Early warning method for sediments in vacuum furnace based on multi-modal sensing

Abstract

The invention relates to the technical field of vacuum furnace early warning, and provides a vacuum furnace internal sediment early warning method based on multi-mode perception. The method comprises the following steps: acquiring a historical material record; extracting a first historical record; performing multi-dimensional feature collection to obtain first material feature information; traversing the deposition database to obtain target deposition data; obtaining a first key feature, and selectively activating the multi-mode sensing device to obtain first key sensing information; taking the first key sensing information as input information of an in-furnace deposition discriminator to obtain output information; and when the output information accords with the discrimination constraint, sending out an early warning instruction, and carrying out early warning on existence of the first sediment. The technical problem that sediment early warning is inaccurate and not timely due to the fact that a traditional early warning technology cannot comprehensively monitor the formation condition of the sediment in the vacuum furnace is solved, and the effects of recognizing the sediment through multi-mode sensing and improving the accuracy and timeliness of sediment early warning are achieved.

Description

Technical Field

[0001] This application relates to the technical field of industrial furnace warning, specifically to the technical field of vacuum furnace warning, and particularly to a method for warning of in-furnace deposits in a vacuum furnace based on multi-modal perception. Background Art

[0002] With the continuous progress of industrial production and the rapid development of process technologies, vacuum furnaces, as important heat treatment equipment, play a crucial role in fields such as material processing, metallurgy, and semiconductor manufacturing. However, with the continuous expansion of production scale and the increasingly strict process requirements, the problems faced by vacuum furnaces during operation are becoming increasingly complex, especially the accumulation and adhesion of in-furnace deposits. During the operation of a vacuum furnace, the formation of in-furnace deposits is inevitable. These deposits may originate from the volatilization of raw materials, the precipitation of reaction products, and the pollution of the furnace environment, etc. The accumulation of deposits not only affects the heating efficiency of the vacuum furnace but may also have a serious impact on product quality and even lead to equipment failures and safety accidents. Traditional methods for warning of in-furnace deposits in vacuum furnaces are inaccurate and untimely in warning of deposits because they cannot comprehensively monitor the formation of in-furnace deposits. Therefore, how to achieve real-time and accurate warning of in-furnace deposits in vacuum furnaces has become an urgent problem to be solved in current industrial production. Summary of the Invention

[0003] This application provides a method for warning of in-furnace deposits in a vacuum furnace based on multi-modal perception, aiming to solve the technical problems that traditional warning technologies are inaccurate and untimely in warning of deposits because they cannot comprehensively monitor the formation of in-furnace deposits in vacuum furnaces.

[0004] In view of the above problems, this application provides a method for warning of in-furnace deposits in a vacuum furnace based on multi-modal perception.

[0005] The present application provides a method for warning of deposits in a vacuum furnace based on multi-modal perception. The method includes: obtaining historical processing material records, which refer to the material processing records of a vacuum furnace after a predetermined maintenance inspection; extracting a first historical record from the historical processing material records, where the first historical record refers to the historical processing record of a first material; reading a predetermined feature dimension, and collecting multi-dimensional features of the first material based on the predetermined feature dimension to obtain first material feature information; traversing a deposition database based on the first material feature information to obtain target deposit data of a target material; obtaining a first key feature of a first deposit in the target deposit data, and selectively activating a multi-modal perception device according to the first key feature to obtain first key perception information; using the first key perception information as input information of an in-furnace deposition discriminator to obtain output information, where the in-furnace deposition discriminator is an intelligent model trained based on the principle of support vector machines; when the output information meets the discrimination constraint, issuing a warning instruction, and warning of the existence of the first deposit in the vacuum furnace based on the warning instruction.

[0006] One or more technical solutions provided in the present application have at least the following technical effects or advantages: For the above method for warning of deposits in a vacuum furnace based on multi-modal perception, the method obtains the material processing records of a vacuum furnace after a predetermined maintenance inspection, that is, historical processing material records, and then extracts the first historical record of the first material from them to analyze the possible deposits. To comprehensively analyze the characteristics of the first material, multi-dimensional feature collection is performed according to the predetermined feature dimension to obtain first material feature information. Subsequently, based on these feature information, the deposition database is traversed to retrieve the target deposit data related to the first material. After obtaining the target deposit data, the key features of the first deposit are extracted. According to these key features, the corresponding multi-modal perception device is selected for activation, and the first key perception information is obtained. Then, these perception information is input into an in-furnace deposition discriminator trained based on the principle of support vector machines to determine whether there is a first deposit in the furnace. If the output information of the in-furnace deposition discriminator meets the preset discrimination constraint, a warning instruction is issued to remind the operator of the risk of the first deposit in the vacuum furnace. Through this process, the effect of improving the accuracy and timeliness of deposit warning can be achieved, ensuring that the operator takes corresponding measures for processing in a timely manner.

[0007] The above description is only an overview of the technical solutions of the present application. In order to be able to understand the technical means of the present application more clearly, it can be implemented according to the content of the specification. And in order to make the above and other purposes, features and advantages of the present application more obvious and understandable, the following specifically illustrates the embodiments of the present application. Description of the Drawings

[0008] To more clearly illustrate the technical solutions in the embodiments of the present invention, the following will briefly introduce the accompanying drawings required for description in the embodiments. Obviously, the accompanying drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other accompanying drawings can be obtained based on these drawings.

[0009] Figure 1 It is a schematic flow chart of a method for warning of deposits in a vacuum furnace based on multi-modal perception in an embodiment; Figure 2 It is a schematic flow chart of obtaining target deposit data of a method for warning of deposits in a vacuum furnace based on multi-modal perception in an embodiment. Specific embodiments

[0010] By providing a method for warning of deposits in a vacuum furnace based on multi-modal perception in the embodiments of the present application, the technical problem that traditional warning technologies cannot comprehensively monitor the formation of deposits in the vacuum furnace, resulting in inaccurate and untimely deposit warnings, is solved.

[0011] The following will clearly and completely describe the technical solutions in the embodiments of the present application with reference to the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only some embodiments of the present application, rather than all embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present application without creative efforts belong to the scope of protection of the present application.

[0012] It should be noted that the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product or server including a series of steps or units does not necessarily have to be limited to those steps or units clearly listed, but may include other steps or modules not clearly listed or inherent to these processes, methods, products or devices.

[0013] Embodiment As Figure 1 shown, the present application provides a method for warning of deposits in a vacuum furnace based on multi-modal perception, and the method includes: Obtain historical processing material records, where the historical processing material records refer to the material processing records of a vacuum furnace after undergoing predetermined maintenance and inspection processing.

[0014] Vacuum furnaces operate in high-temperature and high-vacuum environments, with extremely high requirements for the purity of the furnace interior environment and the stability of equipment operation. However, in actual production processes, due to various factors such as impure raw materials, generation of reaction by-products, and equipment aging, deposits often form inside vacuum furnaces. These deposits not only affect the normal operation of the equipment but may also have a serious impact on product quality.

[0015] In the embodiments of this application, in order to achieve effective early warning of deposits inside the vacuum furnace, the system terminal first obtains the material processing records of the vacuum furnace after a predetermined maintenance inspection, that is, historical processed material records. This historical processed material record includes the processing records of different materials between each two maintenances. This predetermined maintenance inspection refers to regularly performing maintenance inspections on the vacuum furnace and ensuring that there are no deposits inside the furnace after the maintenance inspection is completed. By collecting and organizing the vacuum furnace processing data before the maintenance inspection, historical processed material records can be generated, and these records will serve as the data source for subsequent analysis of possible deposits.

[0016] Extract the first historical record from the historical processed material record, where the first historical record refers to the historical processing record of the first material.

[0017] In one embodiment, after obtaining the historical processed material record, the system terminal parses the historical processed material record. Different material corresponding historical processing records are stored in this historical processed material record. The system terminal extracts the historical processing record of the first material from these historical processing records and denotes it as the first historical record. The first material refers to the material processed by the vacuum furnace for the first time after a predetermined maintenance inspection. By extracting this first historical record, the system terminal can understand the material characteristics of the first material and, in combination with professional knowledge and historical experience, analyze possible carbides. If carbides may exist, the system terminal will detect the carbides in the vacuum furnace to determine whether carbide deposits actually exist. If the detection result is negative, the system terminal will repeat this process, extract the second historical record, and perform carbide deposit judgment until a new predetermined maintenance inspection.

[0018] Read the predetermined feature dimension and collect multi-dimensional features of the first material based on the predetermined feature dimension to obtain the first material feature information.

