Optionally, constructing a data preprocessing function of a first preset processing model according to the data to be processed, including:
acquiring a data sequence x to be processed according to the data to be processed;
and carrying out data preprocessing on the data sequence x to be processed to obtain a data preprocessing function of a first preset processing model.
Optionally, based on the data preprocessing function, an encoder for constructing a first preset processing model includes:
determining the number of encoder layers N_encoder;
constructing an initial encoder, and performing N_encoder iterations on the initial encoder based on the data preprocessing function to obtain an encoder of a first preset processing model;
wherein each iteration performs multi-head attention mechanism, layer normalization and position feed-forward network operation on the current encoder output result.
Optionally, based on the encoder, a decoder for constructing a first preset processing model includes:
determining the number of decoder layers N_decoder;
constructing an initial decoder according to the encoder of the first preset processing model;
performing N_decoder iterations on the initial decoder based on the output result of the encoder of the first preset processing model to obtain a decoder of the first preset processing model;
wherein each iteration performs masking multi-head attention mechanisms, layer normalization, encoding-decoding multi-head attention mechanisms, layer normalization, and position feed-forward network operations on the current decoder output result.
Optionally, training the first preset processing model according to a historical processing data set of the spacecraft data to be processed to obtain the intelligent processing model, including:
acquiring a historical processing data set of the spacecraft data to be processed;
constructing an autonomous decision engine, integrating the autonomous decision engine with the first preset processing model, and acquiring a second preset processing model;
and training the second preset processing model through the historical processing data set to obtain an intelligent processing model.
Optionally, training the second preset processing model through the historical processing data set to obtain an intelligent processing model, including:
dividing the historical processing data set into a historical processing data training set and a historical processing data testing set;
training the second preset processing model through an antagonistic neural network based on the historical processing data training set to obtain an optimized processing model;
and repeatedly evaluating and iterating the optimization processing model based on the historical processing data test set to obtain an intelligent processing model.
It should be noted that, the device is a device corresponding to the method for determining the rocket simulation model, and all implementation manners in the method are applicable to the embodiment of the device, so that the same technical effects can be achieved.
Embodiments of the present invention also provide a computing device comprising: the intelligent processing method for the aerospace data comprises a processor and a memory storing a computer program, wherein the computer program is executed by the processor. All the implementation manners in the method embodiment are applicable to the embodiment, and the same technical effect can be achieved.
Embodiments of the present invention also provide a computer-readable storage medium including: and storing instructions, and enabling the computer to execute the intelligent processing method of the aerospace data when the instructions run on the computer. All the implementation manners in the above method embodiments are applicable to the embodiment, and the same technical effects can be achieved.
Those of ordinary skill in the art will appreciate that the various illustrative elements and algorithm steps described in connection with the embodiments disclosed herein may be implemented as electronic hardware, or combinations of computer software and electronic hardware. Whether such functionality is implemented as hardware or software depends upon the particular application and design constraints imposed on the solution. Skilled artisans may implement the described functionality in varying ways for each particular application, but such implementation decisions should not be interpreted as causing a departure from the scope of the present invention.
It will be clear to those skilled in the art that, for convenience and brevity of description, specific working procedures of the above-described systems, apparatuses and units may refer to corresponding procedures in the foregoing method embodiments, and are not repeated herein.
In the embodiments provided in the present invention, it should be understood that the disclosed apparatus and method may be implemented in other manners. For example, the apparatus embodiments described above are merely illustrative, e.g., the division of the units is merely a logical function division, and there may be additional divisions when actually implemented, e.g., multiple units or components may be combined or integrated into another system, or some features may be omitted or not performed. Alternatively, the coupling or direct coupling or communication connection shown or discussed with each other may be an indirect coupling or communication connection via some interfaces, devices or units, which may be in electrical, mechanical or other form.
The units described as separate units may or may not be physically separate, and units shown as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.
In addition, each functional unit in the embodiments of the present invention may be integrated in one processing unit, or each unit may exist alone physically, or two or more units may be integrated in one unit.
The functions, if implemented in the form of software functional units and sold or used as a stand-alone product, may be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention may be embodied essentially or in a part contributing to the prior art or in a part of the technical solution, in the form of a software product stored in a storage medium, comprising several instructions for causing a computer device (which may be a personal computer, a server, a network device, etc.) to perform all or part of the steps of the method according to the embodiments of the present invention. And the aforementioned storage medium includes: a usb disk, a removable hard disk, a ROM, a RAM, a magnetic disk, or an optical disk, etc.
Furthermore, it should be noted that in the apparatus and method of the present invention, it is apparent that the components or steps may be disassembled and/or assembled. Such decomposition and/or recombination should be considered as equivalent aspects of the present invention. Also, the steps of performing the series of processes described above may naturally be performed in chronological order in the order of description, but are not necessarily performed in chronological order, and some steps may be performed in parallel or independently of each other. It will be appreciated by those of ordinary skill in the art that all or any of the steps or components of the methods and apparatus of the present invention may be implemented in hardware, firmware, software, or a combination thereof in any computing device (including processors, storage media, etc.) or network of computing devices, as would be apparent to one of ordinary skill in the art after reading this description of the invention.
