Intelligent maintenance method, system and equipment for optical astronomical telescope
Through deep learning and knowledge graph technology, the imaging quality of astronomical telescopes is monitored in real time and the failures are diagnosed, solving the complex and time-consuming maintenance problems in the existing technology, and improving the maintenance efficiency and data output of the telescope.
Patent Information
- Application Number
- CN202111588927.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2021-12-23
- Publication Date
- 2024-11-05
- Estimated Expiration
- 2041-12-23
AI Technical Summary
Existing astronomical telescope maintenance technology is difficult to monitor imaging quality in real time and diagnose faults efficiently, resulting in complex and time-consuming maintenance processes, affecting data output.
Deep learning and knowledge graph technology are used to classify and extract star images through random deep forest algorithms and convolutional neural networks, and fault diagnosis is performed by combining fully connected neural networks to monitor imaging quality in real time and provide fault cause analysis.
Real-time monitoring of telescope imaging quality and rapid diagnosis of faults are achieved, maintenance operations are simplified, maintenance personnel burden is reduced, and telescope maintenance efficiency is improved.
Smart Images

Figure CN114299333B_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the technical field of astronomical telescopes, and in particular relates to an intelligent maintenance method, system and equipment for an optical astronomical telescope. Background Art
[0002] Astronomical telescope maintenance can ensure that the telescope works in the best condition and outputs high-quality data. The daily maintenance of astronomical telescopes is divided into the inspection of various equipment before observation and fault maintenance during observation. The efficiency of equipment inspection before observation generally does not affect night observation. Fault maintenance during observation is divided into fault monitoring, fault diagnosis, and replacement of faulty equipment. The maintenance process takes up observation time and affects data output. Failure to discover faults in time will directly affect the imaging quality of the telescope.
[0003] Efficient telescope maintenance can save observation time and increase data output. The key to efficient maintenance of astronomical telescopes lies in real-time monitoring of imaging quality and efficient diagnosis of faults. In order to achieve efficient maintenance of astronomical telescopes, the following technologies are used in the existing technology:
[0004] The Very Large Telescope and the Sloan Digital Sky Survey Telescope install sensors at key locations on the telescope to construct telescope performance monitors. When the sensor value exceeds the set threshold, a warning is displayed to remind maintenance personnel to perform maintenance.
[0005] The Laser Interferometer Gravitational-Wave Observatory organizes data from a large number of observation logs to construct a data query and suggestion system, which returns matching text information based on user input to help improve maintenance efficiency.
[0006] However, the monitoring results of the above-mentioned traditional telescope maintenance methods cannot directly reflect the imaging quality of the final telescope product, and the efficiency of fault diagnosis depends on the experience of the maintenance personnel. With the application of more and more advanced technologies on telescopes, such as active optics technology and adaptive optics technology, telescope maintenance tasks have become complicated and difficult, and it is difficult to adapt to the increasingly complex telescope maintenance tasks. Intelligent telescope maintenance technology is urgently needed. Summary of the invention
[0007] The purpose of the present invention is to provide an intelligent maintenance method, system and equipment for an optical astronomical telescope in response to the deficiencies in the prior art. The method, system and equipment are used to perform real-time monitoring and fault diagnosis of the imaging quality of the telescope based on artificial intelligence technologies such as deep learning and knowledge graphs for fault maintenance during the operation of the optical astronomical telescope, thereby simplifying maintenance operations, reducing the burden on maintenance personnel and improving the maintenance and operation efficiency of the telescope.
[0008] Specifically, the present invention is implemented by adopting the following technical solutions.
[0009] On the one hand, the present invention provides an optical astronomical telescope intelligent maintenance method, comprising:
[0010] The images obtained by the telescope terminal instrument are used as input, and the input images are classified using the preset random forest algorithm, and the images in which the telescope works normally and contains bright stars are retained as the first star image;
[0011] Performing star image extraction on the first star image to obtain a second star image, where the second star image is a small picture containing only a single bright star; using a preset convolutional neural network to obtain a spot shape of the star image in the second star image as a first image feature of the second star image; and obtaining a spot size of the second star image as a second image feature of the second star image;
[0012] The imaging quality data is output as the target image feature; the imaging quality data is represented by the first image feature of the second star image and the second image feature of the second star image.
[0013] Furthermore, the step of acquiring the image spot size of the second star image is to calculate the full width at half maximum of the second star image.
[0014] Furthermore, the optical astronomical telescope intelligent maintenance method also includes: issuing an alarm when the imaging quality data deviates from a preset normal threshold.
