Method, device, equipment and system for monitoring converter steelmaking process

The converter flame spectral images are processed through optical artificial neural network chips and pre-trained artificial neural network models, which solves the instability and safety of steel quality caused by artificial experience, and achieves accurate and safe monitoring of the converter steelmaking process.

CN116083676BActive Publication Date: 2024-11-26TSINGHUA UNIVERSITY
View PDF 8 Cites 0 Cited by

Patent Information

Application Number
CN202111314402.5
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2021-11-08
Publication Date
2024-11-26
Estimated Expiration
2041-11-08

AI Technical Summary

Technical Problem

In the prior art, the quality of steel is uneven due to different manual experience during the converter steelmaking process, and it is difficult for manual experience to accurately react in the reactor, resulting in the safety of the steelmaking process being unable to be guaranteed.

Method used

Optical artificial neural network chip and pre-trained artificial neural network model are used to receive and process the spectral image information of the converter furnace entrance flame to realize abnormal prediction and monitoring of reactions in the converter, and generate monitoring information to guide adjustments.

Benefits of technology

It improves the quality of steel and the safety of the steelmaking process, accelerates processing speed through hardware calculations, meets real-time monitoring needs, and achieves accurate, safe and reliable monitoring of the converter steelmaking process.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN116083676B_ABST
    Figure CN116083676B_ABST
Patent Text Reader

Abstract

The embodiment of the present invention provides a method, device, equipment and system for monitoring the converter steelmaking process, wherein the method includes: receiving transmission information from an optical artificial neural network chip during the converter production process, the optical artificial neural network chip is used to implement a calculation strategy of a part of a pre-trained artificial neural network model, the transmission information is obtained by the optical artificial neural network chip based on the calculation strategy, and the spectrum image of the furnace mouth flame of the converter is collected, and the spectrum image contains spectrum information; the transmission information received at the current moment is input into another part of the artificial neural network model to obtain an abnormal prediction result of the reaction in the converter; and monitoring information corresponding to the abnormal prediction result is generated. In this way, accurate, safe and reliable monitoring of the converter steelmaking process is achieved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of smelting technology, and in particular to a converter steelmaking process monitoring method, device, equipment and system. Background Art

[0002] Converter steelmaking is a widely used steelmaking method. During steelmaking, molten iron and raw materials such as scrap steel are first added to the converter, and then oxygen is blown and smelted to form molten steel, that is, blowing. The different components and quantities of the raw materials added to the converter, and the different degrees of reaction during the blowing process, will make the steelmaking process complex and diverse. Accurately knowing the reaction conditions during the steelmaking process in the converter is important for producing high-quality steel and ensuring the safety of the steelmaking process. For example, the inability to accurately know the reaction conditions in the converter makes it difficult for workers to determine the appropriate amount of oxygen blowing. If the oxygen blowing amount is inappropriate, it may affect the quality of the steel, and may cause abnormal conditions such as molten steel splashing, causing safety accidents, etc. Among them, if the oxygen blowing amount is too low, it may cause insufficient reaction. If the oxygen blowing amount is too high, it may cause overoxidation or molten steel splashing. For another example, if the equipment in the furnace is damaged, causing abnormal reactions in the converter, molten steel splashing may also occur. In severe cases, it may even cause the converter to explode, causing safety problems. Therefore, it is necessary to accurately monitor the reaction conditions in the converter during the converter steelmaking process in order to make timely adjustments.

[0003] At present, the reaction conditions in the converter are generally judged and adjusted based on manual experience. However, on the one hand, different workers have different experiences, which is difficult to standardize, and the quality of the produced steel varies. On the other hand, manual experience is difficult to accurately reflect the reaction conditions in the furnace, resulting in the inability to ensure the safety of the steelmaking process. Summary of the invention

[0004] The embodiments of the present invention provide a method, device, equipment and system for monitoring the converter steelmaking process, so as to solve the problems of the existing method of judging the reaction conditions in the converter by manual experience, the uneven quality of steel due to different manual experience and the inability to ensure safety due to inaccurate manual experience.

[0005] In a first aspect, an embodiment of the present invention provides a method for monitoring a converter steelmaking process, comprising:

[0006] In the converter production process, receiving transmission information from an optical artificial neural network chip, the optical artificial neural network chip is used to implement a calculation strategy of a part of a pre-trained artificial neural network model, the transmission information is obtained by the optical artificial neural network chip processing a collected spectral image of the furnace mouth flame of the converter based on the calculation strategy, and the spectral image contains spectral information;

[0007] Inputting the transmission information received at the current moment into another part of the artificial neural network model to obtain an abnormal prediction result of the reaction in the converter;

[0008] Generate monitoring information corresponding to the abnormal prediction result.