[0019] In one embodiment, the system terminal reads the pre-set characteristic dimensions, which are important bases for evaluating the properties of the material. Subsequently, according to these pre-set characteristic dimensions, detailed feature collection is performed on the first material to obtain specific information of the first material in each characteristic dimension. After multi-dimensional feature collection, the system terminal will obtain detailed feature information about the first material and organize this information to obtain the first material feature information. This first material feature information will be used to traverse the deposition database subsequently to ensure the accurate acquisition of sensing information.

[0020] Furthermore, the present application provides pre-set characteristic dimensions, including: The pre-set characteristic dimensions include the in-furnace process dimension, the material property dimension, and the in-furnace atmosphere dimension.

[0021] Preferably, the pre-set characteristic dimensions are a set of key indicators set for comprehensively evaluating the material processing situation in the vacuum furnace. These dimensions cover the in-furnace process dimension, the material property dimension, and the in-furnace atmosphere dimension. The in-furnace process dimension focuses on the specific process parameters and procedures involved in the material processing in the vacuum furnace, such as heating temperature, heating time, cooling rate, etc. These process parameters directly affect the reaction and change of the material in the furnace, so they are important aspects for evaluating the material processing effect. The material property dimension refers to some inherent properties of the material itself, such as chemical composition, physical state, particle size distribution, etc. These properties determine the behavior and reaction degree of the material in the vacuum furnace and are important bases for analyzing the material processing results. The in-furnace atmosphere dimension focuses on the gas environment and composition inside the vacuum furnace, such as gas type, gas pressure, gas purity, etc. The in-furnace atmosphere has an important impact on the reaction of the material and the formation of deposits, so it is also an aspect that cannot be ignored when evaluating the material processing situation. By comprehensively collecting the relevant information of these three dimensions, the effect of material processing in the vacuum furnace and the risk of deposit formation can be comprehensively evaluated, providing strong data support for subsequent analysis and early warning.

[0022] Traverse the deposition database based on the first material feature information to obtain the target deposit data of the target material.

[0023] In one embodiment, after obtaining the first material characteristic information, the system terminal traverses the deposition database. This deposition database stores the deposition data of all materials except the first material processed by the vacuum furnace. During the traversal process, the system terminal calls the predetermined feature dimension to perform multi-dimensional feature collection on the traversed first deposition data group to obtain the corresponding material characteristic information. Subsequently, the first material characteristic information and the material characteristic information corresponding to the first deposition data are compared and analyzed according to the predetermined feature comparison strategy. If the comparison result is similar, the first deposition data group is used as the target deposition data, and the material corresponding to the first deposition data group is used as the target material. Repeat this process until all the data in the deposition database is traversed, thereby gradually improving the target deposition data of the target material.

[0024] Further, if Figure 2 As shown, the present application provides methods for obtaining target sediment data, including: A first deposition data group is extracted from the deposition database, wherein the first deposition data group refers to deposition data of a second material processed by the vacuum furnace; and multi-dimensional features of the second material are collected based on the predetermined feature dimension to obtain feature information of the second material.

[0025] Preferably, the system terminal extracts the deposition data group traversed in the deposition database to obtain a first deposition data group. This first deposition data group is deposition data about the vacuum furnace when processing the second material, that is, the second material. Subsequently, using the same method as the aforementioned method for obtaining the first material feature information, multi-dimensional feature collection is performed on the second material through a predetermined feature dimension to obtain the second material feature information, which is used to calculate the material feature similarity index with the first material feature information.

[0026] A predetermined feature comparison strategy is read, and a comparison analysis is performed between the first material feature information and the second material feature information based on the predetermined feature comparison strategy to obtain a material feature similarity index; when the material feature similarity index reaches a similarity index limit, the second material is used as the target material, and the first deposition data group is used as the target deposition data.

[0027] Preferably, after obtaining the second material characteristic information, the system terminal reads a predetermined characteristic comparison strategy. This predetermined characteristic comparison strategy is used to guide how to compare the characteristic information of different materials and how to quantify the differences between these characteristics. Subsequently, according to the read predetermined characteristic comparison strategy, the obtained first material characteristic information and second material characteristic information are compared and analyzed. This comparison and analysis involve information marking, mark statistics, and ratio calculation. After that, based on the comparison and analysis, the system terminal can calculate a material characteristic similarity index. This index is a quantitative value used to represent the similarity degree of the characteristics of the first material and the second material and is the result of ratio calculation. Then, the calculated material characteristic similarity index is compared with a preset similarity index limit value. This similarity index limit value is set according to historical experience and is used to judge whether the characteristics of the two materials are similar enough. If the material characteristic similarity index reaches the similarity index limit value, then the system terminal determines that the second material is similar enough to the first material in characteristics. Therefore, the second material is used as the target material, and the first deposition data set related to the second material is used as the target deposition data. This is because if two materials are similar enough in characteristics, then the generated deposition data sets will also be similar enough. Therefore, the latest deposition data set is used as the target deposition data, and the latest material is used as the target material.

[0028] Furthermore, the present application provides a method for calculating a material characteristic similarity index, including: The predetermined characteristic comparison strategy includes a predetermined characteristic marking scheme; based on the predetermined characteristic marking scheme, the first material characteristic information and the second material characteristic information are sequentially marked with information to obtain a first material marking vector and a second material marking vector respectively.

[0029] Optionally, during the comparative analysis of material characteristic information, the system terminal reads a predetermined characteristic comparison strategy, which includes a predetermined characteristic marking scheme. This scheme details how to mark various characteristics of the material. The purpose of marking is to convert each characteristic of the material into a specific and quantifiable vector. The system terminal marks the characteristic information of the first material and the second material respectively according to this predetermined characteristic marking scheme. First, specific characteristic items under the in-furnace process dimension, material property dimension, and in-furnace atmosphere dimension are extracted from the characteristic information of the first material and the second material respectively, and the characteristic set to be marked is determined. Subsequently, each characteristic item is quantified. For numerical characteristics, such as temperature in the in-furnace process dimension and pressure in the in-furnace atmosphere dimension, their measured values are directly used; for non-numerical characteristics, such as gas types in the in-furnace atmosphere dimension and physical states in the material property dimension, one-hot encoding is used to convert them into numerical forms. Suppose there are three categories of gas types: nitrogen, oxygen, and argon. Then nitrogen may be encoded as [1, 0, 0], oxygen as [0, 1, 0], and argon as [0, 0, 1]. After that, all the marked characteristic values of the first material and the second material are grouped and arranged according to the dimension, forming two multi-dimensional vectors. These two multi-dimensional vectors are the first material marking vector and the second material marking vector, which represent the characteristic distributions of the first material and the second material on the selected characteristic set.

[0030] Construct a first set of information marking pairs with consistent markings for the first material marking vector and the second material marking vector, and count the first quantity of the first set of information marking pairs; construct a second set of information marking pairs with inconsistent markings for the first material marking vector and the second material marking vector, and count the second quantity of the second set of information marking pairs.

[0031] Optionally, after obtaining the first material marking vector and the second material marking vector, the system terminal compares each marking in the first material marking vector and the second material marking vector one by one. If the markings at a certain position in the two vectors are the same, that is, the represented codes are consistent, the system terminal regards this pair of markings as an information marking pair with consistent markings and adds it to the first set of information marking pairs. Subsequently, count the number of information marking pairs in this set to obtain the first quantity of the first set of information marking pairs. This first quantity reflects the degree of consistency of the two materials in multiple characteristics. After that, use the same method to obtain the positions with inconsistent codes, and add these information marking pairs with inconsistent markings to the set of information marking pairs with inconsistent markings, thereby generating the second quantity. The second quantity reveals the characteristics in which the two materials are different.

[0032] According to the predetermined feature comparison strategy, the ratio of the first quantity to the sum of the first quantity and the second quantity is denoted as the material feature similarity index.

[0033] Optionally, in order to quantitatively evaluate the similarity degree of the features between the first material and the second material, the system terminal calculates the material feature similarity index according to the calculation method recorded in the predetermined feature comparison strategy, that is, calculates the ratio of the first quantity to the sum of the first quantity and the second quantity, and records the result of this ratio calculation as the material feature similarity index.

[0034] Obtain the first key feature of the first sediment in the target sediment data, and selectively activate the multi-modal sensing device according to the first key feature to obtain the first key sensing information.

[0035] In one embodiment, when processing the target sediment data, the system terminal calculates the material quantity ratio by statistically analyzing the similarities and differences between the first sediment features of the first sediment and the other sediments. If this material quantity ratio meets the requirements of the defined ratio threshold, the first sediment is added to the first key feature. If not, another one is extracted from the first sediment and the same operation is performed. This process will continue until all sediment features of the first sediment have been analyzed for similarities and differences. Subsequently, in order to obtain detailed information related to this first key feature, the system terminal performs selective activation according to this first key feature. The selective activation is based on the functional characteristics of each sensor in the multi-modal sensing device. Each key feature corresponds to one or more sensors that can monitor it. For example, if the key feature is the humidity of the vacuum furnace, the system terminal will activate the sensors related to humidity; if it is the particle size distribution of the sediment, the sensors related to vision will be activated. By selectively activating the sensors corresponding to the first key feature, data closely related to this feature can be collected specifically, that is, the first key sensing information. This method not only improves the efficiency of data collection, but also ensures the accuracy and relevance of the data, providing strong support for subsequent analysis.

[0036] Furthermore, the present application provides obtaining the first key feature, including: Arbitrarily obtain the first sediment feature of the first sediment; judge whether the third material in the sediment database has the first sediment feature; if it has, add the third material to the first list, if not, add the third material to the second list.

[0037] Preferably, when processing sediment data, the system terminal arbitrarily obtains a feature of the first sediment as the first sediment feature. Subsequently, it checks whether the third material in the sediment database has this first sediment feature, that is, whether the sediment of the third material exhibits characteristics similar to those of the first sediment. For example, if the color feature of the first sediment is black carbon sediment, then the system terminal will check whether the sediment of the third material also presents the key feature of being black. According to the inspection results, corresponding list addition operations are performed. If the sediment of the third material has the characteristics of the first sediment, then this feature of the first sediment will be added to the first list, indicating that this feature of the first sediment has a certain similarity or relevance. If the sediment of the third material does not have the characteristics of the first sediment, then it will be added to the second list, indicating that this feature of the first sediment is different and unique to the first sediment. In this way, the sediment characteristics of the first material can be quickly and effectively classified, providing a data basis for subsequent key feature judgment.

[0038] Obtain the material quantity ratio of the first list to the second list; when the material quantity ratio is within the defined ratio threshold, add the first sediment feature to the first key feature of the first sediment.

[0039] Preferably, after the feature classification of the first sediment is completed, the system terminal calculates the ratio of the quantity of the first list to the data volume of the second list to generate the material quantity ratio. If the calculated material quantity ratio reaches a predefined defined ratio threshold, this represents that these features of the first sediment are unique. Therefore, the system terminal adds the first sediment feature to the first key feature of the first sediment.

[0040] Furthermore, the present application provides a multimodal sensing device, including: The multimodal sensing device at least includes a visual sensor, a humidity sensor, a mass sensor, and a position sensor.

[0041] Preferably, the multi-modal sensing device is a device integrating multiple different types of sensors, which is used to capture and analyze various types of information of the sediment and the internal environment of the vacuum furnace. The multi-modal sensing device at least includes a visual sensor, a humidity sensor, a mass sensor, and a position sensor. Among them, the visual sensor: can capture the visual images of the inside of the vacuum furnace and the sediment, and help the system terminal understand information such as the state, color, and shape of the sediment. The humidity sensor is responsible for monitoring the humidity level inside the vacuum furnace, which is crucial for analyzing the environmental conditions for sediment formation and possible chemical reaction processes. The mass sensor can measure the mass of the sediment in real time, which helps to monitor changes in the deposition process, such as deposition rate, deposition amount, etc. The position sensor is used to determine the specific position of the sediment in the vacuum furnace and provide detailed information on the sediment distribution to the system terminal. Through the combined use of these sensors, the multi-modal sensing device can provide rich data support for the system terminal, which helps to understand various characteristics and changes of the sediment and the internal environment of the vacuum furnace.

[0042] Use the first key sensing information as the input information of the in-furnace deposition discriminator to obtain the output information, and the in-furnace deposition discriminator is an intelligent model trained based on the principle of support vector machine.

[0043] In one embodiment, to evaluate the sediment condition inside the vacuum furnace, the system terminal normalizes the first key perception information, that is, extracts the maximum and minimum values of the corresponding features in the first key perception information, calculates the difference between each feature value and the corresponding feature minimum value, and then calculates the ratio of the calculated difference to the difference between the maximum and minimum values of the corresponding feature to obtain the normalized feature value of each feature value. Repeating this process, the system terminal can obtain the normalized first key perception information. Subsequently, according to expert advice, a weight is assigned to each feature, and the feature values in the normalized first key perception information are weighted with the corresponding weights to form the weighted first key perception information. Then, this weighted first key perception information is input into the in-furnace deposition discriminator, and the in-furnace deposition discriminator calculates based on the learned knowledge for the input weighted first key perception information to obtain the output information, and this output information includes the first deposition severity index. Among them, the in-furnace deposition discriminator is an intelligent model trained based on the support vector machine principle. Specifically, the system terminal first extracts historical perception information from the multimodal perception device, and according to the time identifier of this historical perception information, filters out the corresponding historical deposition severity index from the experimental database. Subsequently, using the same method as above, the historical perception information is normalized and weighted, and then the weighted historical perception information and the corresponding historical deposition severity index are divided into data to form a training set and a test set. After that, the weighted historical perception information in the training set is used as the input data, and the corresponding historical deposition severity index is used as the target output. Then, the support vector machine algorithm is used to train the input weighted historical perception information and the historical deposition severity index. During the training process, the support vector machine algorithm will learn how to distinguish different categories of data and find the optimal hyperplane. By adjusting the parameters of the support vector machine, such as the penalty coefficient C, the type of kernel function, etc., the performance of the in-furnace deposition discriminator is gradually optimized. After the training is completed, the system terminal uses the test set to evaluate the adjusted discriminator, calculates evaluation metrics such as accuracy, recall rate, F1 score, etc. Then, the calculated evaluation metrics are compared with the preset metrics to determine whether the in-furnace deposition discriminator meets the requirements. If not, reselect hyperparameters such as the penalty coefficient and the kernel function, and use the training data to train again. Otherwise, the current in-furnace deposition discriminator is output.

[0044] When the output information meets the discrimination constraint, a warning instruction is issued, and based on the warning instruction, a warning of the existence of the first sediment is given to the vacuum furnace.

[0045] In one embodiment, after the system terminal obtains the output information, the system terminal compares the first deposition severity index in the output information with the discrimination constraint. If the first deposition severity index meets the requirements of the discrimination constraint, it represents that the deposition in the vacuum furnace is severe. At this time, the system terminal will issue a warning instruction and trigger the corresponding warning mechanism according to this warning instruction. This includes displaying a warning message on the control interface, emitting a sound, sending a notification to the operator, etc. In this way, the operator can understand the deposition situation in the furnace and can quickly take actions, such as suspending the operation of the furnace, cleaning the deposits, etc., to avoid damage or production interruption caused.

[0046] Further, the present application provides issuing the warning instruction, including: When the first deposition severity index does not reach the severity index limit value, issue a key monitoring instruction, and perform real-time perception monitoring on the first deposit based on the key monitoring instruction.

[0047] Preferably, when the first deposition severity index does not reach the preset severity index limit value, the system terminal will judge that although the deposition degree of this deposit has not reached the level of emergency treatment, it has attracted enough attention. Therefore, a key monitoring instruction will be issued. The purpose of this key monitoring instruction is to enable the system terminal to perform more detailed and frequent perception monitoring on the first deposit, so as to timely detect the change trend of the deposit and whether it may reach the severity index limit value at a certain time in the future.

[0048] In summary, the embodiments of the present application at least have the following technical effects: The embodiments of the present application first obtain the historical processing material record of the vacuum furnace after predetermined maintenance and inspection processing, and extract the historical processing record of the first material. After reading the predetermined characteristic dimension, multi-dimensional characteristic collection is performed on the first material to obtain the first material characteristic information. Based on the first material characteristic information, the deposition database is traversed to obtain the target deposit data of the target material. The first key characteristic of the first deposit in the target deposit data is obtained, and the multi-modal perception device is selectively activated according to the first key characteristic to obtain the first key perception information. The first key perception information is used as the input information of the in-furnace deposition discriminator, and the output information is obtained through an intelligent model trained by the support vector machine principle. When the output information meets the discrimination constraint, a warning instruction is issued, and the vacuum furnace is warned based on the warning instruction. These technical effects together solve the technical problem that the traditional warning technology cannot comprehensively monitor the formation situation of deposits in the vacuum furnace, resulting in inaccurate and untimely deposit warnings, and achieve the effect of improving the accuracy and timeliness of deposit warnings through multi-modal perception.

[0049] It should be noted that the above order of the embodiments of the present application is only for description and does not represent the superiority or inferiority of the embodiments. And the above specific embodiments of this specification have been described. The processes depicted in the drawings do not necessarily require the specific order and continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0050] The above are only the preferred embodiments of the present application and are not intended to limit the present application. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present application shall be included within the protection scope of the present application.

[0051] This specification and the drawings are only exemplary descriptions of the present application and are considered to have covered any and all modifications, variations, combinations, or equivalents within the scope of the present application. Obviously, those skilled in the art can make various changes and modifications to the present application without departing from the scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the present application and its equivalent technologies, the present application is intended to include these changes and modifications.

Claims

1. A method for early warning of deposits in a vacuum furnace based on multimodal sensing, characterized in that: include: Obtaining a historical material processing record, wherein the historical material processing record refers to a material processing record of the vacuum furnace after a predetermined maintenance and inspection process; Extracting a first historical record from the historical material processing record, where the first historical record refers to a historical processing record of a first material; Reading a predetermined feature dimension, and collecting multi-dimensional features of the first material based on the predetermined feature dimension to obtain first material feature information; Traversing a deposition database based on the first material characteristic information to obtain target deposition data of a target material; Acquire a first key feature of a first sediment in the target sediment data, and selectively activate a multimodal sensing device according to the first key feature to obtain first key sensing information; The first key perception information is used as input information of a furnace deposition discriminator to obtain output information, wherein the furnace deposition discriminator is an intelligent model trained based on the support vector machine principle; When the output information meets the discrimination constraint, a warning instruction is issued, and based on the warning instruction, a warning is issued to the vacuum furnace that the first deposit exists.

2. According to claim 1, a vacuum furnace deposit early warning method based on multimodal sensing is characterized in that: The predetermined characteristic dimensions include furnace process dimensions, material property dimensions and furnace atmosphere dimensions.

3. According to claim 1, a vacuum furnace deposit early warning method based on multimodal sensing is characterized in that: include: Extracting a first deposition data group from the deposition database, where the first deposition data group refers to deposition data of a second material processed by the vacuum furnace; Based on the predetermined characteristic dimension, multi-dimensional characteristics of the second material are collected to obtain characteristic information of the second material; Reading a predetermined feature comparison strategy, and performing a comparative analysis of the first material feature information and the second material feature information based on the predetermined feature comparison strategy to obtain a material feature similarity index; When the material characteristic similarity index reaches the similarity index limit, the second material is used as the target material, and the first deposition data group is used as the target deposition data.

4. According to claim 3, a vacuum furnace deposit early warning method based on multimodal sensing is characterized in that: include: The predetermined feature comparison strategy includes a predetermined feature marking scheme; Based on the predetermined feature marking scheme, the first material feature information and the second material feature information are marked in sequence to obtain a first material marking vector and a second material marking vector respectively; Forming a first information tag pair set having the same tags as the first material tag vector and the second material tag vector, and counting a first number of the first information tag pair sets; forming a second information tag pair set in which the tags of the first material tag vector and the second material tag vector are inconsistent, and counting a second number of the second information tag pair set; According to the predetermined feature comparison strategy, the ratio of the first quantity to the sum of the first quantity and the second quantity is recorded as the material feature similarity index.

5. The method for early warning of deposits in a vacuum furnace based on multimodal sensing according to claim 1, characterized in that: include: arbitrarily obtaining a first sediment characteristic of the first sediment; Determining whether a third material in the deposition database has the first deposition feature; If available, add the third material to the first list; if not available, add the third material to the second list; Obtaining a material quantity ratio between the first list and the second list; When the material quantity ratio is at a defined ratio threshold, the first deposit characteristic is added to the first key characteristic of the first deposit.

6. The method for early warning of deposits in a vacuum furnace based on multimodal sensing according to claim 1 is characterized in that: The multimodal sensing device includes at least a visual sensor, a humidity sensor, a quality sensor and a position sensor.

7. The method for early warning of deposits in a vacuum furnace based on multimodal sensing according to claim 1, characterized in that: Also includes: The first key sensing information is weighted and normalized to obtain a first deposition severity index, where the first deposition severity index is used to characterize the deposition severity of the first deposit in the vacuum furnace; When the first deposition severity index reaches a severity index limit, the warning instruction is issued.

8. The method for early warning of deposits in a vacuum furnace based on multimodal sensing according to claim 7 is characterized in that: When the first deposition severity index does not reach the severity index limit, a key monitoring instruction is issued, and real-time perception monitoring of the first deposition is performed based on the key monitoring instruction.

Images