Method and device for artificial intelligence voice system to participate in control of operation process of casting piece furnace

Abstract

The invention discloses a method and a device for an artificial intelligence voice system to participate in control of an operation process of a casting piece furnace, relates to the technical field of vacuum induction casting piece furnaces, and aims at processing a voice instruction of an operator in real time to quickly respond to a command of the operator and improving the production efficiency and the response speed. The method comprises the steps that to-be-recognized audio is acquired, feature extraction is carried out on the to-be-recognized audio, a to-be-recognized feature vector corresponding to the to-be-recognized audio is obtained, and the to-be-recognized audio is based on a voice signal of an operator acquired by a voice acquisition device; determining a text instruction corresponding to the to-be-recognized audio based on the to-be-recognized feature vector; when the instruction category corresponding to the text instruction is a command word category, command character data corresponding to the text instruction are transmitted to a programmable logic controller (PLC), so that the PLC controls the casting piece furnace to act based on the command character data to execute the text instruction; and feedback information transmitted by the PLC is received, and audios associated with the feedback information are played based on preset audio playing equipment.

Description

Technical Field

[0001] The present application relates to the technical field of vacuum induction casting furnaces, and in particular to a method and device for controlling the operation process of a casting furnace using an artificial intelligence voice system. Background Art

[0002] The casting furnace, also known as the SC furnace, is a device that melts, refines, casts at a constant flow rate, crushes, and cools pure NdFeB metal raw materials in a crucible under vacuum or inert atmosphere conditions through medium frequency power induction heating to form flakes. In the daily production process of the SC furnace, the actions of the control execution components are mainly operated through the soft switch on the touch screen. The touch screen communicates with the PLC via Modbus UDP to achieve human-computer interaction, visualize the control software inside the PLC, and control the start and stop of the motor and the opening and closing of the valve.

[0003] In the process of implementing this application, the applicant found that the related technology has at least the following problems:

[0004] The touch screen control system of the SC furnace is centrally located in a specific location, which means that the operator must go to this designated location to operate the touch screen in order to issue corresponding instructions. This design causes the operator to frequently move back and forth between the workbench and the touch screen, making the operation process more complicated and increasing the workload of the operator. Summary of the invention

[0005] In view of this, the present application provides a method and device for controlling the operation process of a casting furnace by an artificial intelligence voice system, the main purpose of which is to solve the problem that the touch screen control system of the current SC furnace is centrally set in a specific location, which means that the operator must go to this designated location to operate the touch screen in order to issue corresponding instructions. This design causes the operator to frequently move back and forth between the workbench and the touch screen, making the operation process more complicated and increasing the workload of the operator.

[0006] According to the first aspect of the present application, a method for an artificial intelligence voice system to participate in controlling the operation process of a casting furnace is provided, the method comprising:

[0007] Acquire audio to be recognized, perform feature extraction on the audio to be recognized, and obtain a feature vector to be recognized corresponding to the audio to be recognized, wherein the audio to be recognized is based on a voice signal of an operator collected by a voice collection device, and the voice collection device is set on a plurality of pre-designated operating stations;

[0008] Based on the feature vector to be identified, determining a text instruction corresponding to the audio to be identified;

[0009] When the instruction category corresponding to the text instruction is a command word category, the command character data corresponding to the text instruction is transmitted to a programmable logic controller (PLC), so that the PLC controls the action of the casting furnace based on the command character data to execute the text instruction;

[0010] Receive feedback information transmitted by the PLC, and play audio associated with the feedback information based on a preset audio playback device, wherein the feedback information is generated by the PLC according to the operating state of the casting furnace.

[0011] Optionally, the acquiring the audio to be recognized, performing feature extraction on the audio to be recognized, and obtaining a feature vector to be recognized corresponding to the audio to be recognized includes:

[0012] Preprocessing the audio to be recognized to obtain audio to be extracted, wherein the preprocessing includes but is not limited to noise reduction processing, gain adjustment processing and frame division processing;

[0013] Performing short-time Fourier transform on each frame of the audio signal to be extracted to obtain a frequency spectrum corresponding to the audio to be extracted, and filtering the frequency spectrum using a Mel frequency filter group, calculating the logarithmic energy of each Mel frequency filter output, and obtaining a logarithmic energy sequence;

[0014] Performing discrete cosine transform (DCT) on the logarithmic energy sequence to obtain Mel frequency cepstral coefficient (MFCC) features, inputting the MFCC features into a pre-trained speech recognition model for recognition, and obtaining a text to be recognized corresponding to the audio to be extracted;

[0015] Feature extraction is performed on the text to be recognized to obtain the feature vector to be recognized.

[0016] Optionally, determining the text instruction corresponding to the audio to be recognized based on the feature vector parameter to be recognized includes:

[0017] Calculating the feature similarity between the feature vector to be identified and each feature model stored in a reference model library, and determining a target feature model whose feature similarity is higher than other feature similarities, wherein the reference model library is obtained by extracting features from audio information in a speech playback library and modeling the extracted audio features, and the speech playback library is constructed based on a speech chip of an intelligent speech recognition model;

[0018] The text information mapped by the target feature model is searched in the voice playback library, and the text information is used as the text instruction corresponding to the audio to be recognized.

[0019] Optionally, after determining the text instruction corresponding to the audio to be recognized based on the feature vector to be recognized, the method further includes:

[0020] When the instruction category corresponding to the text instruction is a management category, querying the instruction audio associated with the text instruction in the voice playback library, and using the preset audio playback device to play the instruction audio;

[0021] The text instruction is transmitted to the touch screen for visual display.

[0022] Optionally, the receiving feedback information transmitted by the PLC, and playing audio associated with the feedback information based on a preset audio playback device, includes:

[0023] Extracting features from the feedback information to obtain a feedback text feature vector;

[0024] Calculating the feature similarity between the feedback text feature vector and each text feature vector in the voice playback library, and determining a target text feature vector whose feature similarity is higher than other feature similarities;

[0025] Determine the front audio and action audio associated with the target text feature vector, connect the front audio with the action audio, and generate audio to be played;

[0026] The to-be-played audio is played based on the preset audio playback device.

[0027] Optionally, after playing the to-be-played audio based on the preset audio playback device, the method further includes:

[0028] Determine text information associated with the target text feature vector, and transmit the text information to the touch screen for visual display.

[0029] Optionally, the method further comprises:

[0030] Receiving alarm information transmitted by the PLC, wherein the alarm information is generated by the PLC based on an alarm signal transmitted by the casting furnace equipment;

[0031] Performing feature extraction on the alarm information, determining an alarm text feature vector corresponding to the alarm information, calculating feature similarity between the alarm text feature vector and each text feature vector in the voice playback library, determining a target text feature vector having a feature similarity higher than other feature similarities, and determining an alarm audio associated with the target text feature vector, and playing the alarm audio based on the preset audio playback device; and / or,

[0032] The alarm information is transmitted to the touch screen for visual display.

[0033] According to the second aspect of the present application, there is provided a device for controlling the operation process of a casting furnace using an artificial intelligence voice system, the device comprising:

[0034] An extraction module, used to obtain audio to be recognized, perform feature extraction on the audio to be recognized, and obtain a feature vector to be recognized corresponding to the audio to be recognized, wherein the audio to be recognized is based on a voice signal of an operator collected by a voice collection device, and the voice collection device is set on a plurality of pre-specified operating stations;

[0035] A determination module, used for determining a text instruction corresponding to the audio to be identified based on the feature vector to be identified;

[0036] An execution module, for transmitting command symbol data corresponding to the text instruction to a programmable logic controller (PLC) when the instruction category corresponding to the text instruction is a command word category, so that the PLC controls the action of the casting furnace based on the command symbol data to execute the text instruction;

[0037] The playing module is used to receive the feedback information transmitted by the PLC and play the audio associated with the feedback information based on a preset audio playing device, wherein the feedback information is generated by the PLC according to the action state of the casting furnace.

[0038] Optionally, the extraction module is used to preprocess the audio to be recognized to obtain the audio to be extracted, and the preprocessing includes but is not limited to noise reduction processing, gain adjustment processing and frame processing; perform short-time Fourier transform on each frame of audio signal of the audio to be extracted to obtain the spectrum corresponding to the audio to be extracted, and use a Mel frequency filter group to filter the spectrum, calculate the logarithmic energy of each Mel frequency filter output, and obtain a logarithmic energy sequence; perform discrete cosine transform DCT transform on the logarithmic energy sequence to obtain Mel frequency cepstral coefficient MFCC features, input the MFCC features into a pre-trained speech recognition model for recognition, and obtain the text to be recognized corresponding to the audio to be extracted; perform feature extraction on the text to be recognized to obtain the feature vector to be recognized.

[0039] Optionally, the determination module is used to calculate the feature similarity between the feature vector to be identified and each feature model stored in a reference model library, and determine a target feature model whose feature similarity is higher than other feature similarities, wherein the reference model library is obtained by extracting features from audio information in a speech playback library and modeling the extracted audio features, and the speech playback library is constructed based on a speech chip of an intelligent speech recognition model; the text information mapped by the target feature model is queried in the speech playback library, and the text information is used as the text instruction corresponding to the audio to be identified.

[0040] Optionally, the device further comprises:

[0041] A query module, used for, when the instruction category corresponding to the text instruction is a management category, querying the instruction audio associated with the text instruction in the voice playback library, and using the preset audio playback device to play the instruction audio;

[0042] The first display module is used to transmit the text instruction to the touch screen for visual display.

[0043] Optionally, the playback module is used to perform feature extraction on the feedback information to obtain a feedback text feature vector; calculate the feature similarity between the feedback text feature vector and each text feature vector in the voice playback library, and determine a target text feature vector whose feature similarity is higher than other feature similarities; determine the front audio and action audio associated with the target text feature vector, connect the front audio with the action audio, and generate audio to be played; and play the audio to be played based on the preset audio playback device.

[0044] Optionally, the device further comprises:

[0045] The second display module is used to determine the text information associated with the target text feature vector and transmit the text information to the touch screen for visual display.

[0046] Optionally, the device further comprises:

[0047] A receiving module, used for receiving alarm information transmitted by PLC, wherein the alarm information is generated by the PLC based on an alarm signal transmitted by the casting furnace equipment;

[0048] A calculation module is used to extract features from the alarm information, determine an alarm text feature vector corresponding to the alarm information, calculate feature similarity between the alarm text feature vector and each text feature vector in the voice playback library, determine a target text feature vector whose feature similarity is higher than other feature similarities, and determine an alarm audio associated with the target text feature vector, and play the alarm audio based on the preset audio playback device; and / or,

[0049] The third display module is used to transmit the alarm information to the touch screen for visual display.

[0050] According to a third aspect of the present application, a computer device is provided, comprising a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of any one of the methods described in the first aspect when executing the computer program.

[0051] According to a fourth aspect of the present application, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps of any one of the methods described in the first aspect are implemented.

[0052] By means of the above technical scheme, the present application provides a method and device for an artificial intelligence voice system to participate in controlling the operation process of a casting furnace. The present application first obtains the audio to be recognized, and extracts the features of the audio to be recognized to obtain the feature vector to be recognized corresponding to the audio to be recognized, wherein the audio to be recognized is based on the voice signal of the operator collected by the voice acquisition device, and the voice acquisition device is set on a plurality of pre-specified operating stations. Further, based on the feature vector to be recognized, the text instruction corresponding to the audio to be recognized is determined. When the instruction category corresponding to the text instruction is the command word category, the command character data corresponding to the text instruction is transmitted to the programmable logic controller PLC, so that the PLC controls the action of the casting furnace based on the command character data to execute the text instruction. Finally, the feedback information transmitted by the PLC is received, and the audio associated with the feedback information is played based on the preset audio playback device, and the feedback information is generated by the PLC according to the action state of the casting furnace. In the embodiment of the present application, since the voice recognition system can process the voice instructions of the operator in real time, the action of the casting furnace can respond quickly to the operator's command, thereby improving production efficiency and response speed. In addition, during the production process of the casting furnace, the actions performed and the alarms that occur are played out through the loudspeaker, so that no matter where the operator is near the casting furnace, he can immediately understand the operating status of the equipment and the alarm content, thereby reducing labor intensity, shortening labor time, and improving work efficiency.

[0053] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are listed below. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] Various other advantages and benefits will become apparent to those of ordinary skill in the art by reading the detailed description of the preferred embodiments below. The accompanying drawings are only for the purpose of illustrating the preferred embodiments and are not to be considered as limiting the present application. Also, the same reference symbols are used throughout the accompanying drawings to represent the same components. In the accompanying drawings:

[0055] Figure 1 A schematic flow chart of a method for an artificial intelligence voice system to participate in controlling the operation process of a casting furnace provided in an embodiment of the present application is shown;

[0056] Figure 2 A schematic flow chart of a method for an artificial intelligence voice system to participate in controlling the operation process of a casting furnace provided in an embodiment of the present application is shown;

[0057] Figure 3A schematic diagram of the structure of a device for controlling the operation process of a casting furnace using an artificial intelligence voice system provided in an embodiment of the present application is shown;

[0058] Figure 4 A schematic diagram of the device structure of a computer device provided in an embodiment of the present application is shown. DETAILED DESCRIPTION

[0059] The exemplary embodiments of the present application will be described in more detail below with reference to the accompanying drawings. Although the exemplary embodiments of the present application are shown in the accompanying drawings, it should be understood that the present application can be implemented in various forms and should not be limited by the embodiments set forth herein. On the contrary, these embodiments are provided in order to enable a more thorough understanding of the present application and to fully convey the scope of the present application to those skilled in the art.

[0060] The casting furnace, also known as the SC furnace, is a device that melts, refines, casts at a constant flow rate, crushes, and cools pure NdFeB metal raw materials in a crucible under vacuum or inert atmosphere conditions through medium frequency power induction heating to form scales. In the daily production process of the SC furnace, the actions of the control execution components are mainly operated through the soft switches on the touch screen. The touch screen communicates with the PLC via Modbus UDP to achieve human-computer interaction, visualize the control software inside the PLC, and control the start and stop of the motor and the opening and closing of the valve. However, the applicant realized that the touch screen control system of the SC furnace is centrally set in a specific location, which means that the operator must go to this designated location to operate the touch screen in order to issue corresponding instructions. This design causes the operator to frequently move back and forth between the workbench and the touch screen, making the operation process more complicated and increasing the workload of the operator. Therefore, the present application provides a method for an artificial intelligence voice system to participate in controlling the operation process of a casting furnace. The present application first obtains the audio to be recognized, and extracts the features of the audio to be recognized to obtain the feature vector to be recognized corresponding to the audio to be recognized, wherein the audio to be recognized is based on the voice signal of the operator collected by the voice acquisition device, and the voice acquisition device is set on multiple pre-specified operating stations. Further, based on the feature vector to be recognized, the text instruction corresponding to the audio to be recognized is determined. When the instruction category corresponding to the text instruction is the command word category, the command character data corresponding to the text instruction is transmitted to the programmable logic controller PLC, so that the PLC controls the action of the casting furnace based on the command character data to execute the text instruction. Finally, the feedback information transmitted by the PLC is received, and the audio associated with the feedback information is played based on the preset audio playback device. The feedback information is generated by the PLC according to the action state of the casting furnace. In the embodiment of the present application, since the voice recognition system can process the voice instructions of the operator in real time, the action of the casting furnace can respond quickly to the operator's command, thereby improving production efficiency and response speed. In addition, during the production process of the casting furnace, the actions performed and the alarms that occur are played out through the loudspeaker, so that no matter where the operator is near the casting furnace, he can immediately understand the operating status of the equipment and the alarm content, thereby reducing labor intensity, shortening labor time, and improving work efficiency.

[0061] The present application embodiment provides a method for an artificial intelligence voice system to participate in controlling the operation process of a casting furnace, such as Figure 1 As shown, the method includes:

[0062] S10, obtaining audio to be recognized, performing feature extraction on the audio to be recognized, and obtaining a feature vector to be recognized corresponding to the audio to be recognized, wherein the audio to be recognized is based on a voice signal of an operator collected by a voice collection device, and the voice collection device is set on multiple pre-designated operating stations.

[0063] The casting furnace, also known as the SC furnace, is a device that melts, refines, casts at a constant flow rate, crushes, and cools pure NdFeB metal raw materials in a crucible under vacuum or inert atmosphere conditions through medium frequency power induction heating, thereby forming flakes. In the daily production process of the SC furnace, the SC furnace control system is used to control the casting furnace equipment. The system includes a touch screen and a PLC. The PLC hardware configuration uses Mitsubishi Q series PLC, the CPU unit is Q06UVDCPU, the communication unit includes QJ71C24N, and the PLC programming software is Works2. The hardware of the touch screen control system is the PS6000 series industrial computer of the Pro face brand, the running software is Win GP of the Pro face brand, and the development software is Pro EX. The operator controls the action of the execution component mainly through the soft switch on the touch screen. The touch screen communicates with the PLC via Modbus UDP, thereby achieving human-computer interaction, visualizing the control software inside the PLC, and controlling the start and stop of the motor and the opening and closing of the valve. While the action is being executed, the touch screen will execute historical action records for the convenience of the operator to query. In the daily production process of SC furnace, when the equipment alarms, the alarm buzzer of the control system will sound to remind the operator that an alarm has occurred, and the alarm interface of the touch screen will have an alarm pop-up window to remind the operator of the specific alarm content. When the alarm occurs, the touch screen will execute the historical alarm record to facilitate the operator to query.

[0064] This application adds an intelligent voice recognition module to the existing SC furnace control system. The model of the intelligent voice recognition module is WTK6900HM01. WTK6900HM01 communicates with the QJ71C24N serial communication module of the Mitsubishi Q series PLC in the existing control system via RS485 (initializing the serial port, setting the baud rate, sending and receiving data), so that WTK6900HM01 establishes data interaction with the PLC and realizes the reception and execution of voice commands. WTK6900HM01 integrates a voice chip, and a voice playback library for daily operation actions of the SC furnace is established based on the voice chip, including actions such as "starting the sliding valve pump", "turning off the intermediate frequency power supply", and "executing the recovery container rise". In the "Audio Settings" under the "General Settings" of the Pro EX programming software of the touch screen, the audio of the execution action and alarm content in the WAV file format is registered and downloaded to Win GP, ​​such as "temperature measurement occasional break alarm", "crucible leakage alarm", "execute melting chamber emptying", "execute guide cylinder closing" and other preset audio content. Among them, the registered audio content is edited and designed through the "Text-to-Speech Master" APP. Plug the sound card into the hardware USB interface of the PS6000 industrial computer, and connect the sound card to the preset audio playback device, such as a speaker, through an audio cable, so that the corresponding audio can be played based on the preset audio playback device.

[0065] On this basis, the operator needs to install voice acquisition equipment at the key operating stations on each floor of the SC furnace before the actual operation of the system, wherein the voice acquisition equipment may be a microphone, a sound acquisition pickup, an audio power amplifier, an audio analyzer, etc. In the embodiment of the present application, the voice acquisition equipment may select a microphone. The key operating station may be an operating station pre-designated by the operator according to actual needs, specifically an operating station in the smelting control area, an operating station in the casting control area, etc. The present application does not specifically limit the pre-designated operating station.

[0066] After completing a series of necessary preparations, when operators need to control the SC furnace and perform specific operations during actual operation, they can directly use the voice acquisition device to issue corresponding voice commands at their respective operating stations. For example, the operator can say a voice command such as "start the sliding valve pump" to the voice acquisition device. The voice acquisition device will capture the voice signal issued by the operator in real time and extract the audio information to be recognized from it. Once the audio information to be recognized is captured, the system will process it according to the established steps S11 to S14, as shown below. Figure 2 As shown, feature extraction is performed on the audio information to obtain a feature vector corresponding to the audio to be identified.

[0067] S11, preprocessing the audio to be recognized to obtain the audio to be extracted, the preprocessing including but not limited to noise reduction processing, gain adjustment processing and frame division processing.

[0068] In this step, the intelligent speech recognition module performs a series of preprocessing operations on the audio to be recognized, in order to improve the quality of the speech signal, so that the subsequent feature extraction process is smoother and more efficient. Specifically, the continuous audio signal is first segmented into several shorter frames, and the duration of each frame can be set between 20 and 40 milliseconds according to actual needs. In order to better capture the details of the audio signal, a certain overlap between frames is usually set, such as an overlap rate of 50%. This ensures that important speech information will not be missed during the segmentation process. Next, in order to further improve the signal-to-noise ratio, some noise reduction algorithms are applied to remove environmental noise in the audio signal. Common noise reduction algorithms include spectral subtraction and Wiener filtering, which can effectively identify and eliminate background noise, making the speech signal clearer. Through noise reduction processing, the speech recognition module can more accurately capture the key information in the speech and improve the accuracy of recognition. Finally, in order to ensure the stability of feature extraction, the audio signal is gain-adjusted. By adjusting the gain of each frame, the energy of each frame is kept consistent to avoid instability in the feature extraction process caused by excessive energy fluctuations. In this way, the feature extraction algorithm can operate on a relatively stable signal basis, thereby improving the accuracy and reliability of feature extraction. Through these preprocessing steps, the intelligent speech recognition module can lay a solid foundation for the subsequent feature extraction and recognition process.

[0069] S12, performing short-time Fourier transform on each frame of the audio signal to be extracted to obtain a frequency spectrum corresponding to the audio to be extracted, and filtering the frequency spectrum using a Mel frequency filter group, calculating the logarithmic energy of each Mel frequency filter output, and obtaining a logarithmic energy sequence.

[0070] Furthermore, in order to better match the acoustic model, it is necessary to extract representative features from the preprocessed audio signal. Specifically, firstly, a short-time Fourier transform is performed on each frame of the audio signal to convert the signal originally in the time domain into the frequency domain, so as to obtain the spectrum corresponding to each frame of audio. This process can reveal the distribution of the audio signal at different frequencies. Next, the obtained spectrum is further filtered using a Mel frequency filter bank. The Mel frequency filter bank is a filter bank that simulates the auditory characteristics of the human ear, which can better capture the key information in the audio signal. In this way, the logarithmic energy of each filter output can be obtained, and these logarithmic energy values ​​constitute a sequence, namely the logarithmic energy sequence, which reflects the energy distribution of the audio signal in different frequency bands. Subsequently, by analyzing and processing these logarithmic energy sequences, more refined and discriminative features can be extracted, which will be used in the subsequent acoustic model matching process, thereby improving the accuracy and reliability of speech recognition.

[0071] S13, performing discrete cosine transform (DCT) on the logarithmic energy sequence to obtain Mel-frequency cepstral coefficient (MFCC) features, inputting the MFCC features into a pre-trained speech recognition model for recognition, and obtaining the text to be recognized corresponding to the audio to be extracted.

[0072] Further, the logarithmic energy sequence is transformed by DCT (Discrete Cosine Transform) to obtain MFCC (Mel Frequency Cepstral Coefficients) features. It should be noted that in the actual operation process, in addition to the static MFCC features, dynamic features (such as first-order differences and second-order differences) can also be extracted. The extraction of these dynamic features is to capture the changes between frames, so as to better capture the dynamic characteristics of the speech signal. Next, the extracted MFCC features are input into the pre-trained speech recognition model for recognition. Specifically, the extracted feature sequence can be matched one by one with the phoneme or phrase sequence in the acoustic model, and the acoustic matching score is calculated. Finally, the text to be recognized corresponding to the audio to be extracted is determined according to the acoustic matching score. In the actual operation process, the text to be recognized can also be post-processed, spelling errors can be corrected using a spelling check algorithm (such as an edit distance algorithm), and the recognition results can be normalized, etc., to correct errors and improve readability.

[0073] S14, extracting features from the text to be recognized to obtain a feature vector to be recognized.

[0074] In this step, the feature extraction of the text to be recognized can be performed by using the Bag of Words model. The core idea of ​​this method is to count the words in the text, ignore the order and grammatical structure of the words, and convert the text into a feature vector. In this vector, each dimension corresponds to a specific word in the vocabulary, and the value of the dimension represents the frequency or number of times the word appears in the text to be recognized. In this way, the text data can be converted into a numerical form that can be processed by the machine learning model, that is, converted into a feature vector to be recognized.

[0075] In addition, word embedding technology can also be used to extract features from the text to be recognized. Word embedding is a method of mapping words to low-dimensional dense vector space. These vectors can capture the semantic relationship and contextual information between words. Unlike the bag-of-words model, word embedding not only considers the frequency of occurrence of words, but also the semantic and contextual information of words, so that it can better represent the meaning of the text. Common word embedding methods include Word2Vec, GloVe, and BERT. Word2Vec generates word vectors by training neural networks to capture the similarity and contextual relationship between words. GloVe generates word vectors based on global word frequency statistics, which can effectively capture the co-occurrence relationship between words. BERT (Bidirectional Encoder Representations from Transformers) is a pre-trained language representation model based on bidirectional Transformers, which can generate more complex and rich word embedding representations and capture the meaning of words in different contexts. Through these methods, more accurate and rich feature vectors to be recognized can be obtained, thereby improving the performance of machine learning models.

[0076] S20: Determine the text instruction corresponding to the audio to be recognized based on the feature vector to be recognized.

[0077] Specifically, the system will compare the feature vector to be identified with each feature model stored in the reference model library one by one. By calculating the similarity between the feature vector to be identified and each feature model, the system can identify a specific feature model whose similarity value is significantly higher than all other feature models. This reference model library is constructed by extracting features from the audio information in the voice playback library and modeling the extracted audio features. The voice playback library is built on a voice chip based on the intelligent voice recognition model. After determining the target feature model, the system will perform further queries in the voice playback library to obtain detailed text information corresponding to the target feature model. These text information are actually text instructions corresponding to the audio to be identified. Through this series of complex and precise processes, the system can convert voice information into text instructions, thereby realizing efficient voice recognition function.

[0078] After determining the text instructions corresponding to the audio to be recognized, the text instructions need to be recognized next to determine the instruction category of the text instructions. Instruction categories include management categories and command word categories. Management categories include power-on language instructions, exit wake-up instructions, and wake-up word instructions, etc. Command word categories include "lubrication pump start instruction", "melting chamber argon gas introduction instruction", "furnace cover opening instruction", "melting heating start instruction", "intermediate chamber vent valve opening instruction", "start temperature measurement instruction", "start automatic recovery instruction", etc.

[0079] When the instruction category corresponding to the text instruction is the management category, the system will query the instruction audio associated with the text instruction in the voice playback library, and use the preset audio playback device to play the instruction audio. At the same time, the text instruction is passed to the touch screen for visual display. For example, when the instruction category corresponding to the text instruction is the management category, if the text instruction indicates a power-on language instruction, the system will determine the instruction audio associated with the power-on language instruction, and play the audio of "Welcome to the SC furnace intelligent operation housekeeper switch panel". If the text instruction indicates an exit wake-up instruction, the system will determine the instruction audio associated with the exit wake-up instruction and play the "Exit" audio. If the text instruction indicates a wake-up word instruction, the system will determine the instruction audio associated with the wake-up word instruction and play the "I am here" audio. In this way, the system can effectively convert text instructions into audio instructions and display them visually on the touch screen, thereby achieving effective management and control of the equipment.

[0080] S30. When the instruction category corresponding to the text instruction is a command word category, the command character data corresponding to the text instruction is transmitted to the programmable logic controller PLC, so that the PLC controls the action of the casting furnace based on the command character data to execute the text instruction.

[0081] In this step, when the instruction category corresponding to the text instruction is the command word category, the system will automatically transmit the command symbol data corresponding to the text instruction to the programmable logic controller PLC. After receiving these command symbol data, the PLC will execute the corresponding application instructions in the PLC ladder diagram based on these data to control the corresponding actions of the casting furnace, ensuring that the casting furnace can perform the corresponding operations according to the requirements of the text instruction. This process not only ensures the accurate transmission of instructions, but also ensures the accurate execution of instructions, so that the entire production process can proceed smoothly, improving production efficiency and operation accuracy.

[0082] S40, receiving feedback information transmitted by the PLC, and playing audio associated with the feedback information based on a preset audio playback device, wherein the feedback information is generated by the PLC according to the operating state of the casting furnace.

[0083] When processing text information in the voice playback database, special attention should be paid to distinguishing between voice playback and receiving language commands. In order to prevent the phenomenon of infinite loop caused by audio input and output confusion, specific measures are taken in this step. Specifically, while the SC furnace terminal command is executed, the PLC will generate feedback information. Further, the system will extract features from the feedback information to obtain a feedback text feature vector. Next, the system will calculate the feature similarity between this feedback text feature vector and each text feature vector in the voice playback library. Then determine a target text feature vector with a feature similarity higher than all other text feature vectors. Once the target text feature vector is determined, the system will further associate it with the related pre-audio and action audio. The pre-audio here refers to the audio played before performing a specific action, such as the "execution" audio; and the action audio refers to the audio played when performing a specific action, such as the "crucible tilting forward" audio. The system connects these two audios to generate an audio file to be played "execution" + "crucible tilting forward". Based on the preset audio playback device, the system will play this audio file to be played. In addition, the system can also determine the text information associated with the target text feature vector. This text information will be transmitted to the touch screen for visual display so that the user can intuitively see the relevant information, which can be displayed visually in the form of pop-up windows. In this way, the system can not only guide the user to perform specific operations through audio playback, but also enhance the user's understanding and accuracy of operation through visual display.

[0084] In the daily production process of the SC furnace, when a specific action is performed or an alarm event occurs, the speaker will respond immediately and trigger the corresponding audio information through the instructions of the PLC. These audio information are quickly transmitted to the operators on site through the speaker system, ensuring that they can grasp the operating status of the equipment and any abnormal conditions at the first time. Specifically, the system receives alarm information from the PLC in real time, which is generated by the PLC according to the alarm signal transmitted by the casting furnace equipment. Subsequently, the system extracts the features of these alarm information and extracts the text feature vectors corresponding to the alarm information. Next, the system calculates the feature similarity between these alarm text feature vectors and each text feature vector stored in the voice playback library. Then, a target text feature vector with a feature similarity higher than all other feature vectors is determined, and the system retrieves the alarm audio file associated with the target text feature vector from the voice playback library. Then, the system plays this alarm audio through the preset audio playback device so that the operator can quickly identify and respond to the alarm event through hearing. In addition, the system can also transmit the alarm information to the touch screen interface for visual display, so that the operator can understand the specific content and severity of the alarm information not only through sound but also through vision. This multi-sensory alarm transmission method greatly improves the operator's perception and response speed to equipment status, thereby ensuring the safety and efficiency of the production process.

[0085] In summary, in the actual operation process, the system receives the operator's voice command through the microphone near the SC furnace operation station, and sends the audio to be recognized to the intelligent voice recognition module through the microphone. The system recognizes the audio to be recognized based on the intelligent voice recognition module and converts the audio to be recognized into text to be recognized. Furthermore, based on the intelligent voice recognition module, feature extraction is performed on the text to be recognized to obtain feature vector parameters, and the similarity metric is compared with each feature model in the reference model library mapped by the voice playback library of the voice chip, and the text instruction with the highest similarity is output. Then the command symbol of the text instruction is sent to the PLC, and the PLC executes the corresponding application instruction in the ladder diagram. The PLC controls the SC furnace terminal to perform corresponding actions, such as starting the motor, closing the valve, the trolley travels, and performing temperature measurement and other processes. Finally, while the SC furnace terminal command is executed, the process to be executed is output by audio to prompt the operator. In addition, when the SC furnace alarm occurs, the alarm content is also output by audio to prompt the operator.

[0086] The method provided in the embodiment of the present application first obtains the audio to be recognized, and extracts the features of the audio to be recognized to obtain the feature vector to be recognized corresponding to the audio to be recognized, wherein the audio to be recognized is based on the voice signal of the operator collected by the voice acquisition device, and the voice acquisition device is set on multiple pre-specified operating stations. Further, based on the feature vector to be recognized, the text instruction corresponding to the audio to be recognized is determined. When the instruction category corresponding to the text instruction is the command word category, the command character data corresponding to the text instruction is transmitted to the programmable logic controller PLC, so that the PLC controls the action of the casting furnace based on the command character data to execute the text instruction. Finally, the feedback information transmitted by the PLC is received, and the audio associated with the feedback information is played based on the preset audio playback device. The feedback information is generated by the PLC according to the action state of the casting furnace. In the embodiment of the present application, since the voice recognition system can process the voice instructions of the operator in real time, the action of the casting furnace can respond quickly to the operator's command, thereby improving production efficiency and response speed. In addition, during the production process of the casting furnace, the actions performed and the alarms that occur are played out through the loudspeaker, so that no matter where the operator is near the casting furnace, he can immediately understand the operating status of the equipment and the alarm content, thereby reducing labor intensity, shortening labor time, and improving work efficiency.

[0087] Further, as Figure 1 The specific implementation of the method, the embodiment of the present application provides a device for an artificial intelligence voice system to participate in controlling the operation process of a casting furnace, such as Figure 3 As shown, the device includes: an extraction module 301, a determination module 302, an execution module 303, and a playback module 304.

[0088] The extraction module 301 is used to obtain the audio to be recognized, perform feature extraction on the audio to be recognized, and obtain a feature vector to be recognized corresponding to the audio to be recognized, wherein the audio to be recognized is based on the voice signal of the operator collected by the voice collection device, and the voice collection device is set on multiple pre-specified operating stations;

[0089] The determination module 302 is used to determine the text instruction corresponding to the audio to be recognized based on the feature vector to be recognized;

[0090] The execution module 303 is used to transmit the command symbol data corresponding to the text instruction to the programmable logic controller PLC when the instruction category corresponding to the text instruction is the command word category, so that the PLC controls the casting furnace action based on the command symbol data to execute the text instruction;

[0091] The playing module 304 is used to receive feedback information transmitted by the PLC, and play audio associated with the feedback information based on a preset audio playing device, wherein the feedback information is generated by the PLC according to the operation status of the casting furnace.

[0092] In a specific application scenario, the extraction module 301 is used to preprocess the audio to be recognized to obtain the audio to be extracted, and the preprocessing includes but is not limited to noise reduction processing, gain adjustment processing and frame processing; short-time Fourier transform is performed on each frame of the audio signal of the audio to be extracted to obtain the spectrum corresponding to the audio to be extracted, and the spectrum is filtered using a Mel frequency filter group, and the logarithmic energy output by each Mel frequency filter is calculated to obtain a logarithmic energy sequence; discrete cosine transform DCT is performed on the logarithmic energy sequence to obtain Mel frequency cepstrum coefficient MFCC features, and the MFCC features are input into a pre-trained speech recognition model for recognition to obtain the text to be recognized corresponding to the audio to be extracted; feature extraction is performed on the text to be recognized to obtain the feature vector to be recognized.

[0093] In a specific application scenario, the determination module 302 is used to calculate the feature similarity between the feature vector to be identified and each feature model stored in the reference model library, and determine a target feature model whose feature similarity is higher than other feature similarities. The reference model library is obtained by extracting features from the audio information in the speech playback library and modeling the extracted audio features. The speech playback library is constructed based on a speech chip of an intelligent speech recognition model. The text information mapped by the target feature model is queried in the speech playback library, and the text information is used as a text instruction corresponding to the audio to be identified.

[0094] In a specific application scenario, the device further includes: a query module 305 and a first display module 306 .

[0095] The query module 305 is used to query the instruction audio associated with the text instruction in the voice playback library when the instruction category corresponding to the text instruction is the management category, and use the preset audio playback device to play the instruction audio;

[0096] The first display module 306 is used to transmit the text instruction to the touch screen for visual display.

[0097] In a specific application scenario, the playback module 304 is used to perform feature extraction on the feedback information to obtain a feedback text feature vector; calculate the feature similarity between the feedback text feature vector and each text feature vector in the voice playback library, and determine a target text feature vector whose feature similarity is higher than other feature similarities; determine the front audio and action audio associated with the target text feature vector, connect the front audio with the action audio, and generate audio to be played; and play the audio to be played based on the preset audio playback device.

[0098] In a specific application scenario, the device further includes: a second display module 307 .

[0099] The second display module 307 is used to determine the text information associated with the target text feature vector, and transmit the text information to the touch screen for visual display.

[0100] In a specific application scenario, the device further includes: a receiving module 308 , a calculating module 309 , and a third display module 310 .

[0101] The receiving module 308 is used to receive the alarm information transmitted by the PLC, wherein the alarm information is generated by the PLC based on the alarm signal transmitted by the casting furnace equipment;

[0102] The calculation module 309 is used to extract features from the alarm information, determine the alarm text feature vector corresponding to the alarm information, calculate the feature similarity between the alarm text feature vector and each text feature vector in the voice playback library, determine a target text feature vector whose feature similarity is higher than other feature similarities, and determine the alarm audio associated with the target text feature vector, and play the alarm audio based on the preset audio playback device; and / or,

[0103] The third display module 310 is used to transmit the alarm information to the touch screen for visual display.

[0104] The device provided in the embodiment of the present application first obtains the audio to be recognized, and extracts the features of the audio to be recognized to obtain the feature vector to be recognized corresponding to the audio to be recognized, wherein the audio to be recognized is based on the voice signal of the operator collected by the voice acquisition device, and the voice acquisition device is set on multiple pre-specified operating stations. Further, based on the feature vector to be recognized, the text instruction corresponding to the audio to be recognized is determined. When the instruction category corresponding to the text instruction is the command word category, the command character data corresponding to the text instruction is transmitted to the programmable logic controller PLC, so that the PLC controls the action of the casting furnace based on the command character data to execute the text instruction. Finally, the feedback information transmitted by the PLC is received, and the audio associated with the feedback information is played based on the preset audio playback device. The feedback information is generated by the PLC according to the action state of the casting furnace. In the embodiment of the present application, since the voice recognition system can process the voice instructions of the operator in real time, the action of the casting furnace can respond quickly to the operator's command, thereby improving production efficiency and response speed. In addition, during the production process of the casting furnace, the actions performed and the alarms that occur are played out through the loudspeaker, so that no matter where the operator is near the casting furnace, he can immediately understand the operating status of the equipment and the alarm content, thereby reducing labor intensity, shortening labor time, and improving work efficiency.

[0105] It should be noted that for other corresponding descriptions of the functional units involved in the device for controlling the operation process of the casting furnace provided by the embodiment of the present application, reference can be made to Figure 1 and Figure 2 The corresponding description in will not be repeated here.

[0106] In an exemplary embodiment, see Figure 4 , and also provides a device, which includes a communication bus, a processor, a memory and a communication interface, and may also include an input and output interface and a display device, wherein each functional unit can communicate with each other through the bus. The memory stores a computer program, and the processor is used to execute the program stored in the memory and execute the method in the above embodiment where the artificial intelligence voice system participates in controlling the operation process of the casting furnace.

[0107] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of the method in which an artificial intelligence voice system participates in controlling the operation process of a casting furnace.

[0108] Through the description of the above implementation methods, those skilled in the art can clearly understand that the present application can be implemented by hardware, or by means of software plus a necessary general hardware platform. Based on such an understanding, the technical solution of the present application can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (which can be a CD-ROM, a USB flash drive, a mobile hard disk, etc.), including a number of instructions for enabling a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each implementation scenario of the present application.

[0109] Those skilled in the art will appreciate that the accompanying drawings are merely schematic diagrams of a preferred implementation scenario, and the modules or processes in the accompanying drawings are not necessarily required for implementing the present application.

[0110] Those skilled in the art will appreciate that the modules in the device in the implementation scenario can be distributed in the device in the implementation scenario according to the implementation scenario description, or can be changed accordingly and located in one or more devices different from the implementation scenario. The modules in the above implementation scenario can be combined into one module, or can be further split into multiple sub-modules.

[0111] The above application serial numbers are for description only and do not represent the advantages or disadvantages of the implementation scenarios.

[0112] The above disclosure only discloses several specific implementation scenarios of the present application. However, the present application is not limited thereto, and any changes that can be conceived by technicians in this field should fall within the scope of protection of the present application.

Claims

1. A method for controlling the operation process of a casting furnace using an artificial intelligence voice system, characterized in that: include: Acquire audio to be recognized, perform feature extraction on the audio to be recognized, and obtain a feature vector to be recognized corresponding to the audio to be recognized, wherein the audio to be recognized is based on a voice signal of an operator collected by a voice collection device, and the voice collection device is set on a plurality of pre-designated operating stations; Based on the feature vector to be identified, determining a text instruction corresponding to the audio to be identified; When the instruction category corresponding to the text instruction is a command word category, the command character data corresponding to the text instruction is transmitted to a programmable logic controller (PLC), so that the PLC controls the action of the casting furnace based on the command character data to execute the text instruction; Receive feedback information transmitted by the PLC, and play audio associated with the feedback information based on a preset audio playback device, wherein the feedback information is generated by the PLC according to the operating state of the casting furnace.

2. The method according to claim 1, characterized in that The step of obtaining the audio to be recognized, performing feature extraction on the audio to be recognized, and obtaining a feature vector to be recognized corresponding to the audio to be recognized includes: Preprocessing the audio to be recognized to obtain audio to be extracted, wherein the preprocessing includes but is not limited to noise reduction processing, gain adjustment processing and frame division processing; Performing short-time Fourier transform on each frame of the audio signal to be extracted to obtain a frequency spectrum corresponding to the audio to be extracted, and filtering the frequency spectrum using a Mel frequency filter group, calculating the logarithmic energy of each Mel frequency filter output, and obtaining a logarithmic energy sequence; Performing discrete cosine transform (DCT) on the logarithmic energy sequence to obtain Mel frequency cepstral coefficient (MFCC) features, inputting the MFCC features into a pre-trained speech recognition model for recognition, and obtaining a text to be recognized corresponding to the audio to be extracted; Feature extraction is performed on the text to be recognized to obtain the feature vector to be recognized.

3. The method according to claim 1, characterized in that The step of determining the text instruction corresponding to the audio to be recognized based on the feature vector parameter to be recognized includes: Calculating the feature similarity between the feature vector to be identified and each feature model stored in a reference model library, and determining a target feature model whose feature similarity is higher than other feature similarities, wherein the reference model library is obtained by extracting features from audio information in a speech playback library and modeling the extracted audio features, and the speech playback library is constructed based on a speech chip of an intelligent speech recognition model; The text information mapped by the target feature model is searched in the voice playback library, and the text information is used as the text instruction corresponding to the audio to be recognized.

4. The method according to claim 1, characterized in that: After determining the text instruction corresponding to the audio to be recognized based on the feature vector to be recognized, the method further includes: When the instruction category corresponding to the text instruction is a management category, querying the instruction audio associated with the text instruction in the voice playback library, and using the preset audio playback device to play the instruction audio; The text instruction is transmitted to the touch screen for visual display.

5. The method according to claim 1, characterized in that The receiving feedback information transmitted by the PLC, and playing the audio associated with the feedback information based on a preset audio playback device, includes: Extracting features from the feedback information to obtain a feedback text feature vector; Calculating the feature similarity between the feedback text feature vector and each text feature vector in the voice playback library, and determining a target text feature vector whose feature similarity is higher than other feature similarities; Determine the front audio and action audio associated with the target text feature vector, connect the front audio with the action audio, and generate audio to be played; The to-be-played audio is played based on the preset audio playback device.

6. The method according to claim 5, characterized in that After playing the audio to be played based on the preset audio playback device, the method further includes: Determine text information associated with the target text feature vector, and transmit the text information to the touch screen for visual display.

7. The method according to claim 1, characterized in that The method further comprises: Receiving alarm information transmitted by PLC, wherein the alarm information is generated by the PLC based on an alarm signal transmitted by the casting furnace equipment; Performing feature extraction on the alarm information, determining an alarm text feature vector corresponding to the alarm information, calculating feature similarity between the alarm text feature vector and each text feature vector in the voice playback library, determining a target text feature vector having a feature similarity higher than other feature similarities, and determining an alarm audio associated with the target text feature vector, and playing the alarm audio based on the preset audio playback device; and / or, The alarm information is transmitted to the touch screen for visual display.

8. An artificial intelligence voice system participates in controlling the operation process of a casting furnace, characterized in that: include: An extraction module, used to obtain audio to be recognized, perform feature extraction on the audio to be recognized, and obtain a feature vector to be recognized corresponding to the audio to be recognized, wherein the audio to be recognized is based on a voice signal of an operator collected by a voice collection device, and the voice collection device is set on a plurality of pre-specified operating stations; A determination module, used for determining a text instruction corresponding to the audio to be identified based on the feature vector to be identified; An execution module, for transmitting command symbol data corresponding to the text instruction to a programmable logic controller (PLC) when the instruction category corresponding to the text instruction is a command word category, so that the PLC controls the action of the casting furnace based on the command symbol data to execute the text instruction; The playing module is used to receive the feedback information transmitted by the PLC and play the audio associated with the feedback information based on a preset audio playing device, wherein the feedback information is generated by the PLC according to the action state of the casting furnace.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images