[0009] Optionally, the part of the artificial neural network model includes an input layer and a first fully connected layer;

[0010] The optical artificial neural network chip includes a light modulation layer and an image sensor;

[0011] Wherein, the light modulation layer is used to modulate the incident light from the furnace mouth flame based on the calculation strategy of the input layer;

[0012] The image sensor is used to sense the light signal modulated by the light modulation layer based on the computing strategy of the first fully connected layer, and process the sensed light signal into the transmission information in the form of an electrical signal.

[0013] Optionally, also include:

[0014] Inputting the spectral image samples of the furnace mouth flame collected in advance by using the spectral imaging chip into the preset artificial neural network model for training to obtain the artificial neural network model, wherein the spectral image samples are marked with the actual results of the reaction in the converter;

[0015] Performing optical simulation on the light modulation layer, and adjusting the parameters of the structure of the simulated light modulation layer so that the transmittance of the simulated light modulation layer to different wavelength components in the incident light is a target transmittance;

[0016] Wherein, the target transmittance is determined based on a connection weight from the input layer to the first fully connected layer;

[0017] The structural parameters of the simulated light modulation layer are used to manufacture the light modulation layer on the image sensor to obtain the optical artificial neural network chip.

[0018] Optionally, another part of the artificial neural network model includes a second fully connected layer and a nonlinear activation layer;

[0019] The step of inputting the transmission information received at the current moment into another part of the artificial neural network model to obtain an abnormal prediction result of the reaction in the converter includes:

[0020] Inputting the transmission information received at the current moment into the second fully connected layer to obtain a fully connected result;

[0021] The fully connected result is input into the nonlinear activation layer to obtain the abnormal prediction result.

[0022] Optionally, the abnormal prediction result includes the abnormal prediction result corresponding to the current moment and / or the abnormal prediction result corresponding to a preset time period after the current moment.

[0023] Optionally, the abnormality prediction result includes a normal response, or an abnormality of a preset type occurs.

[0024] Optionally, the generating monitoring information corresponding to the abnormal prediction result includes:

[0025] In response to the abnormality prediction result including the occurrence of an abnormality of a preset type, early warning information corresponding to the abnormality of the preset type is generated.

[0026] In a second aspect, an embodiment of the present invention provides a converter steelmaking process monitoring device, comprising:

[0027] An information receiving module is used to receive transmission information from an optical artificial neural network chip during the converter production process, wherein the optical artificial neural network chip is used to implement a calculation strategy of a part of a pre-trained artificial neural network model, wherein the transmission information is obtained by the optical artificial neural network chip processing a collected spectral image of the furnace mouth flame of the converter based on the calculation strategy, wherein the spectral image contains spectral information;

[0028] An abnormality prediction module, used for inputting the transmission information received at the current moment into another part of the artificial neural network model to obtain an abnormality prediction result of the reaction in the converter;

[0029] The information generation module is used to generate monitoring information corresponding to the abnormal prediction result.

[0030] In a third aspect, an embodiment of the present invention provides an electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the steps of the converter steelmaking process monitoring method provided in the first aspect are implemented.

[0031] In a fourth aspect, an embodiment of the present invention provides a converter steelmaking process monitoring system, comprising:

[0032] Optical artificial neural network chip and electronic device as provided in the third aspect.

[0033] In a fifth aspect, an embodiment of the present invention provides a non-transitory computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the steps of the method provided in the first aspect.

[0034] A converter steelmaking process monitoring method provided by an embodiment of the present invention receives transmission information from an optical artificial neural network chip during the converter production process. The optical artificial neural network chip is used to implement a calculation strategy of a part of a pre-trained artificial neural network model. The transmission information is obtained by the optical artificial neural network chip based on the calculation strategy by processing the spectrum image of the furnace mouth flame of the converter. The spectrum image contains spectrum information. The transmission information received at the current moment is input into another part of the artificial neural network model to obtain the abnormal prediction result of the reaction in the converter, and generate monitoring information corresponding to the abnormal prediction result, so as to automatically realize the accurate prediction of the abnormal situation in the converter, so as to make timely adjustments, and avoid the uneven quality of steel caused by the difference in the existing artificial experience and the problem that the safety cannot be guaranteed due to the inaccurate artificial experience, thereby improving the quality of steel and improving safety. In addition, the calculation strategy of a part of the artificial neural network model is transferred to the optical artificial neural network chip for hardware calculation, and the processing speed is greatly improved, which can meet the real-time requirements for monitoring the converter steelmaking process. The calculation strategy of another part of the artificial neural network model is calculated by software, and the abnormal prediction result is more accurate. In this way, accurate, safe and reliable monitoring of the converter steelmaking process is achieved. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0036] Figure 1 A schematic flow chart of a converter steelmaking process monitoring method provided by an embodiment of the present invention;