[0015] Furthermore, the random forest algorithm and convolutional neural network are trained in the following way:
[0016] Classifying the images obtained by the telescope terminal instrument to obtain a corresponding relationship between the images and the operating status of the telescope;
[0017] Automatically classify the first image features of the second star image using a K-nearest neighbor clustering algorithm, evaluate the automatic classification results according to preset rules, and determine the correspondence between the first image features of the second star image and the cause of the fault;
[0018] According to the relationship between the picture obtained by the telescope terminal instrument and the operating status of the telescope, and the correspondence between the first image feature of the second star image and the cause of the fault, the star image corresponding to the fault is obtained using historical data in a database of telescope faults and fault causes, the picture obtained by the telescope terminal instrument and the second star image are calibrated to construct a training set; the training set is used to train a random deep forest algorithm and a convolutional neural network algorithm to obtain a preset random deep forest algorithm and a convolutional neural network algorithm.
[0019] Furthermore, the optical astronomical telescope intelligent maintenance method further includes:
[0020] Taking the target image feature as a telescope imaging quality entity, querying the telescope imaging quality entity in a preset telescope maintenance knowledge graph, obtaining a node matching the telescope imaging quality entity, and obtaining a knowledge graph subgraph pointing to the node according to a preset matching depth, that is, obtaining a knowledge graph subgraph corresponding to the telescope imaging quality entity;
[0021] Converting the knowledge graph subgraph into events represented by a node chain, wherein the events are all possible fault causes that cause the target image feature to deviate from a normal threshold;
[0022] Use the telescope's sensor to obtain the properties of the nodes on the node chain [x11, x12…, x1 m ], where m is the number of node features, the attributes are used as node features, and the features of all nodes in a node chain are used as the feature vector of the node chain [x11, x12, … x1 m ,x21,x22…xn1,xn2…xn m ] and output it to the preset fully connected neural network, where n is the number of nodes on a node chain. The fully connected neural network gives the probability of the event represented by the node chain and outputs the node chain and its probability.
[0023] Furthermore, the method for constructing the preset telescope maintenance knowledge graph includes:
[0024] The target image feature is used as an imaging quality entity, and the telescope maintenance related entities are classified and sorted to obtain the relationship between the telescope maintenance related entities, and the relationship between the telescope maintenance related entities is sorted into a subject-predicate-object form; wherein the subject and the object are represented by the telescope maintenance related entities, and the predicate is represented by the relationship between the telescope maintenance related entities.
[0025] Furthermore, the event of converting the knowledge graph subgraph into a node chain representation includes converting the knowledge graph subgraph from a mesh structure to a tree structure, and retaining the relationship between nodes, and all nodes included in the path from all leaf nodes to the root node constitute a node chain.
[0026] On the other hand, the present invention also provides an intelligent maintenance system for an optical astronomical telescope, comprising a storage layer, a working layer and an interface layer;
[0027] The storage layer stores data and knowledge, including telescope maintenance knowledge graph, random deep forest algorithm, fully connected neural network and convolutional neural network algorithm, database of telescope failures and failure causes, and imaging quality records;
[0028] The interface layer displays the imaging quality and fault causes obtained by the intelligent maintenance system of the optical astronomical telescope, the environment and equipment parameters obtained by the sensor, and the pictures obtained by the telescope terminal instrument;
[0029] The working layer processes data during the working process of the optical astronomical telescope intelligent maintenance system, including an imaging quality monitoring module, wherein:
[0030] Taking the pictures obtained by the telescope terminal instrument as input, classifying the input pictures using a preset random deep forest algorithm, and retaining the pictures in which the telescope works normally and contains bright stars as the first star image;
[0031] Performing star image extraction on the first star image to obtain a second star image, where the second star image is a small picture containing only a single bright star; using a preset convolutional neural network to obtain a spot shape of the star image in the second star image as a first image feature of the second star image; and obtaining a spot size of the second star image as a second image feature of the second star image;
[0032] The imaging quality data is output as the target image feature; the imaging quality data is represented by the first image feature of the second star image and the second image feature of the second star image.
[0033] Furthermore, the working layer further comprises a fault diagnosis module, wherein:
[0034] Taking the target image feature as a telescope imaging quality entity, querying the telescope imaging quality entity in a preset telescope maintenance knowledge graph, obtaining a node matching the telescope imaging quality entity, and obtaining a knowledge graph subgraph pointing to the node according to a preset matching depth, that is, obtaining a knowledge graph subgraph corresponding to the telescope imaging quality entity;
[0035] Converting the knowledge graph subgraph into events represented by a node chain, wherein the events are all possible fault causes that cause the target image feature to deviate from a normal threshold;
[0036] Use the telescope's sensor to obtain the properties of the nodes on the node chain [x11, x12…, x1 m ], where m is the number of node features, the attributes are used as node features, and the features of all nodes in a node chain are used as the feature vector of the node chain [x11, x12, … x1 m ,x21,x22…xn1,xn2…xn m ] and output it to the preset fully connected neural network, where n is the number of nodes on a node chain. The fully connected neural network gives the probability of the event represented by the node chain and outputs the node chain and its probability.
[0037] On the other hand, the present invention also provides an intelligent maintenance device for an optical astronomical telescope, the device comprising a memory and a processor; the memory stores a computer program for implementing the intelligent maintenance method for the optical astronomical telescope, and the processor executes the computer program to implement the steps of the above method.
[0038] The beneficial effects of the optical astronomical telescope intelligent maintenance method, system and device of the present invention are as follows:
[0039] The optical astronomical telescope intelligent maintenance method, system and equipment of the present invention use the star image data obtained by the terminal instrument of the astronomical telescope as input through the imaging quality monitoring module, use the random deep forest algorithm to classify the star images, obtain pictures containing bright stars, and then use the convolutional neural network algorithm to classify the shapes of the star images, and output the star spot shape and size of the star image, so as to timely detect the abnormality of the star spot shape or size of the star image, and realize real-time monitoring of the imaging quality of the telescope.
[0040] The optical astronomical telescope intelligent maintenance method, system and equipment of the present invention use the star spot shape and size of the star image output by the imaging quality monitoring module as the telescope imaging quality entity through the fault diagnosis module, and use the knowledge graph to query to obtain the knowledge graph subgraph corresponding to the telescope imaging quality entity; then convert the knowledge graph subgraph into an event represented by a node chain; use the telescope's sensor to obtain the attributes of the nodes on the node chain, use the attributes as the node features to obtain the feature vector of the node chain, output the feature vector to the fully connected neural network, obtain the probability of the node chain representing the event being established, and finally output the node chain and its probability of establishment to the maintenance personnel as the basis for maintenance, thereby realizing the telescope imaging fault diagnosis.
[0041] The present invention combines artificial intelligence technology with the needs of telescope maintenance. It uses artificial intelligence technology to perform data analysis and decision support to replace manual monitoring and diagnosis of telescope faults. It gives full play to the advantages of deep learning in shape classification and knowledge graphs in knowledge representation. It uses the shape of star images to monitor imaging quality. The star spot shape of star images obtained by deep learning is used as the input of the knowledge graph for fault diagnosis. It solves the difficulties in real-time monitoring of telescope imaging quality and real-time diagnosis of faults. It can simplify maintenance operations, reduce the burden on maintenance personnel, and help improve the efficiency of telescope maintenance. BRIEF DESCRIPTION OF THE DRAWINGS
[0042] Figure 1 It is a schematic diagram of the system architecture of the present invention.
[0043] Figure 2 Schematic diagram of the workflow of the imaging quality monitoring module of this embodiment.
[0044] Figure 3It is a schematic diagram of the relationship between entities related to telescope maintenance in this embodiment.
[0045] Figure 4 Schematic diagram of the fault diagnosis module workflow of the embodiment.
[0046] Figure 5 It is a schematic diagram of the knowledge graph subgraph and the transformed tree structure of this embodiment. DETAILED DESCRIPTION
[0047] The present invention is further described in detail below in conjunction with embodiments and with reference to the accompanying drawings.
[0048] Embodiment 1:
[0049] One embodiment of the present invention is a method, system and device for intelligent maintenance of optical astronomical telescopes. Figure 1 As shown, the architecture of the intelligent maintenance system for optical astronomical telescopes includes a storage layer, a working layer and an interface layer.
[0050] The interface layer is used to display the imaging quality and fault causes obtained by the intelligent maintenance system of the optical astronomical telescope, the environmental and equipment parameters obtained by the sensor, and the star images obtained by the telescope terminal instrument. It feeds back the imaging quality and fault causes to the maintenance personnel, and issues an alarm when a telescope fails to remind the maintenance personnel to perform maintenance.
[0051] The storage layer is used to store data and knowledge, including telescope maintenance knowledge graph, random deep forest algorithm, fully connected neural network and convolutional neural network (CNN) algorithm, database of telescope failure and failure cause (the database contains historical data during the operation of the optical astronomical telescope intelligent maintenance system and is continuously updated), and image quality records. Among them, the knowledge graph is a semantic network that reveals the relationship between entities and can formally describe things in the real world and their relationships.
[0052] In the present invention, a telescope maintenance knowledge graph is preset to represent the relationships between telescope maintenance related entities. Telescope maintenance related entities refer to all entities related to telescope maintenance, which are divided into four categories: telescope imaging quality, site environment, telescope equipment, and staff. Each category includes many entities. For example, equipment can be divided into optical systems, original tops and auxiliary facilities, racks, control systems, terminal instruments, etc. The relationships between telescope maintenance related entities include the relationships between telescope imaging quality, telescope equipment, site environment, and staff.
[0053] The telescope maintenance knowledge graph is constructed in the following way:
[0054] The star spot shape and size obtained from the imaging quality monitoring module are used as imaging quality entities. Then, based on expert experience and observation logs accumulated during telescope maintenance, the telescope maintenance related entities are classified and sorted (for example, into four categories: imaging quality, site environment, equipment, and staff), and the relationships between telescope maintenance related entities are obtained, such as Figure 3 As shown. The relationships between telescope maintenance related entities are organized into the form of subject-predicate-object. Among them, the subject and object are represented by telescope maintenance related entities, and the predicate is represented by the relationship between telescope maintenance related entities (for example, including the impact of the site environment on imaging quality, the impact of equipment failure on imaging quality, the poor imaging quality caused by the staff's operation affecting equipment performance, and the impact of the site environment on the staff, to construct a complete telescope maintenance knowledge graph).
[0055] The working layer is used to process the data during the operation of the intelligent maintenance system of the optical astronomical telescope, including the imaging quality monitoring module.
[0056] Deep learning algorithms are widely used in pattern recognition, mining implicit rules from a large amount of historical data and using them for prediction or classification. In the present invention, real-time monitoring of the imaging quality of the telescope is carried out in order to detect faults in a timely manner and ensure high-quality imaging of the telescope. The normal star images obtained by the telescope and the star images caused by different faults have different characteristics. Therefore, the real-time monitoring problem of the telescope imaging quality can be converted into a feature classification problem through deep learning algorithms. The higher the classification accuracy, the higher the imaging quality monitoring accuracy.
[0057] The imaging quality monitoring module takes the image obtained by the terminal instrument of the astronomical telescope as input, uses the random deep forest algorithm to classify the star images, obtains the images containing bright stars, and then uses the convolutional neural network algorithm to classify the shapes of the star images and outputs the shape and size of the star images.
[0058] The specific workflow of the imaging quality monitoring module is as follows: Figure 2 As shown. Includes:
[0059] The first step is to classify the input image and obtain the first star image. That is, the image obtained by the telescope terminal instrument is used as input, and the input image is classified using the preset random forest algorithm, and the image of the telescope working normally and containing bright stars is retained as the first star image.
[0060] The second step is to obtain the first image feature and the second star image feature of the second star image. That is, extract the star image from the picture containing the bright star retained in the previous step to obtain a small picture containing only a single bright star (the second star image), and use the preset convolutional neural network to obtain the image spot shape of the star image in the small picture as the first image feature of the second star image. Calculate the full width at half maximum (FWHM, Full Width Half Maximum) of the second star image image to represent the image spot size of the star image as the second image feature of the second star image.
[0061] In the third step, the imaging quality data is represented by the first image feature of the second star image (the image spot shape of the star image) and the second image feature of the second star image (the image spot size of the star image), and the imaging quality data is output as the target image feature. Preferably, an alarm module can also be used to issue an alarm when the imaging quality data deviates from a preset normal threshold.
[0062] The above random forest algorithm and convolutional neural network are trained in the following way:
[0063] 1. Classify the images obtained by the telescope terminal instrument according to preset rules (such as maintenance expert experience and historical data) to obtain the corresponding relationship between the images and the telescope operation status (including changing sky areas, no bright stars, and normal observation); use the K-nearest neighbor clustering algorithm to automatically classify the image spot shapes of star images, evaluate the automatic classification results according to preset rules (such as expert experience), and obtain the corresponding relationship between the image spot shapes of star images and the cause of the fault;
[0064] 2. According to the relationship between the image obtained by the telescope terminal instrument and the operating status of the telescope, and the correspondence between the image spot shape of the star image and the cause of the fault, the star image corresponding to the fault is obtained by using the historical data in the database of telescope faults and fault causes, and the image obtained by the telescope terminal instrument and the star image obtained by cutting are calibrated to construct a training set and a test set; the training set is used to train the random deep forest algorithm and the convolutional neural network algorithm to obtain the preset random deep forest algorithm and the convolutional neural network algorithm. Among them, the preset random deep forest algorithm is used for real-time classification of the images obtained by the telescope terminal instrument, and the preset convolutional neural network algorithm is used for real-time classification of the image spot shape of the star image in the second star image.
[0065] Preferably, in another embodiment, the imaging quality detection module uses the above method to monitor the imaging of the guide camera of the astronomical telescope. At the same time, the excessive size of the star image spot may be caused by environmental factors such as poor seeing, or it may be caused by equipment such as ventilation and refrigeration in the dome. The imaging quality detection module obtains the image spot size of the star image by calculating the full width at half maximum (FWHM) of the star image spot in the imaging of the guide camera; according to the quality of the site environment, the size threshold of the image spot of different star images is set to determine whether the image spot shape of the star image deviates from normal or whether the image spot size of the star image is too large. When the imaging quality data deviates from the preset normal threshold (the image spot shape of the star image deviates from normal or the image spot size of the star image is too large), an alarm is issued.
[0066] The fundamental purpose of constructing the telescope fault diagnosis system of the present invention is to construct a telescope maintenance expert system. In order to improve maintenance efficiency and reduce the burden on maintenance personnel, in another embodiment, the working layer also includes a fault diagnosis module. When the quality monitoring module detects that the imaging quality data deviates from a preset normal threshold, the fault diagnosis module performs fault diagnosis.
[0067] The fault diagnosis module uses the image spot shape and size of the star image output by the imaging quality monitoring module as the telescope imaging quality entity, and uses the knowledge graph to query to obtain the knowledge graph subgraph corresponding to the telescope imaging quality entity; then the knowledge graph subgraph is converted into an event represented by a node chain; the telescope's sensors are used to obtain the attributes of the nodes on the node chain, and the attributes are used as node features to obtain the feature vector of the node chain, and the feature vector is output to the fully connected neural network to obtain the probability of the node chain representing the event being established, and finally the node chain and its probability of establishment (that is, all possible fault causes represented by the node chain) are output to the maintenance personnel as the basis for maintenance, thereby realizing telescope imaging fault diagnosis.
[0068] The specific workflow of the fault diagnosis module is as follows: Figure 4 As shown. Includes:
[0069] The first step is to receive the image spot shape and size of the star image output by the imaging quality monitoring module as the telescope imaging quality entity, query the telescope imaging quality entity in the preset telescope maintenance knowledge graph, obtain the node matching the telescope imaging quality entity, and obtain the knowledge graph subgraph pointing to the node according to the preset matching depth, that is, obtain the knowledge graph subgraph corresponding to the telescope imaging quality entity.
[0070] The second step is to convert the knowledge graph subgraph into events represented by node chains (i.e. possible fault causes). For example, convert the knowledge graph subgraph from a mesh structure to a tree structure, such as Figure 5 As shown, the relationship between nodes is retained, and all nodes included in the path from leaf nodes to root nodes constitute a node chain (such as Figure 5 5→1→A0). A node chain is an event that contains a causal relationship (the relationship between the deviation of the telescope imaging quality data from the normal threshold and the cause of the failure, represented by the subject, predicate, and object). It can be used to represent all possible causes of failure that cause the telescope imaging quality data to deviate from the normal threshold.
[0071] The third step is to use the telescope’s sensor to obtain the attributes of the nodes on the node chain [x11, x12…, x1 m ], where m is the number of node features. For example, the displacement actuator can obtain the force magnitude and actuator installation time. The number of node features of the displacement actuator is 2. The attributes are used as node features, and the features of all nodes in a node chain are used as the feature vector of the node chain [x11, x12, … x1 m ,x21,x22…xn1,xn2…xn m ], and output it to the preset fully connected neural network (FCNN, Fully Connected Feedforward Neural Network), where n is the number of nodes in a node chain. The results of the fully connected neural network are set as two categories: event establishment and failure establishment. The fully connected neural network gives the establishment probability of the event represented by the node chain (i.e., the possible cause of the fault), and finally outputs the node chain and its probability to the maintenance personnel as the basis for maintenance.
[0072] Embodiment 2:
[0073] Another embodiment of the present invention takes LAMOST (Large Sky Area Multi-Object FiberSpectroscopy Telescope) as an example to describe the optical astronomical telescope intelligent maintenance method, system and equipment of the present invention. LAMOST is a large-field telescope (Wang-Su active reflection Schmidt system) with China's independent intellectual property rights and original creation, the largest aperture in the world, which breaks the bottleneck that large field of view and large aperture cannot be achieved at the same time in the world, and leads the trend of world spectral surveys. LAMOST has an equivalent aperture of 3.6-4.9 meters and a field of view diameter of 5 degrees, and can obtain spectra of up to 4,000 celestial bodies at the same time. The 8 guide cameras on the focal plane of LAMOST can be used for real-time star image quality monitoring to construct the LAMOST star image quality real-time monitoring system.
[0074] The machine learning algorithms in this embodiment require a large amount of data for training and testing, including random deep forest algorithms, convolutional neural networks, and fully connected neural networks, and the tensorflow software package is used to build the algorithms. The historical data accumulated by LAMOST over the years is used to calibrate the historical data, construct a data set containing various types of data, and divide the data set into a test set and a training set at a ratio of 3:7. The random deep forest algorithm is used to roughly classify the pictures obtained by the telescope, and classify the pictures obtained by the telescope into pictures in the process of changing the sky area, pictures containing bright stars, and pictures without bright stars; the convolutional neural network classifies the pictures obtained by the telescope into round, rod-shaped, two-point-shaped, donut-shaped, and cluster-shaped. The fully connected network is used to estimate the probability of the event represented by the node chain, and classify the historical data according to whether the event is true or not.
[0075] The training set and the test set are used to adjust the parameters of the machine learning algorithm in this embodiment, and the trained machine learning parameters are saved. The knowledge graph of this embodiment also requires a lot of expert experience to build. After obtaining the entity and the relationship data between entities, it is built using the neo4j industrial knowledge graph software. After building the required artificial intelligence algorithms, the specific steps of the optical astronomical telescope intelligent maintenance method of this embodiment are as follows:
[0076] 1. Take the pictures obtained by the 8 guide cameras as the input of the imaging quality monitoring system, extract the features of the pictures, and use the random forest algorithm to classify the pictures. If the classification result shows that the picture contains bright stars obtained by normal observation, proceed to the next step; if the classification result shows that the telescope is in the sky change area or there are no bright stars, directly display the current state of the telescope as in the sky change area or no bright stars.
[0077] 2. Extract star images from pictures containing bright stars to obtain small pictures containing only one bright star. Use convolutional neural network to obtain the star spot shape of the star image. Use half-height full width to represent the star spot size of the star image. Compare the obtained star spot size and shape with the preset threshold (for example, the star image is circular and the star spot size of the star image is set to 3 arc seconds). If the star spot size and shape of the telescope star image are normal, the output is that the telescope status is normal; otherwise, the alarm system sounds an alarm and outputs the star spot shape and size of the star image to the fault diagnosis module.
[0078] 3. For the input star spot shape and size, the fault diagnosis module gives a knowledge graph subgraph containing nodes matching the input, converts the knowledge graph subgraph into a knowledge tree represented by a tree structure, and takes the node chain composed of the nodes on the path from the leaf node to the root node as the cause of the telescope failure.
[0079] 4. Use the telescope's installed environment sensors and the sensors of the telescope device to obtain the properties of the nodes on the node chain (such as wind speed, temperature, force, etc.), use the node properties as node features, and use multiple node features of the same node chain as a feature vector, and output them to the fully connected neural network. The fully connected neural network gives the probability of the event represented by the node chain. Finally, the node chain and the probability of the node chain are output to the maintenance personnel as the diagnosis results of the telescope failure.
[0080] The optical astronomical telescope intelligent maintenance method, system and equipment of the present invention use the star image data obtained by the terminal instrument of the astronomical telescope as input through the imaging quality monitoring module, use the random deep forest algorithm to classify the star images, obtain pictures containing bright stars, and then use the convolutional neural network algorithm to classify the shapes of the star images, and output the star spot shape and size of the star image, so as to timely detect the abnormality of the star spot shape or size of the star image, and realize real-time monitoring of the imaging quality of the telescope.
[0081] The optical astronomical telescope intelligent maintenance method, system and equipment of the present invention use the star spot shape and size of the star image output by the imaging quality monitoring module as the telescope imaging quality entity through the fault diagnosis module, and use the knowledge graph to query to obtain the knowledge graph subgraph corresponding to the telescope imaging quality entity; then convert the knowledge graph subgraph into an event represented by a node chain; use the telescope's sensor to obtain the attributes of the nodes on the node chain, use the attributes as the node features to obtain the feature vector of the node chain, output the feature vector to a fully connected neural network, obtain the probability of the node chain representing the event being established, and finally output the node chain and its probability of establishment to the maintenance personnel as the basis for maintenance, thereby realizing real-time diagnosis of telescope imaging faults.
[0082] The present invention combines artificial intelligence technology with the needs of telescope maintenance. It uses artificial intelligence technology to perform data analysis and decision support to replace manual monitoring and diagnosis of telescope faults. It gives full play to the advantages of deep learning in shape classification and knowledge graphs in knowledge representation. It uses the shape of star images to monitor imaging quality. The star spot shape of star images obtained by deep learning is used as the input of the knowledge graph for fault diagnosis. It solves the difficulties in real-time monitoring of telescope imaging quality and real-time diagnosis of faults. It can simplify maintenance operations, reduce the burden on maintenance personnel, and help improve the efficiency of telescope maintenance.
[0083] In some embodiments, certain aspects of the above-mentioned techniques may be implemented by one or more processors of a processing system executing software. The software includes one or more executable instruction sets stored or otherwise tangibly implemented on a non-transitory computer-readable storage medium. The software may include instructions and certain data that manipulate one or more processors to perform one or more aspects of the above-mentioned techniques when executed by one or more processors. Non-transitory computer-readable storage media may include, for example, magnetic or optical disk storage devices, solid-state storage devices such as flash memory, cache, random access memory (RAM), or other non-volatile memory devices. The executable instructions stored on the non-transitory computer-readable storage medium may be source code, assembly language code, object code, or other instruction formats interpreted or otherwise executed by one or more processors.
[0084] Computer-readable storage media may include any storage medium or combination of storage media that can be accessed by a computer system during use to provide instructions and / or data to the computer system. Such storage media may include, but are not limited to, optical media (e.g., compact disks (CDs), digital versatile disks (DVDs), Blu-ray disks), magnetic media (e.g., floppy disks, tapes, or magnetic hard drives), volatile memory (e.g., random access memory (RAM) or cache), non-volatile memory (e.g., read-only memory (ROM) or flash memory), or micro-electromechanical system (MEMS)-based storage media. Computer-readable storage media may be embedded in a computing system (e.g., system RAM or ROM), fixedly attached to a computing system (e.g., a magnetic hard drive), removably attached to a computing system (e.g., an optical disk or a universal serial bus (USB)-based flash memory), or coupled to a computer system via a wired or wireless network (e.g., network accessible storage (NAS)).
[0085] Please note that not all activities or elements in the general description above are required, that a portion of a particular activity or device may not be required, and that one or more further activities or included elements may be performed in addition to those described. Further, the order in which the activities are listed is not necessarily the order in which they are performed. Moreover, these concepts have been described with reference to specific embodiments. However, those of ordinary skill in the art recognize that various modifications and changes may be made without departing from the scope of the present disclosure as set forth in the following claims. Therefore, the specification and drawings are to be regarded as illustrative rather than restrictive, and all such modifications are included within the scope of the present disclosure.
[0086] Benefits, other advantages, and solutions to problems have been described above with respect to specific embodiments. However, benefits, advantages, solutions to problems, and any features that may cause any benefit, advantage, or solution to occur or become more apparent should not be construed as key, essential, or essential features of any or all claims in any or other respects. In addition, the specific embodiments disclosed above are merely illustrative, as the disclosed subject matter may be modified and implemented in different but equivalent ways that are apparent to those skilled in the art having the benefit of the teachings herein. There is no intention to limit the details of construction or design shown herein, except as described in the claims. It is therefore apparent that the specific embodiments disclosed above may be altered or modified, and all such variations are considered to be within the scope of the disclosed subject matter.
Claims
1. An intelligent maintenance method for an optical astronomical telescope, characterized in that: include: The images obtained by the telescope terminal instrument are used as input, and the input images are classified using the preset random forest algorithm, and the images in which the telescope works normally and contains bright stars are retained as the first star image; Performing star image extraction on the first star image to obtain a second star image, where the second star image is a small picture containing only a single bright star; using a preset convolutional neural network to obtain a spot shape of the star image in the second star image as a first image feature of the second star image; and obtaining a spot size of the second star image as a second image feature of the second star image; Outputting imaging quality data as target image features; the imaging quality data is represented by the first image features of the second star image and the second image features of the second star image; Taking the target image feature as a telescope imaging quality entity, querying the telescope imaging quality entity in a preset telescope maintenance knowledge graph, obtaining a node matching the telescope imaging quality entity, and obtaining a knowledge graph subgraph pointing to the node according to a preset matching depth, that is, obtaining a knowledge graph subgraph corresponding to the telescope imaging quality entity; Converting the knowledge graph subgraph into events represented by a node chain, wherein the events are all possible fault causes that cause the target image feature to deviate from a normal threshold; Use the telescope's sensor to obtain the properties of the nodes on the node chain [x11, x12…, x1 m ], where m is the number of node features, the attributes are used as node features, and the features of all nodes in a node chain are used as the feature vector of the node chain [x11, x12, … x1 m ,x21,x22…xn1,xn2…xn m ], output to the preset fully connected neural network, where n is the number of nodes on a node chain, the fully connected neural network gives the probability of the event represented by the node chain, and outputs the node chain and its probability; The method for constructing the preset telescope maintenance knowledge graph includes: The target image feature is used as an imaging quality entity, and the telescope maintenance related entities are classified and sorted to obtain the relationship between the telescope maintenance related entities, and the relationship between the telescope maintenance related entities is sorted into a subject-predicate-object form; wherein the subject and the object are represented by the telescope maintenance related entities, and the predicate is represented by the relationship between the telescope maintenance related entities; The event of converting the knowledge graph subgraph into a node chain representation includes converting the knowledge graph subgraph from a mesh structure to a tree structure, and retaining the relationship between the nodes, and all nodes included in the path from all leaf nodes to the root node constitute a node chain.
2. The intelligent maintenance method for an optical astronomical telescope according to claim 1, characterized in that: The step of obtaining the image spot size of the second star image is to calculate the full width at half maximum of the second star image.
3. The intelligent maintenance method for an optical astronomical telescope according to claim 1, characterized in that: Also includes: An alarm is issued when the imaging quality data deviates from a preset normal threshold.
4. The intelligent maintenance method for an optical astronomical telescope according to claim 1, characterized in that: The random forest algorithm and convolutional neural network are trained in the following way: Classifying the images obtained by the telescope terminal instrument to obtain a corresponding relationship between the images and the operating status of the telescope; Automatically classify the first image features of the second star image using a K-nearest neighbor clustering algorithm, evaluate the automatic classification results according to preset rules, and determine the correspondence between the first image features of the second star image and the cause of the fault; According to the relationship between the picture obtained by the telescope terminal instrument and the operating status of the telescope, and the correspondence between the first image feature of the second star image and the cause of the fault, the star image corresponding to the fault is obtained using historical data in a database of telescope faults and fault causes, the picture obtained by the telescope terminal instrument and the second star image are calibrated to construct a training set; the training set is used to train a random deep forest algorithm and a convolutional neural network algorithm to obtain a preset random deep forest algorithm and a convolutional neural network algorithm.
5. An intelligent maintenance system for an optical astronomical telescope, characterized in that: Includes storage layer, working layer and interface layer; The storage layer stores data and knowledge, including telescope maintenance knowledge graph, random deep forest algorithm, fully connected neural network and convolutional neural network algorithm, database of telescope failures and failure causes, and imaging quality records; The interface layer displays the imaging quality and fault causes obtained by the intelligent maintenance system of the optical astronomical telescope, the environment and equipment parameters obtained by the sensor, and the pictures obtained by the telescope terminal instrument; The working layer processes data during the working process of the intelligent maintenance system for optical astronomical telescopes, and includes an imaging quality monitoring module and a fault diagnosis module. The imaging quality monitoring module: Taking the pictures obtained by the telescope terminal instrument as input, classifying the input pictures using a preset random deep forest algorithm, and retaining the pictures in which the telescope works normally and contains bright stars as the first star image; Performing star image extraction on the first star image to obtain a second star image, where the second star image is a small picture containing only a single bright star; using a preset convolutional neural network to obtain a spot shape of the star image in the second star image as a first image feature of the second star image; and obtaining a spot size of the second star image as a second image feature of the second star image; Outputting imaging quality data as target image features; the imaging quality data is represented by the first image features of the second star image and the second image features of the second star image; The fault diagnosis module: Taking the target image feature as a telescope imaging quality entity, querying the telescope imaging quality entity in a preset telescope maintenance knowledge graph, obtaining a node matching the telescope imaging quality entity, and obtaining a knowledge graph subgraph pointing to the node according to a preset matching depth, that is, obtaining a knowledge graph subgraph corresponding to the telescope imaging quality entity; Converting the knowledge graph subgraph into events represented by a node chain, wherein the events are all possible fault causes that cause the target image feature to deviate from a normal threshold; Use the telescope's sensor to obtain the properties of the nodes on the node chain [x11, x12…, x1 m ], where m is the number of node features, the attributes are used as node features, and the features of all nodes in a node chain are used as the feature vector of the node chain [x11, x12, … x1 m ,x21,x22…xn1,xn2…xn m ], output to the preset fully connected neural network, where n is the number of nodes on a node chain, the fully connected neural network gives the probability of the event represented by the node chain, and outputs the node chain and its probability; The method for constructing the preset telescope maintenance knowledge graph includes: The target image feature is used as an imaging quality entity, and the telescope maintenance related entities are classified and sorted to obtain the relationship between the telescope maintenance related entities, and the relationship between the telescope maintenance related entities is sorted into a subject-predicate-object form; wherein the subject and the object are represented by the telescope maintenance related entities, and the predicate is represented by the relationship between the telescope maintenance related entities; The event of converting the knowledge graph subgraph into a node chain representation includes converting the knowledge graph subgraph from a mesh structure to a tree structure, and retaining the relationship between the nodes, and all nodes included in the path from all leaf nodes to the root node constitute a node chain.
6. An intelligent maintenance device for an optical astronomical telescope, characterized in that: The device includes a memory and a processor; the memory stores a computer program for implementing an intelligent maintenance method for an optical astronomical telescope, and the processor executes the computer program to implement the steps of any method according to claims 1-4.