The object of the invention can thus also be achieved by running a program or a set of programs on any computing device. The computing device may be a well-known general purpose device. The object of the invention can thus also be achieved by merely providing a program product containing program code for implementing said method or apparatus. That is, such a program product also constitutes the present invention, and a storage medium storing such a program product also constitutes the present invention. It is apparent that the storage medium may be any known storage medium or any storage medium developed in the future. It should also be noted that in the apparatus and method of the present invention, it is apparent that the components or steps may be disassembled and/or assembled. Such decomposition and/or recombination should be considered as equivalent aspects of the present invention. The steps of executing the series of processes may naturally be executed in chronological order in the order described, but are not necessarily executed in chronological order. Some steps may be performed in parallel or independently of each other.
While the foregoing is directed to the preferred embodiments of the present invention, it will be appreciated by those skilled in the art that various modifications and adaptations can be made without departing from the principles of the present invention, and such modifications and adaptations are intended to be comprehended within the scope of the present invention.
Claims (10)
1. An intelligent processing method of aerospace data, which is characterized by being applied to a spacecraft, comprises the following steps:
acquiring data to be processed of a spacecraft;
performing data preprocessing on the data to be processed to obtain text data of the data to be processed;
inputting the text data into an intelligent processing model for data analysis processing to obtain a processing result of the data to be processed, and outputting the processing result;
the intelligent processing model is obtained by training a first preset processing model according to a historical processing data set of spacecraft data to be processed, and the first preset processing model is obtained by constructing the first preset processing model according to the data to be processed.
2. The method for intelligently processing the space data according to claim 1, wherein constructing a first preset processing model according to the data to be processed comprises:
constructing a data preprocessing function of a first preset processing model according to the data to be processed;
constructing an encoder of a first preset processing model based on the data preprocessing function;
a decoder for constructing a first preset processing model based on the encoder;
constructing a linear transformation function and a connection function of a first preset processing model based on the decoder;
constructing a loss function of a first preset processing model based on the connection function;
and constructing a back propagation function and a parameter updating function of a first preset processing model based on the loss function, and acquiring the first preset processing model.
3. The method for intelligently processing aerospace data according to claim 1, wherein constructing a data preprocessing function of a first preset processing model according to the data to be processed comprises:
acquiring a data sequence x to be processed according to the data to be processed;
and carrying out data preprocessing on the data sequence x to be processed to obtain a data preprocessing function of a first preset processing model.
4. The method of intelligent processing of aerospace data of claim 3, wherein constructing an encoder of a first pre-set processing model based on the data preprocessing function comprises:
determining the number of encoder layers N_encoder;
constructing an initial encoder, and performing N_encoder iterations on the initial encoder based on the data preprocessing function to obtain an encoder of a first preset processing model;
wherein each iteration performs multi-head attention mechanism, layer normalization and position feed-forward network operation on the current encoder output result.
5. The method of intelligent processing of aerospace data of claim 4, wherein constructing a decoder of a first pre-set processing model based on the encoder comprises:
determining the number of decoder layers N_decoder;
constructing an initial decoder according to the encoder of the first preset processing model;
performing N_decoder iterations on the initial decoder based on the output result of the encoder of the first preset processing model to obtain a decoder of the first preset processing model;
wherein each iteration performs masking multi-head attention mechanisms, layer normalization, encoding-decoding multi-head attention mechanisms, layer normalization, and position feed-forward network operations on the current decoder output result.
6. The method for intelligently processing aerospace data according to claim 1, wherein training the first preset processing model according to a historical processing data set of data to be processed of the spacecraft to obtain the intelligent processing model comprises:
acquiring a historical processing data set of the spacecraft data to be processed;
constructing an autonomous decision engine, integrating the autonomous decision engine with the first preset processing model, and acquiring a second preset processing model;
and training the second preset processing model through the historical processing data set to obtain an intelligent processing model.
7. The method of claim 6, wherein training the second predetermined processing model through the historical processing dataset to obtain an intelligent processing model comprises:
dividing the historical processing data set into a historical processing data training set and a historical processing data testing set;
training the second preset processing model through an antagonistic neural network based on the historical processing data training set to obtain an optimized processing model;
and repeatedly evaluating and iterating the optimization processing model based on the historical processing data test set to obtain an intelligent processing model.
8. An intelligent processing device for aerospace data, which is characterized by comprising:
the acquisition module is used for acquiring data to be processed of the spacecraft, carrying out data preprocessing on the data to be processed, and acquiring text data of the data to be processed;
the processing module is used for inputting the text data into an intelligent processing model for data analysis and processing, obtaining a processing result of the data to be processed and outputting the processing result; the intelligent processing model is obtained by training a first preset processing model according to a historical processing data set of spacecraft data to be processed, and the first preset processing model is obtained by constructing the first preset processing model according to the data to be processed.
9. A computing device, comprising: a processor, a memory storing a computer program which, when executed by the processor, performs the method of any one of claims 1 to 7.