[0037] Figure 2 A schematic diagram of the structure of an optical artificial neural network chip provided by an embodiment of the present invention;

[0038] Figure 3 A schematic diagram of an application scenario of a converter steelmaking process monitoring system provided by an embodiment of the present invention;

[0039] Figure 4 A schematic diagram of the structure of a converter steelmaking process monitoring device provided by an embodiment of the present invention;

[0040] Figure 5 A schematic diagram of the structure of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0041] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0042] Figure 1 A schematic diagram of a converter steelmaking process monitoring method provided by an embodiment of the present invention is shown in FIG. Figure 1 As shown, the method includes:

[0043] Step 110, during the converter production process, receiving transmission information from the optical artificial neural network chip, the optical artificial neural network chip is used to implement a calculation strategy of a part of a pre-trained artificial neural network model, the transmission information is obtained by the optical artificial neural network chip based on the calculation strategy, by processing the collected spectral image of the furnace mouth flame of the converter, and the spectral image contains spectral information.

[0044] Among them, the optical artificial neural network chip, that is, the optical neural network chip, can use optical components to realize the calculation of the neural network. The processing speed of the calculation based on the hardware of the optical component is much faster than the processing speed of the computer software. In this step, the calculation strategy of a part of the pre-trained artificial neural network model is set in the optical artificial neural network chip, and the processing speed is greatly improved, which can meet the real-time requirements for monitoring the converter steelmaking process.

[0045] In addition to spatial information (i.e., coordinate information of the two-dimensional plane), the pixel information of each pixel of the spectral image also contains spectral information, which includes the intensity information of each wavelength component. The spectrum is the fingerprint of the substance, so the spectral information of the flame at the mouth of the converter can reflect the substance in the converter. Since the substances in different reaction conditions in the converter are different, the spectral information of the flame at the mouth of the converter can further reflect the reaction conditions in the converter.

[0046] The converter steelmaking process monitoring method provided in this embodiment can be executed by electronic devices such as computers or a combination of software and / or hardware therein.

[0047] In actual applications, when the converter starts production, the optical artificial neural network chip can start to process the collected spectral image of the converter's furnace mouth flame in real time to obtain transmission information, and transmit it to the electronic device, and the electronic device can receive the above transmission information from the optical artificial neural network chip.

[0048] Step 120, input the transmission information received at the current moment into another part of the artificial neural network model to obtain the abnormal prediction result of the reaction in the converter.

[0049] The transmission information received at the current moment can reflect the current reaction situation in the converter, the trend of the reaction, and whether the reaction will be abnormal. Therefore, based on the above transmission information received at the current moment, the abnormal prediction result corresponding to the current moment can be predicted, and the abnormal prediction result corresponding to the preset time period after the current moment can also be predicted. Exemplarily, the preset time period after the current moment can include the next moment of the current moment. Based on this, the abnormal prediction result output by the artificial neural network model in this step includes the abnormal prediction result corresponding to the current moment, and / or the abnormal prediction result corresponding to the preset time period after the current moment.

[0050] Step 130: Generate monitoring information corresponding to the abnormal prediction result.

[0051] In the present embodiment, on the one hand, since the substances in different reaction conditions in the converter are different, and the spectral information contained in the spectral image of the furnace mouth flame can reflect the substances in the converter, and then can reflect the reaction conditions in the converter, therefore, based on the spectral image of the furnace mouth flame, an abnormality prediction can be performed through a pre-trained artificial neural network model to generate monitoring information corresponding to the abnormal prediction result, so that the staff can understand whether there will be abnormal conditions in the reaction in the converter through the monitoring information, thus automatically and accurately predicting the abnormal conditions in the converter so as to make timely adjustments, thereby avoiding the uneven quality of steel caused by the differences in the existing manual experience and the problem of inability to guarantee safety caused by the inaccurate manual experience, thereby improving the quality of steel and improving safety. On the other hand, the calculation strategy of a part of the pre-trained artificial neural network model is implemented by the optical artificial neural network chip, and the optical artificial neural network chip processes the collected spectral image of the furnace mouth flame of the converter based on the calculation strategy to obtain transmission information, that is, the calculation strategy of a part of the artificial neural network model is transferred to the optical artificial neural network chip for hardware calculation, and the processing speed is greatly improved, which can meet the real-time requirements of the converter steelmaking process monitoring. After receiving the transmission information, the transmission information received at the current moment is continuously input into another part of the artificial neural network model, so as to obtain the abnormal prediction result of the reaction in the converter, that is, the calculation strategy of another part of the artificial neural network model is calculated by software, and the abnormal prediction result is more accurate. In this way, accurate, safe and reliable converter steelmaking process monitoring is achieved.

[0052] In practical applications, after the monitoring information corresponding to the abnormal prediction result is generated, the monitoring information can be displayed for viewing by the staff. Specifically, the monitoring information can be displayed on the screen of the electronic device.

[0053] Exemplarily, the abnormal prediction result may include a normal reaction, or an abnormality of a preset type occurs. In practical applications, the abnormality of the preset type may be set according to actual needs. Exemplarily, the abnormality of the preset type may be molten steel splashing, etc. If the abnormal prediction result includes a normal reaction, no adjustment may be made. If the abnormal prediction result includes an abnormality of a preset type occurs, an adjustment may be made in a timely manner. Exemplarily, if molten steel splashing occurs, the oxygen blowing amount may be adjusted.

[0054] Based on the above embodiment, monitoring information corresponding to the abnormality prediction result is generated, and its specific implementation method may include: in response to the abnormality prediction result including the occurrence of a preset type of abnormality, generating warning information corresponding to the preset type of abnormality.

[0055] For example, when the abnormal prediction results include molten steel splashing, a warning message can be generated "Molten steel splashing may occur soon, please pay attention!" In this way, the staff can understand the possibility of molten steel splashing through the displayed warning information and make timely adjustments.

[0056] If the abnormal prediction result includes a normal reaction, correspondingly, monitoring information corresponding to the abnormal prediction result is generated, and the specific implementation method thereof may include: in response to the abnormal prediction result including a normal reaction, generating a prompt message for prompting that the reaction in the converter is normal. Exemplarily, the prompt message may be "The reaction in the converter is normal!" At this time, the staff member knows that the reaction in the converter is normal through the displayed prompt message, and no adjustment is required.

[0057] If the abnormal prediction result output by the artificial neural network model includes the abnormal prediction result corresponding to the current moment, and / or the abnormal prediction result corresponding to the preset time period after the current moment. Then, when generating monitoring information corresponding to the abnormal prediction result, it can specifically include: generating monitoring information corresponding to the abnormal prediction result corresponding to the current moment, and / or generating monitoring information corresponding to the abnormal prediction result corresponding to the preset time period after the current moment.

[0058] Among them, when generating monitoring information corresponding to the abnormal prediction result corresponding to the current moment, it can specifically include: in response to the abnormal prediction result corresponding to the current moment including the occurrence of a preset type of abnormality, generating early warning information corresponding to the preset type of abnormality for the current moment; in response to the abnormal prediction result corresponding to the current moment including a normal reaction, generating prompt information for prompting that the reaction in the converter is normal for the current moment.

[0059] When generating monitoring information corresponding to the abnormal prediction result corresponding to the preset time period after the current moment, it can specifically include: in response to the abnormal prediction result corresponding to the preset time period after the current moment including the occurrence of a preset type of abnormality, generating early warning information corresponding to the preset type of abnormality for the preset time period after the current moment; in response to the abnormal prediction result corresponding to the preset time period after the current moment including a normal reaction, generating prompt information for prompting that the reaction in the converter is normal for the preset time period after the current moment.

[0060] In this way, the staff can understand the reaction conditions in the converter at the current moment and in a preset time period after the current moment more comprehensively.

[0061] Based on the above embodiment, a part of the artificial neural network model may include an input layer and a first fully connected layer. Figure 2 As shown, the optical artificial neural network chip 210 may include an optical modulation layer 211 and an image sensor 212. The optical modulation layer 211 is used to modulate the incident light from the furnace flame based on the calculation strategy of the input layer. The image sensor 212 is used to sense the optical signal modulated by the optical modulation layer 211 based on the calculation strategy of the first fully connected layer, and process the sensed optical signal into transmission information in the form of an electrical signal.

[0062] The light modulation layer 211 has different modulation effects on light of different wavelengths, so that the transmittance of light of different wavelengths is different. The light modulation layer 211 may include a plurality of modulation units 2111, and the modulation unit 2111 may include a modulation hole array, and the shape of the modulation hole may be, but is not limited to, a circle, a cross, a triangle, a star, or a rectangle, etc.

[0063] The image sensor 212 has a photosensitive area, and the light modulation layer 211 is disposed on the photosensitive area of ​​the image sensor 212. The image sensor 212 can sense the light signal modulated by the light modulation layer 211 through the photosensitive area, and processes the sensed light signal to obtain transmission information in the form of an electrical signal, and transmits the transmission information to the electronic device in the form of an electrical signal.

[0064] Since the optical artificial neural network chip can process optical signals, the input layer of the artificial neural network model can be directly set on the optical artificial neural network chip to process the incident light from the furnace mouth flame, that is, start processing the collected spectral image of the furnace mouth flame. In addition, the first fully connected layer of the artificial neural network model can be set on the optical artificial neural network chip, and dimensionality reduction processing can be achieved through preliminary full connection. In this way, through the preprocessing of the optical artificial neural network chip, the calculation of subsequent software is reduced, thereby improving the processing speed.

[0065] Based on any of the above embodiments, the above converter steelmaking process monitoring method may further include:

[0066] Firstly, the spectral image samples of the furnace mouth flame collected in advance by the spectral imaging chip are input into the preset artificial neural network model for training to obtain the artificial neural network model, in which the actual results of the reaction in the converter are marked in the spectral image samples.

[0067] Then, an optical simulation is performed on the light modulation layer, and the parameters of the structure of the simulated light modulation layer are adjusted so that the transmittance of the simulated light modulation layer to different wavelength components in the incident light is the target transmittance. The target transmittance is determined based on the connection weight from the input layer to the first fully connected layer. The parameters of the structure of the simulated light modulation layer are used to make a light modulation layer on an image sensor to obtain an optical artificial neural network chip.

[0068] Among them, the spectral imaging chip can collect incident light and output the original spectral image. In practical applications, the existing spectral imaging chip can be directly used to collect spectral image samples of the furnace mouth flame, and then the actual results of the reaction in the converter corresponding to each spectral image sample are collected, and the actual results of the reaction in the converter are marked in the spectral image sample, for example, including the actual results corresponding to the current moment, and the actual results corresponding to the preset time period after the current moment (for example, including the next moment of the current moment), and the actual results may include normal reaction or a preset type of abnormality.

[0069] An artificial neural network model can be built in advance, that is, a preset artificial neural network model is obtained, and the spectral image samples at each moment in the converter production process are sequentially input into the preset artificial neural network model for training until convergence, and a trained artificial neural network model is obtained. Exemplarily, the artificial neural network model may include an input layer, a first fully connected layer, a second fully connected layer, and a nonlinear activation layer. The calculation strategy of the input layer and the first fully connected layer of the artificial neural network model is transferred to the optical artificial neural network chip, wherein the optical modulation layer of the optical artificial neural network chip corresponds to the input layer, and the image sensor corresponds to the first fully connected layer. After the image sensor is selected, the structure of the optical modulation layer can be optically simulated by computer simulation technology, and the parameters of the structure of the simulated optical modulation layer can be adjusted so that the transmittance of the optical modulation layer to different wavelength components in the incident light is the target transmittance, and the target transmittance can be the connection weight from the input layer to the first fully connected layer. Finally, the parameters of the structure of the adjusted simulated optical modulation layer are used to make the optical modulation layer on the selected image sensor.

[0070] In this embodiment, after the training of the artificial neural network model is completed, the structural parameters of the light modulation layer are obtained by optical simulation, so that the optical artificial neural network chip is obtained more quickly and accurately.

[0071] In practice, a light modulation layer can be fabricated on a wafer-level image sensor using a complementary metal oxide semiconductor (CMOS) tape-out process, thereby realizing a monolithic integrated optical artificial neural network chip. The CMOS tape-out process is a very mature process that greatly reduces the size, power consumption and cost of the chip, can be applied on a large scale, and the quality of the manufactured optical artificial neural network chip is good, thereby ensuring the processing speed and accuracy of the optical artificial neural network chip.

[0072] Based on any of the above embodiments, another part of the artificial neural network model may include a second fully connected layer and a nonlinear activation layer. Accordingly, the transmission information received at the current moment is input into another part of the artificial neural network model to obtain an abnormal prediction result of the reaction in the converter, and its specific implementation method may include: inputting the transmission information received at the current moment into the second fully connected layer to obtain a fully connected result; inputting the fully connected result into the nonlinear activation layer to obtain an abnormal prediction result.

[0073] Here, the second fully connected layer further achieves dimensionality reduction. The nonlinear activation layer can map the fully connected result of the second fully connected layer to the preset abnormal prediction result. In this embodiment, after the transmission information is fully connected and nonlinearly activated, the abnormal prediction result can be quickly obtained.

[0074] The following is a more detailed description of a converter steelmaking process monitoring method provided by an embodiment of the present invention through a specific application scenario.

[0075] The converter steelmaking process monitoring method of this embodiment is implemented based on the converter steelmaking process monitoring system. Figure 3 In the application scenario shown, the converter steelmaking process monitoring system includes an optical artificial neural network chip 210, a transmission line 220 and a monitoring computer 230.

[0076] Among them, the optical artificial neural network chip 210 includes an optical modulation layer 211 and an image sensor 212. The specific structure of the optical artificial neural network chip 210 can refer to the above related embodiments and will not be described here. The optical artificial neural network chip 210 can implement the calculation strategy of a part of the pre-trained artificial neural network model. Among them, the pre-trained artificial neural network model may include an input layer, a first fully connected layer, a second fully connected layer and a nonlinear activation layer. The optical artificial neural network chip 210 corresponds to the input layer and the first fully connected layer of the artificial neural network model.

[0077] A spectrum processing program and a real-time display program are set in the monitoring computer 230. The second fully connected layer and the nonlinear activation layer of the pre-trained artificial neural network model are set in the spectrum processing program.

[0078] The optical artificial neural network chip 210 is arranged at a safe distance on one side of the furnace body of the converter 240. The safe distance can be set according to actual needs. For example, the safe distance can be at least 20m. The surface of the light modulation layer of the optical artificial neural network chip 210 faces the furnace mouth flame 241 of the converter 240. Based on this, the incident light 242 from the furnace mouth flame 241 enters the optical artificial neural network chip 210, is modulated by the light modulation layer 211 of the optical artificial neural network chip 210, and is sensed by the image sensor 212. The modulated light signal is processed into transmission information in the form of an electrical signal, and the transmission information is transmitted to the monitoring computer 230 through the transmission line 220. The processor in the monitoring computer 230 performs full connection and nonlinear activation on the electrical signal by executing the spectrum processing program, obtains the abnormal prediction result of the reaction in the converter, and generates monitoring information corresponding to the abnormal prediction result, and displays the monitoring information by executing the real-time display program. The abnormal prediction result can be the occurrence of molten steel splashing, etc.

[0079] This embodiment uses an optical artificial neural network chip to obtain spectral images, and through fast and accurate calculations, it can realize real-time monitoring of the reaction in the converter during the converter steelmaking process, make predictions on abnormal conditions such as molten steel splashing, and guide the steelmaking operation of the staff, thereby improving the quality of steel and ensuring the safe and reliable operation of the converter. The optical artificial neural network chip used in this solution can be produced using the existing CMOS tape-out process, which greatly reduces the size, power consumption and cost of the chip and can be applied on a large scale.

[0080] The converter steelmaking process monitoring device provided by the present invention is described below. The converter steelmaking process monitoring device described below and the converter steelmaking process monitoring method described above can be referenced to each other.

[0081] Figure 4 A schematic diagram of the structure of a converter steelmaking process monitoring device provided by an embodiment of the present invention is shown in FIG. Figure 4 As shown, the device includes an information receiving module 410, an abnormality prediction module 420 and an information generating module 430;

[0082] The information receiving module 410 is used to receive transmission information from the optical artificial neural network chip during the converter production process. The optical artificial neural network chip is used to implement a calculation strategy of a part of the pre-trained artificial neural network model. The transmission information is obtained by the optical artificial neural network chip based on the calculation strategy by processing the collected spectral image of the furnace mouth flame of the converter.

[0083] The abnormality prediction module 420 is used to input the transmission information received at the current moment into another part of the artificial neural network model to obtain the abnormality prediction result of the reaction in the converter;

[0084] The information generation module 430 is used to generate monitoring information corresponding to the abnormal prediction result.

[0085] Based on any of the above embodiments, a portion of the artificial neural network model includes an input layer and a first fully connected layer;

[0086] The optical artificial neural network chip includes a light modulation layer and an image sensor;

[0087] The light modulation layer is used to modulate the incident light from the furnace flame based on the calculation strategy of the input layer;

[0088] The image sensor is used to sense the light signal modulated by the light modulation layer based on the computing strategy of the first fully connected layer, and process the sensed light signal into transmission information in the form of an electrical signal.

[0089] Based on any of the above embodiments, the converter steelmaking process monitoring device may further include:

[0090] A network training module is used to input the spectral image samples of the furnace mouth flame collected in advance by using the spectral imaging chip into a preset artificial neural network model for training to obtain an artificial neural network model, and the spectral image samples are annotated with actual results of the reaction in the converter;

[0091] An optical simulation module is used to perform optical simulation on the light modulation layer and adjust the parameters of the structure of the simulated light modulation layer so that the transmittance of the simulated light modulation layer to different wavelength components in the incident light is the target transmittance;

[0092] Wherein, the target transmittance is determined based on the connection weight from the input layer to the first fully connected layer;

[0093] The parameters of the structure of the simulated light modulation layer are used to make the light modulation layer on the image sensor to obtain an optical artificial neural network chip.

[0094] Based on any of the above embodiments, another part of the artificial neural network model includes a second fully connected layer and a nonlinear activation layer;

[0095] The abnormality prediction module is specifically used to input the transmission information received at the current moment into the second fully connected layer to obtain a fully connected result; and input the fully connected result into the nonlinear activation layer to obtain an abnormality prediction result.

[0096] Based on any of the above embodiments, the abnormality prediction result includes the abnormality prediction result corresponding to the current moment and / or the abnormality prediction result corresponding to a preset time period after the current moment.

[0097] Based on any of the above embodiments, the abnormality prediction result includes a normal response, or an abnormality of a preset type occurs.

[0098] Based on any of the above embodiments, the information generation module is specifically configured to generate warning information corresponding to the preset type of abnormality in response to the abnormality prediction result including the occurrence of a preset type of abnormality.

[0099] Figure 5 A schematic diagram of the structure of an electronic device provided by an embodiment of the present invention, such as Figure 5 As shown, the electronic device may include: a processor 510, a communication interface 520, a memory 530 and a communication bus 540, wherein the processor 510, the communication interface 520 and the memory 530 communicate with each other through the communication bus 540. The processor 510 may call the logic command in the memory 530 to execute the following method: in the converter production process, receiving transmission information from the optical artificial neural network chip, the optical artificial neural network chip is used to implement the calculation strategy of a part of the pre-trained artificial neural network model, the transmission information is obtained by the optical artificial neural network chip based on the calculation strategy, and the spectrum image of the furnace mouth flame of the converter is collected and processed, and the spectrum image contains spectrum information; the transmission information received at the current moment is input into another part of the artificial neural network model to obtain the abnormal prediction result of the reaction in the converter; and the monitoring information corresponding to the abnormal prediction result is generated.

[0100] In addition, the logic commands in the above-mentioned memory 530 can be implemented in the form of software functional units and can be stored in a computer-readable storage medium when sold or used as an independent product. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product, which is stored in a storage medium and includes several commands to enable a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods of various embodiments of the present invention. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0101] An embodiment of the present invention further provides a converter steelmaking process monitoring system, comprising: an optical artificial neural network chip and the electronic equipment provided by the above embodiments.

[0102] An embodiment of the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, it is implemented to execute the methods provided by the above embodiments, for example including: in the production process of a converter, receiving transmission information from an optical artificial neural network chip, the optical artificial neural network chip is used to implement a calculation strategy of a part of a pre-trained artificial neural network model, the transmission information is obtained by the optical artificial neural network chip based on the calculation strategy, by processing the collected spectral image of the furnace mouth flame of the converter, and the spectral image contains spectral information; the transmission information received at the current moment is input into another part of the artificial neural network model to obtain an abnormal prediction result of the reaction in the converter; and monitoring information corresponding to the abnormal prediction result is generated.

[0103] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.

[0104] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, which can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of commands to enable a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0105] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for monitoring a converter steelmaking process, characterized in that: include: In the converter production process, receiving transmission information from an optical artificial neural network chip, the optical artificial neural network chip is used to implement a calculation strategy of a part of a pre-trained artificial neural network model, the transmission information is obtained by the optical artificial neural network chip processing a collected spectral image of the furnace mouth flame of the converter based on the calculation strategy, the spectral image contains spectral information; the spectral information contains intensity information of each wavelength component; a part of the artificial neural network module is set in the optical artificial neural network chip; The transmission information received at the current moment is input into another part of the artificial neural network model to obtain an abnormal prediction result of the reaction in the converter; the abnormal prediction result includes an abnormal prediction result corresponding to the current moment and / or an abnormal prediction result corresponding to a preset time period after the current moment; the abnormal prediction result includes a normal reaction or an abnormality of a preset type; Generating monitoring information corresponding to the abnormal prediction result; The generating of monitoring information corresponding to the abnormal prediction result includes: In response to the abnormality prediction result including the occurrence of an abnormality of a preset type, generating warning information corresponding to the abnormality of the preset type; In response to the abnormal prediction result including a normal reaction, generating prompt information for prompting that the reaction in the converter is normal; Generate monitoring information corresponding to the abnormal prediction result corresponding to the current moment, and / or generate monitoring information corresponding to the abnormal prediction result corresponding to a preset time period after the current moment; A portion of the artificial neural network model includes an input layer and a first fully connected layer; The optical artificial neural network chip includes a light modulation layer and an image sensor; the light modulation layer corresponds to the input layer, and the image sensor corresponds to the first fully connected layer; the image sensor has a photosensitive area, and the light modulation layer is arranged on the photosensitive area; The light modulation layer is used to modulate the incident light from the furnace mouth flame based on the calculation strategy of the input layer; the light modulation layer has different modulation effects on light of different wavelengths, so that the transmittance of light of different wavelengths is different; the light modulation layer includes a plurality of modulation units; the modulation unit includes a modulation hole array, and the shape of the modulation hole includes a circle, a cross, a triangle, a star or a rectangle; The image sensor is used to sense the light signal modulated by the light modulation layer based on the computing strategy of the first fully connected layer, and process the sensed light signal into the transmission information in the form of an electrical signal; Another part of the artificial neural network model includes a second fully connected layer and a nonlinear activation layer; The step of inputting the transmission information received at the current moment into another part of the artificial neural network model to obtain an abnormal prediction result of the reaction in the converter includes: Inputting the transmission information received at the current moment into the second fully connected layer to obtain a fully connected result; The fully connected result is input into the nonlinear activation layer to obtain the abnormal prediction result.

2. The method for monitoring converter steelmaking process according to claim 1, characterized in that: Also includes: Inputting the spectral image samples of the furnace mouth flame collected in advance by using the spectral imaging chip into the preset artificial neural network model for training to obtain the artificial neural network model, wherein the spectral image samples are marked with the actual results of the reaction in the converter; Performing optical simulation on the light modulation layer, and adjusting the parameters of the structure of the simulated light modulation layer so that the transmittance of the simulated light modulation layer to different wavelength components in the incident light is a target transmittance; wherein the target transmittance is determined based on a connection weight from the input layer to the first fully connected layer; The structural parameters of the simulated light modulation layer are used to manufacture the light modulation layer on the image sensor to obtain the optical artificial neural network chip.

3. A converter steelmaking process monitoring device, characterized in that: include: An information receiving module is used to receive transmission information from an optical artificial neural network chip during the converter production process, wherein the optical artificial neural network chip is used to implement a calculation strategy of a part of a pre-trained artificial neural network model, wherein the transmission information is obtained by processing a spectrum image of a furnace mouth flame of the converter collected by the optical artificial neural network chip based on the calculation strategy, wherein the spectrum image contains spectrum information; the spectrum information contains intensity information of each wavelength component; a part of the artificial neural network module is arranged in the optical artificial neural network chip; an abnormality prediction module, used for inputting the transmission information received at the current moment into another part of the artificial neural network model to obtain an abnormality prediction result of the reaction in the converter; the abnormality prediction result includes an abnormality prediction result corresponding to the current moment and / or an abnormality prediction result corresponding to a preset time period after the current moment; the abnormality prediction result includes a normal reaction, or an abnormality of a preset type occurs; An information generation module, used to generate monitoring information corresponding to the abnormal prediction result; The generating of monitoring information corresponding to the abnormal prediction result includes: In response to the abnormality prediction result including the occurrence of an abnormality of a preset type, generating warning information corresponding to the abnormality of the preset type; In response to the abnormal prediction result including a normal reaction, generating prompt information for prompting that the reaction in the converter is normal; Generate monitoring information corresponding to the abnormal prediction result corresponding to the current moment, and / or generate monitoring information corresponding to the abnormal prediction result corresponding to a preset time period after the current moment; A portion of the artificial neural network model includes an input layer and a first fully connected layer; The optical artificial neural network chip includes a light modulation layer and an image sensor; the light modulation layer corresponds to the input layer, and the image sensor corresponds to the first fully connected layer; the image sensor has a photosensitive area, and the light modulation layer is arranged on the photosensitive area; The light modulation layer is used to modulate the incident light from the furnace mouth flame based on the calculation strategy of the input layer; the light modulation layer has different modulation effects on light of different wavelengths, so that the transmittance of light of different wavelengths is different; the light modulation layer includes a plurality of modulation units; the modulation unit includes a modulation hole array, and the shape of the modulation hole includes a circle, a cross, a triangle, a star or a rectangle; The image sensor is used to sense the light signal modulated by the light modulation layer based on the computing strategy of the first fully connected layer, and process the sensed light signal into the transmission information in the form of an electrical signal; Another part of the artificial neural network model includes a second fully connected layer and a nonlinear activation layer; The step of inputting the transmission information received at the current moment into another part of the artificial neural network model to obtain an abnormal prediction result of the reaction in the converter includes: Inputting the transmission information received at the current moment into the second fully connected layer to obtain a fully connected result; The fully connected result is input into the nonlinear activation layer to obtain the abnormal prediction result.

4. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the program, the steps of the converter steelmaking process monitoring method according to any one of claims 1 to 2 are implemented.

5. A converter steelmaking process monitoring system, characterized in that: include: An optical artificial neural network chip and an electronic device as claimed in claim 4.

Citation Information

Patent Citations

  • Converter steelmaking endpoint determination method and system, control method and control system

    CN102206727A

  • Termination point prediction method and system for converter steelmaking

    CN102392095A

  • Steelmaking end point prediction method based on neural networks

    CN105925750A

  • Converter splashing prediction method and system based on video intelligent algorithm

    CN110309973A

  • Converter smelting splashing monitoring method and system, storage medium and equipment

    CN111104856A