10. A computer-readable storage medium, comprising: instructions stored which, when executed on a computer, cause the computer to perform the method of any one of claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310882558.6A CN116595339A (en) | 2023-07-19 | 2023-07-19 | Intelligent processing method, device and equipment for space data |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310882558.6A CN116595339A (en) | 2023-07-19 | 2023-07-19 | Intelligent processing method, device and equipment for space data |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116595339A true CN116595339A (en) | 2023-08-15 |
Family
ID=87606641
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310882558.6A Pending CN116595339A (en) | 2023-07-19 | 2023-07-19 | Intelligent processing method, device and equipment for space data |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116595339A (en) |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111080032A (en) * | 2019-12-30 | 2020-04-28 | 成都数之联科技有限公司 | Load prediction method based on Transformer structure |
| WO2021139297A1 (en) * | 2020-07-28 | 2021-07-15 | 平安科技（深圳）有限公司 | Question-answer method and question-answer apparatus based on transformer model, and storage apparatus |
| CN115599899A (en) * | 2022-11-08 | 2023-01-13 | 中国空气动力研究与发展中心计算空气动力研究所（Cn） | Intelligent question-answering method, system, equipment and medium based on aircraft knowledge graph |
| CN116383352A (en) * | 2023-03-10 | 2023-07-04 | 江苏科技大学 | Knowledge graph-based method for constructing field intelligent question-answering system by using zero samples |
-
2023
- 2023-07-19 CN CN202310882558.6A patent/CN116595339A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111080032A (en) * | 2019-12-30 | 2020-04-28 | 成都数之联科技有限公司 | Load prediction method based on Transformer structure |
| WO2021139297A1 (en) * | 2020-07-28 | 2021-07-15 | 平安科技（深圳）有限公司 | Question-answer method and question-answer apparatus based on transformer model, and storage apparatus |
| CN115599899A (en) * | 2022-11-08 | 2023-01-13 | 中国空气动力研究与发展中心计算空气动力研究所（Cn） | Intelligent question-answering method, system, equipment and medium based on aircraft knowledge graph |
| CN116383352A (en) * | 2023-03-10 | 2023-07-04 | 江苏科技大学 | Knowledge graph-based method for constructing field intelligent question-answering system by using zero samples |
Non-Patent Citations (3)
| Title |
|---|
| ASHISH VASWANI ET AL.: "Attention is all you need", 《ARXIV[CS.CL]》, pages 1 - 15 * |
| LONG OUYANG ET AL.: "Training language models to follow instructions with human feedback", 《ARXIV[CS.CL]》, pages 1 - 68 * |
| 马克西姆·拉潘: "《深度强化学习实践 原书第2版》", vol. 2, 机械工业出版社, pages: 282 - 284 * |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111124840B (en) | Method and device for predicting alarm in business operation and maintenance and electronic equipment | |
| CN110688288B (en) | Automatic test method, device, equipment and storage medium based on artificial intelligence | |
| CN111650453A (en) | Power equipment diagnosis method and system based on windowing characteristic Hilbert imaging | |
| CN113012811B (en) | Traditional Chinese medicine syndrome diagnosis and health evaluation method combining deep convolutional network and graph neural network | |
| CN112052948A (en) | Network model compression method and device, storage medium and electronic equipment | |
| Anand et al. | Black magic in deep learning: How human skill impacts network training | |
| CN114757432A (en) | Future execution activity and time prediction method and system based on flow log and multi-task learning | |
| CN113128689A (en) | Entity relationship path reasoning method and system for regulating knowledge graph | |
| CN115935277A (en) | Supercharged boiler fault diagnosis method based on small sample learning and training method and testing method of fault diagnosis model | |
| CN115618732A (en) | Nuclear reactor digital twin key parameter autonomous optimization data inversion method | |
| CN114638360A (en) | Computing platform and method for impulse neural network learning and simulation | |
| CN116595339A (en) | Intelligent processing method, device and equipment for space data | |
| CN108363738B (en) | Recommendation method for industrial equipment data analysis algorithm | |
| US20230206054A1 (en) | Expedited Assessment and Ranking of Model Quality in Machine Learning | |
| CN113743461B (en) | Unmanned aerial vehicle cluster health degree assessment method and device | |
| CN115345303A (en) | Convolutional neural network weight tuning method, device, storage medium and electronic equipment | |
| CN114720129A (en) | Rolling bearing residual life prediction method and system based on bidirectional GRU | |
| Arora et al. | A Review on Learning Planning Action Models for Socio-Communicative HRI | |
| CN116527411B (en) | Data security intelligent protection model construction method and device and collaboration platform | |
| CN114400049B (en) | Training method and device for peptide fragment quantitative model, computer equipment and storage medium | |
| Nowak et al. | Deep net triage: Analyzing the importance of network layers via structural compression | |
| CN115831356B (en) | Auxiliary prediction diagnosis method based on artificial intelligence algorithm | |
| CN115329906B (en) | Fault diagnosis method and device based on graph regularization network | |
| CN118155227B (en) | Nuclear power equipment maintenance decision-making method and system based on intelligent technology | |
| CN118246810A (en) | Software class discipline competition performance evaluation method and system based on improved dynamic Bayesian network |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |