Aluminum leakage judgment model generation method, aluminum leakage detection method and equipment

Abstract

The invention relates to the technical field of aluminum processing supervision, in particular to an aluminum leakage judgment model generation method, an aluminum leakage detection method and aluminum leakage detection equipment. Acquiring a first voiceprint signal without aluminum leakage in a casting state and a noise power spectrum density; acquiring a second voiceprint signal when aluminum leakage occurs in the casting state and environmental sound power spectral density; calculating a noise suppression gain function; training voiceprint signals are obtained, and casting voiceprint features are extracted based on the noise suppression gain function; classifying the casting voiceprint features to obtain non-leakage aluminum feature tags and leakage aluminum feature tags, and training to obtain an aluminum leakage judgment model; real-time voiceprint signals in the casting state are collected, and real-time voiceprint features are obtained; the real-time voiceprint features are input into an aluminum leakage judgment model, and a judgment result is waited to be output; and if the judgment result corresponds to the aluminum leakage feature tag, aluminum leakage is determined, and alarm information is output. The method is used for improving the aluminum leakage detection effect in the multi-noise environment.

Description

Technical Field

[0001] The present application relates to the technical field of aluminum processing supervision, and in particular to a method for generating an aluminum leakage determination model, and an aluminum leakage detection method and equipment. Background Art

[0002] The aluminum processing industry is a high-risk industry, especially in the casting workshop, where there is often a high risk of explosion. The safety hazards in the aluminum processing industry are often caused by aluminum leakage. The reasons for aluminum liquid leakage usually include two factors: one is the leakage that is difficult to avoid in the production process itself, and the other is the leakage caused by improper behavior of workers in production operations.

[0003] Initially, the aluminum processing industry had a low level of informatization in production safety monitoring. Production safety warnings mainly relied on experience, that is, they mainly relied on manual monitoring. The safety warnings in this way were mostly experience-driven and heavily dependent on human monitoring and judgment.

[0004] In current technology, with the continuous improvement of computer technology, the construction of workshop safety tends to be intelligent, and the aluminum processing industry has generally realized the method of human-machine joint monitoring. The staff monitors and inspects based on experience, and the monitoring equipment determines whether aluminum leakage occurs through hearing, such as obtaining the sound signal during the processing to determine whether aluminum leakage occurs. However, because the ambient sound during the processing includes the signal emitted by aluminum leakage, it will also include environmental noise, such as water inlet and outlet noise, equipment noise, etc. The current technology cannot effectively and accurately monitor the sound corresponding to aluminum leakage among many messy voiceprints during the actual sound signal acquisition process.

[0005] Therefore, the current technology for aluminum leakage detection has poor detection accuracy and detection effect, and it is impossible to perform rapid and accurate detection of aluminum leakage. Summary of the invention

[0006] In order to improve the aluminum leakage detection effect in a noisy environment, the present application provides a method for generating an aluminum leakage determination model, an aluminum leakage detection method and a device.

[0007] In the first aspect, the present application provides a method for generating an aluminum leakage determination model, which adopts the following technical solution: A method for generating an aluminum leakage determination model comprises the following steps: Acquire a first soundprint signal in a casting state without aluminum leakage, and calculate the noise power spectrum density corresponding to the first soundprint signal; Acquire a second soundprint signal of aluminum leakage in a casting state, and calculate the ambient sound power spectrum density corresponding to the second soundprint signal; Calculating a noise suppression gain function based on the noise power spectrum density and the ambient sound power spectrum density; Acquire a training voiceprint signal, and extract a casting voiceprint feature in the training voiceprint signal based on the noise suppression gain function; The casting voiceprint features are classified to obtain feature labels of no aluminum leakage and feature labels of aluminum leakage, and the casting voiceprint features are trained based on the feature labels of no aluminum leakage and the feature labels of aluminum leakage to obtain an aluminum leakage determination model.

[0008] In some embodiments, calculating the noise power spectrum density corresponding to the first voiceprint signal includes the following steps: Performing Fourier transform on the first voiceprint signal; The power spectrum density of the first voiceprint signal after Fourier transformation is calculated and smoothed to obtain the theoretical noise power spectrum density. The specific formula is as follows: , in, It is represented by the theoretical noise power spectrum density after smoothing, a is represented by the smoothing coefficient, It is represented by the power spectrum density of the frequency domain signal of the first voiceprint signal, and w is represented by the number of frames of the first voiceprint signal.

[0009] In some embodiments, calculating the noise power spectrum density corresponding to the first voiceprint signal further includes the following steps: The casting state is set to multi-furnace casting, and the first voiceprint signal of a preset time is collected; Based on the theoretical noise power spectrum density, the mean value of the noise power spectrum density after smoothing per second in the preset time is calculated, and the mean value of the noise power spectrum density is used as the noise power spectrum density. The specific formula is as follows: , in, It is represented by the mean value of the noise power spectrum density, and V is represented by the length of the preset time.

[0010] In some embodiments, calculating the ambient sound power spectrum density corresponding to the second voiceprint signal includes the following steps: Perform Fourier transform on the second voiceprint signal to obtain: , in, It is characterized by the frequency domain signal corresponding to the second voiceprint signal. It is characterized by the frequency domain signal corresponding to the aluminum leakage soundprint signal. A frequency domain signal characterized as the first voiceprint signal; The ambient sound power spectrum density corresponding to the second voiceprint signal is calculated based on the following formula: , .

[0011] In some embodiments, calculating the noise suppression gain function based on the noise power spectrum density and the ambient sound power spectrum density comprises the following steps: The noise suppression gain function is calculated by the following formula, .

[0012] In some embodiments, obtaining a training voiceprint signal and extracting casting voiceprint features in the training voiceprint signal based on the noise suppression gain function include the following steps: Performing Fourier transform on the training voiceprint signal and calculating its corresponding training power spectrum density; The training power spectrum density is multiplied by the noise suppression gain function to obtain the casting voiceprint feature corresponding to the training voiceprint signal.

[0013] In some embodiments, classifying the casting soundprint features to obtain non-aluminum leakage feature labels and aluminum leakage feature labels includes the following steps: Set the number of label categories to 2 based on the MobileNet2 algorithm; The non-aluminum leakage feature label corresponding to the casting voiceprint feature without aluminum leakage is set to 0; The aluminum leakage feature label corresponding to the casting voiceprint feature where aluminum leakage occurs is set to 1.

[0014] In a second aspect, the present application provides a method for detecting aluminum leakage, which adopts the following technical solution: A method for detecting aluminum leakage is implemented based on the aluminum leakage determination model generated in the above method, and specifically comprises the following steps: Collect real-time voiceprint signals in the casting state; Calculating the real-time power spectrum density corresponding to the real-time voiceprint signal and multiplying it with the noise suppression gain function to obtain the real-time voiceprint feature corresponding to the real-time voiceprint signal; Inputting the real-time voiceprint feature into the aluminum leakage determination model and waiting for output of the determination result; If the determination result corresponds to the non-aluminum leakage characteristic label, it is determined that no aluminum leakage occurs; If the determination result corresponds to the aluminum leakage characteristic tag, it is determined to be suspected aluminum leakage, and after the output result is maintained for a preset time, the suspected aluminum leakage is converted into confirmed aluminum leakage and an alarm message is output.

[0015] In a third aspect, the present application provides an aluminum leakage detection device, which adopts the following technical solution: An aluminum leakage detection device, comprising: Hydrophone, used to collect real-time soundprint signals in the casting state; A voiceprint feature calculation module, used to calculate the real-time power spectrum density corresponding to the real-time voiceprint signal and multiply it by the noise suppression gain function to obtain the real-time voiceprint feature corresponding to the real-time voiceprint signal; A deep learning module, used to generate and train an aluminum leakage determination model, and also used to obtain the input real-time voiceprint features and output a determination result; The aluminum leakage determination module is used to detect aluminum leakage based on the determination result. Specifically, if the determination result corresponds to a non-aluminum leakage feature label, it is determined that no aluminum leakage occurs; if the determination result corresponds to an aluminum leakage feature label, it is determined to be suspected aluminum leakage, and after the output result is maintained for a preset time, the suspected aluminum leakage is converted into confirmed aluminum leakage and an alarm message is output.

[0016] In some of the embodiments, a float device is further included, which includes a float and a guide tube. The float and the hydrophone are slidably arranged in the guide tube. A hanging ring is arranged at the bottom end of the float. The cable of the hydrophone passes through the hanging ring to achieve connection, and there is a preset length of interval between the hydrophone and the float.

[0017] The technical solution provided by the embodiments of the present application has the following technical effects: The suppression gain function is calculated by using the soundprint signals detected in different aluminum leakage environments (aluminum leakage, no aluminum leakage) to estimate the noise and pure soundprints, and based on the estimation results, a large number of training soundprint signals in the training samples are calculated to extract the corresponding casting soundprint features in different aluminum leakage environments. These casting soundprint features are classified to generate aluminum leakage and no aluminum leakage sample sets, and the sample sets are used to train a model for detecting aluminum leakage. The deep learning algorithm is combined to realize the estimation and denoising of noise in a multi-noise environment to accurately and quickly detect aluminum leakage. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 It is a flowchart of the steps of a method for generating an aluminum leakage determination model provided in an embodiment of the present application.

[0019] Figure 2 It is a step flow chart of a method for detecting aluminum leakage provided in an embodiment of the present application.

[0020] Figure 3 This is a schematic diagram of module connections of an aluminum leakage detection device provided in an embodiment of the present application.

[0021] Figure 4 It is a partial structural schematic diagram of a module of an aluminum leakage detection device provided in an embodiment of the present application.

[0022] Explanation of the accompanying reference numerals: 1. hydrophone; 2. float; 3. guide tube; 4. lifting ring. DETAILED DESCRIPTION

[0023] To more clearly understand the purpose, technical solutions and advantages of the present application, the present application is described and illustrated below in conjunction with the accompanying drawings and embodiments. However, it should be understood by those of ordinary skill in the art that the present application can be implemented without these details. In some cases, in order to avoid unnecessary descriptions that make various aspects of the present application obscure, well-known methods, processes, systems, components and / or circuits that have been described at a higher level will not be described in detail. For those of ordinary skill in the art, it is obvious that various changes can be made to the embodiments disclosed in the present application, and without departing from the principles and scope of the present application, the general principles defined in the present application can be applied to other embodiments and application scenarios. Therefore, the present application is not limited to the embodiments shown, but conforms to the broadest scope consistent with the scope claimed for protection of the present application.

[0024] It should be noted that the description of these embodiments is used to help understand the present invention, but does not constitute a limitation of the present invention. In addition, the technical features involved in each embodiment of the present invention described below can be combined with each other as long as there is no conflict between them.

[0025] In the description of the present application, the description with reference to the terms "one embodiment", "some embodiments", "illustrative embodiments", "examples", "specific examples", or "some examples" means that the specific features, structures, materials, or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present application. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials, or characteristics described may be combined in any one or more embodiments or examples.

[0026] The embodiment of the present application discloses a method for generating an aluminum leakage determination model.

[0027] like Figure 1 As shown, a method for generating an aluminum leakage determination model comprises the following steps: S100, obtaining a first soundprint signal in a casting state without aluminum leakage, and calculating a noise power spectrum density corresponding to the first soundprint signal.

[0028] First, it is determined that the casting environment is normal and there is no aluminum leakage. At this time, the first soundprint signal is detected by a hydrophone. When there is no aluminum leakage, the currently detected soundprints are only the noise in the environment, including machine working noise, environmental noise, etc.

[0029] Calculate the noise power spectral density corresponding to the first voiceprint signal.

[0030] S200, obtaining a second soundprint signal indicating aluminum leakage in a casting state, and calculating the ambient sound power spectrum density corresponding to the second soundprint signal.

[0031] It is determined that the casting environment is normal casting and aluminum leakage exists. At this time, the second soundprint signal is detected by the hydrophone 1. When aluminum leakage exists, the detected soundprint contains noise and aluminum leakage sound.

[0032] Calculate the ambient sound spectrum density corresponding to the second voiceprint signal.

[0033] S300: Calculate a noise suppression gain function based on the noise power spectrum density and the ambient sound power spectrum density.

[0034] The suppression gain function is calculated by the power spectrum density of two different state environments. The suppression gain function is mainly used to estimate the noise in the voiceprint information, and the estimation result of the pure voiceprint is derived based on the noise estimation result combined with the real-time received voiceprint signal.

[0035] S400, obtaining a training voiceprint signal, and extracting casting voiceprint features in the training voiceprint signal based on a noise suppression gain function.

[0036] A certain amount of training voiceprint signals are input, which include voiceprints of aluminum leakage and voiceprints of non-aluminum leakage. The pure voiceprints in the voiceprint signals are estimated through the noise suppression gain function to extract the casting voiceprint features in the training voiceprint signals. The casting voiceprint features in different aluminum leakage scenarios are different.

[0037] S500, classifying the casting voiceprint features to obtain non-aluminum leakage feature labels and aluminum leakage feature labels, and training the casting voiceprint features based on the non-aluminum leakage feature labels and the aluminum leakage feature labels to obtain an aluminum leakage determination model.

[0038] A large number of casting voiceprint features extracted from the training voiceprint signals are used as training samples, and different classification labels are provided for different voiceprint features of aluminum leakage and non-aluminum leakage. Each label and the corresponding casting voiceprint feature are put into the model for training. The trained model is defined as an aluminum leakage judgment model, through which the voiceprint signals obtained in real time in the subsequent detection process can be input to obtain the corresponding aluminum leakage detection results.

[0039] Through the above method, the suppression gain function is calculated through the soundprint signals detected in different aluminum leakage environments (aluminum leakage, no aluminum leakage) to estimate the noise and pure soundprints, and based on the estimation results, a large number of training soundprint signals in the training samples are calculated to extract the corresponding casting soundprint features in different aluminum leakage environments, and these casting soundprint features are classified to generate aluminum leakage and no aluminum leakage sample sets, and the sample sets are used to train a model for detecting aluminum leakage. Combined with deep learning algorithms, the estimation and denoising of noise in a multi-noise environment can be realized to accurately and quickly detect aluminum leakage.

[0040] In some other embodiments, calculating the noise power spectrum density corresponding to the first voiceprint signal includes the following steps: S110, performing Fourier transform on the first voiceprint signal.

[0041] The first voiceprint information is collected when there is no aluminum leakage. Since there is no aluminum leakage, the voiceprint signals collected every second are all noise signals, that is: .

[0042] The representation is the currently collected voiceprint signal, which is the first voiceprint information in the embodiment of the present application, and It is characterized as a noise signal.

[0043] After Fourier transforming the first voiceprint signal, we get: .

[0044] is the frequency domain signal of the currently collected voiceprint signal, It represents the frequency domain signal of the noise signal, and w is the frame number corresponding to the first voiceprint signal (noise signal).

[0045] S120, calculating the power spectrum density of the first voiceprint signal after Fourier transformation and performing smoothing processing to obtain a theoretical noise power spectrum density.

[0046] The specific formula is as follows: , in, It is represented by the theoretical noise power spectrum density after smoothing, a is represented by the smoothing coefficient, It is represented by the power spectrum density of the frequency domain signal of the first voiceprint signal, and w is represented by the number of frames of the first voiceprint signal.

[0047] Because when there is no aluminum leakage, the first soundprint signal is a noise signal. = .

[0048] In some other embodiments, calculating the noise power spectrum density corresponding to the first voiceprint signal further includes the following steps: S130, setting the casting state to multi-furnace casting, and collecting the first voiceprint signal of a preset time.

[0049] After the theoretical noise power spectrum density is calculated, it is the power spectrum density corresponding to the first voiceprint signal pair under ideal conditions. In order to improve the accuracy of subsequent training and calculation, it is also necessary to calculate the mean of the noise power spectrum density corresponding to multi-furnace casting.

[0050] In the embodiment of the present application, the multi-furnace casting specifically includes 10 furnaces.

[0051] S140, calculating a mean value of the smoothed noise power spectrum density per second in a preset time based on the theoretical noise power spectrum density, and using the mean value of the noise power spectrum density as the noise power spectrum density.

[0052] The specific formula is as follows: , in, It is represented by the mean value of the noise power spectrum density, and V is represented by the length of the preset time.

[0053] The calculated mean value of noise power spectrum density is calculated on the basis of relatively ideal theoretical noise power spectrum density combined with the actual casting environment, so its accuracy and universality are relatively good. Therefore, the mean value of noise power spectrum density is taken as the final noise power spectrum density.

[0054] In some other embodiments, calculating the ambient sound power spectrum density corresponding to the second voiceprint signal includes the following steps: S210, performing Fourier transform on the second voiceprint signal to obtain: , in, It is characterized by the frequency domain signal corresponding to the second voiceprint signal. It is characterized by the frequency domain signal corresponding to the aluminum leakage soundprint signal. The frequency domain signal characterized as the first voiceprint signal.

[0055] When an aluminum leakage event occurs, the collected second voiceprint signal contains the aluminum leakage voiceprint and the noise signal, that is: .

[0056] Perform Fourier transform on the second voiceprint signal to obtain the above formula.

[0057] S220, calculating the ambient sound power spectrum density corresponding to the second voiceprint signal based on the following formula: , .

[0058] Since the aluminum leakage soundprint and the noise signal are two independent events, the estimated power spectrum density of the aluminum leakage signal is actually the power spectrum density corresponding to the currently collected second voiceprint signal minus the power spectrum density of the noise signal.

[0059] In this way, the value of the pure signal (leaking aluminum sound pattern) can be obtained.

[0060] In some other embodiments, calculating the noise suppression gain function based on the noise power spectrum density and the ambient sound power spectrum density includes the following steps: S310, calculating the noise suppression gain function by the following formula: .

[0061] In some other embodiments, obtaining a training voiceprint signal and extracting casting voiceprint features in the training voiceprint information based on a noise suppression gain function comprises the following steps: S410, performing Fourier transform on the training voiceprint signal and calculating its corresponding training power spectrum density.

[0062] S420, multiplying the training power spectrum density by the noise suppression gain function to obtain a casting voiceprint feature corresponding to the training voiceprint signal.

[0063] After the noise suppression gain function in the casting process is calculated, it is only necessary to multiply the power spectrum density corresponding to the pre-collected training voiceprint signal by the noise suppression gain function to obtain the casting voiceprint characteristics of the casting state. The casting voiceprint characteristics are mainly used to display the characteristic representation of aluminum leakage events and non-aluminum leakage events in a voiceprint signal.

[0064] The specific formula is: .

[0065] Y represents the casting sound pattern feature, It is characterized by the training power spectrum density corresponding to the training voiceprint signal.

[0066] When aluminum leakage does not occur, = = , so is equal to 0, and when aluminum leakage occurs, It is not equal to 0, so whether aluminum leakage occurs can be determined based on the value of the casting soundprint feature.

[0067] In some other embodiments, classifying casting soundprint features to obtain non-aluminum leakage feature labels and aluminum leakage feature labels includes the following steps: S510, based on the MobileNet2 algorithm, sets the number of label categories to 2.

[0068] S520, setting the non-aluminum leakage feature label corresponding to the casting voiceprint feature without aluminum leakage to 0.

[0069] S530, setting the aluminum leakage feature label corresponding to the casting voiceprint feature where aluminum leakage occurs to 1.

[0070] After obtaining the feature labels of non-aluminum leakage and aluminum leakage, the casting voiceprint features of each casting state are trained through the MobileNet2 algorithm to obtain the aluminum leakage judgment model.

[0071] like Figure 2 As shown, an aluminum leakage detection method is also disclosed in an embodiment of the present application, which is implemented based on the aluminum leakage determination model generated by the above method, and specifically includes the following steps: S600 collects real-time voiceprint signals in the casting state.

[0072] S610, calculating the real-time power spectrum density corresponding to the real-time voiceprint signal and multiplying it with the noise suppression gain function to obtain the real-time voiceprint feature corresponding to the real-time voiceprint signal.

[0073] S620, input the real-time voiceprint feature into the aluminum leakage determination model and wait for the output of the determination result.

[0074] S630: If the determination result corresponds to the non-aluminum leakage characteristic label, it is determined that no aluminum leakage occurs.

[0075] S640: If the determination result corresponds to the aluminum leakage feature tag, it is determined to be suspected aluminum leakage, and after the output result is maintained for a preset time, the suspected aluminum leakage is converted into confirmed aluminum leakage and an alarm message is output.

[0076] During the operation of the casting system, it is only necessary to collect real-time voiceprint signals every second (which may or may not be leaking aluminum), perform Fourier transform and power spectrum density calculation on the real-time voiceprint signals, and multiply the calculated real-time power spectrum density with the noise suppression gain function to extract the voiceprint features corresponding to the real-time voiceprint signals.

[0077] The extracted real-time voiceprint features are input into the aluminum leakage judgment model for judgment. If the output result is 0, it corresponds to the feature label of no aluminum leakage, and it is judged that no aluminum leakage occurs; if the output result is 1, it corresponds to the feature label of aluminum leakage, and the judgment result is set to suspected aluminum leakage. In order to reduce the influence of detection errors and other situations on the results, it is necessary to judge whether the result is maintained for a certain time. When it is maintained for a certain time, it can be determined that aluminum leakage occurs, and an alarm is issued.

[0078] In other embodiments, when detecting aluminum leakage, the following steps may also be included: After the system is running, the visual feature acquisition module collects video information in the working environment. The visual feature acquisition module is generally a camera.

[0079] Different visual monitoring target areas are set, and the state of the aluminum processing working environment is determined according to the specific state of the workers and workpieces in the visual monitoring target areas and combined with the corresponding visual processing algorithms.

[0080] Specifically, the indicator light ROI area is set. The ROI area is also called the region of interest. In machine vision and image processing, it is an area to be processed that is manually set and outlined in the processed image in the form of a box, circle, ellipse, irregular polygon, etc.

[0081] In the embodiment of the present application, after the specific position of the visual feature acquisition module is determined, the image area collected by it is obtained, and the corresponding indicator light ROI area is set according to the area where the equipment indicator light is located in the video image. Because the casting equipment is large-scale equipment, the position will not move for a long time, so after the indicator light ROI area is set, it does not need to be revised or changed in general scenarios.

[0082] The video frame is obtained from the video surveillance stream collected by the visual feature acquisition module, and the video frame is cropped. The cropped size is the size corresponding to the indicator light ROI area. After cropping, a picture containing the indicator light is obtained. The indicator light picture is classified into two categories using the MobileNetV2 algorithm. If the classification result is "light on", it indicates that the indicator light of the device is currently in the on state. If the classification result is "light off", it indicates that the indicator light of the device is currently in the off state.

[0083] The current casting state is determined based on the lighting state, and the casting state is used as the visual monitoring result. The casting state is characterized by whether the current monitoring environment is in the casting working state, and the subsequent visual monitoring and sound monitoring are combined to determine the aluminum leakage based on the current casting state. If it is not in the casting state, there is no need to monitor the aluminum leakage.

[0084] The working area to be inspected corresponding to the working environment is artificially selected, generally the area around the casting plate, which is defined as the working range of the casting staff and the monitoring range of the personnel for aluminum leakage.

[0085] The number of categories of the YOLOv5 algorithm is set to 2 to monitor pedestrians (staff) and helmets in the video frame, so as to obtain the first detection frame person_bbox containing the staff and the second detection frame helmet_bbox containing the helmet.

[0086] Calculate the first center point corresponding to the first detection frame and the second center point corresponding to the second detection frame. First, determine whether there is a first center point in the working area to be detected. If so, it means that there are workers in the working range of the casting. Then detect the number of first center points in the working area to be detected, which represents the number of workers around the casting plate.

[0087] Set the number of staff in the work area to be detected according to the requirements. For example, if it is set to 2, then when there are more than two first center points in the work area to be detected, it is considered that the work scope is not in an off-duty state. When the number of first center points in the work area to be detected is less than 2, it is considered that there is an off-duty state, and an alarm is required.

[0088] Secondly, when there is a second center point in the first detection frame, it is considered that the staff is wearing a safety helmet when working. Conversely, if there is no second center point in the first detection frame, it is considered that the staff is not wearing a safety helmet when working, and an alarm signal needs to be issued.

[0089] The above monitoring results are used to monitor the behavior of the staff. When no alarm signal appears in the staff's behavior, it is considered that there is no safety hazard of aluminum leakage that cannot be monitored by the staff in time, and aluminum leakage monitoring based on sound monitoring can be carried out.

[0090] like Figure 3 As shown, an aluminum leakage detection device is also disclosed in the embodiment of the present application, including: The hydrophone 1 is used to collect real-time soundprint signals in the casting state.

[0091] The voiceprint feature calculation module is used to calculate the real-time power spectrum density corresponding to the real-time voiceprint signal and multiply it with the noise suppression gain function to obtain the real-time voiceprint feature corresponding to the real-time voiceprint signal.

[0092] The deep learning module is used to generate and train the aluminum leakage determination model, and is also used to obtain the real-time voiceprint features of the input and output the determination results.

[0093] The aluminum leakage determination module is used to detect aluminum leakage based on the determination result. Specifically, if the determination result corresponds to the non-aluminum leakage feature label, it is determined that no aluminum leakage occurs; if the determination result corresponds to the aluminum leakage feature label, it is determined to be suspected aluminum leakage, and after the output result is maintained for a preset time, the suspected aluminum leakage is converted into confirmed aluminum leakage and an alarm message is output.

[0094] like Figure 4As shown, in other embodiments, a float device is also included, which includes a float 2 and a guide tube 3. The float 2 and the hydrophone 1 are slidably arranged in the guide tube 3. A hanging ring 4 is arranged at the bottom end of the float 2. The cable of the hydrophone 1 passes through the hanging ring 4 to achieve connection, and there is a gap of a preset length between the hydrophone 1 and the float 2.

[0095] Since the distance between the hydrophone 1 and the water surface is different, the size and value of the soundprint signal detected by it are different. In order to reduce the error in the soundprint detection result caused by the change in distance, the hydrophone 1 needs to maintain a certain distance from the water surface.

[0096] The float 2 is a stainless steel hollow float with an outer diameter of 45 mm, a length of 150 mm and a load of 15 kg. A circular lifting ring 4 is fixed to its bottom.

[0097] The guide tube 3 is a hollow long tube made of stainless steel. The float 2 and the hydrophone 1 both slide inside the guide tube 3 .

[0098] The signal transmission line of the hydrophone 1 passes through the lifting ring 4 to be connected to the float 2. The signal transmission line between the float 2 and the hydrophone 1 is wrapped with a coating material of a certain hardness. The length of the coating material is 1m, so that the interval of 1m between the hydrophone 1 and the float 2 can be ensured. The float 2 floats on the water surface by buoyancy, so that the hydrophone 1 can always maintain a distance of 1m from the water surface, ensuring that the hydrophone 1 can stably collect the soundprint information during the casting process. At the same time, the coating material can also protect the signal transmission line and fix the hydrophone 1 at the same time, that is, when the hydrophone 1 moves up and down with the water level change, it will not move left and right in the guide tube 3.

[0099] The implementation principle is: The suppression gain function is calculated by using the soundprint signals detected in different aluminum leakage environments (aluminum leakage, no aluminum leakage) to estimate the noise and pure soundprints, and based on the estimation results, a large number of training soundprint signals in the training samples are calculated to extract the corresponding casting soundprint features in different aluminum leakage environments. These casting soundprint features are classified to generate aluminum leakage and no aluminum leakage sample sets, and the sample sets are used to train a model for detecting aluminum leakage. The deep learning algorithm is combined to realize the estimation and denoising of noise in a multi-noise environment to accurately and quickly detect aluminum leakage.

[0100] It should be understood that although the steps in the flowchart of the accompanying drawings are shown in sequence according to the instructions of the arrows, these steps are not necessarily performed in the order indicated by the arrows. Unless otherwise clearly stated in this document, the execution of these steps is not strictly limited in order and can be performed in other orders.

[0101] The above are all preferred embodiments of the present application, and the protection scope of the present application is not limited thereto. Therefore, any equivalent changes made according to the structure, shape, and principle of the present application should be included in the protection scope of the present application.

Claims

1. A method for generating an aluminum leakage determination model, characterized in that: The following steps are involved: Acquire a first soundprint signal in a casting state without aluminum leakage, and calculate the noise power spectrum density corresponding to the first soundprint signal; Acquire a second soundprint signal of aluminum leakage in a casting state, and calculate the ambient sound power spectrum density corresponding to the second soundprint signal; Calculating a noise suppression gain function based on the noise power spectrum density and the ambient sound power spectrum density; Acquire a training voiceprint signal, and extract a casting voiceprint feature in the training voiceprint signal based on the noise suppression gain function; The casting voiceprint features are classified to obtain feature labels of no aluminum leakage and feature labels of aluminum leakage, and the casting voiceprint features are trained based on the feature labels of no aluminum leakage and the feature labels of aluminum leakage to obtain an aluminum leakage determination model.

2. The method for generating an aluminum leakage determination model according to claim 1, characterized in that: Calculating the noise power spectrum density corresponding to the first voiceprint signal includes the following steps: Performing Fourier transform on the first voiceprint signal; The power spectrum density of the first voiceprint signal after Fourier transformation is calculated and smoothed to obtain the theoretical noise power spectrum density. The specific formula is as follows: O(w)=a|q(w-1)| 2 +(1-a)|q(w)| 2 , Among them, O(w) represents the theoretical noise power spectrum density after smoothing, a represents the smoothing coefficient, |q(w)| 2 It is represented by the power spectrum density of the frequency domain signal of the first voiceprint signal, and w is represented by the number of frames of the first voiceprint signal.

3. The method for generating an aluminum leakage determination model according to claim 2, characterized in that: Calculating the noise power spectrum density corresponding to the first voiceprint signal also includes the following steps: The casting state is set to multi-furnace casting, and the first voiceprint signal of a preset time is collected; Based on the theoretical noise power spectrum density, the mean value of the smoothed noise power spectrum density per second in the preset time is calculated, and the mean value of the noise power spectrum density is used as the noise power spectrum density. The specific formula is as follows: in, It is represented by the mean value of the noise power spectrum density, and V is represented by the length of the preset time.

4. The method for generating an aluminum leakage determination model according to claim 3, characterized in that: Calculating the ambient sound power spectrum density corresponding to the second voiceprint signal includes the following steps: Perform Fourier transform on the second voiceprint signal to obtain: y(w)=s(w)+q(w), Among them, y(w) represents the frequency domain signal corresponding to the second voiceprint signal, s(w) represents the frequency domain signal corresponding to the aluminum leakage voiceprint signal, and q(w) represents the frequency domain signal of the first voiceprint signal; The ambient sound power spectrum density corresponding to the second voiceprint signal is calculated based on the following formula: |y(w)| 2 =|s(w)| 2 +|q(w)| 2 , 5. The method for generating an aluminum leakage determination model according to claim 4, characterized in that: Calculating a noise suppression gain function based on the noise power spectrum density and the ambient sound power spectrum density comprises the following steps: The noise suppression gain function is calculated by the following formula, 6. The method for generating an aluminum leakage determination model according to claim 5, characterized in that: Acquiring a training voiceprint signal and extracting a casting voiceprint feature in the training voiceprint signal based on the noise suppression gain function comprises the following steps: Performing Fourier transform on the training voiceprint signal and calculating its corresponding training power spectrum density; The training power spectrum density is multiplied by the noise suppression gain function to obtain the casting voiceprint feature corresponding to the training voiceprint signal.

7. The method for generating an aluminum leakage determination model according to claim 1, characterized in that: Classifying the casting soundprint features to obtain non-aluminum leakage feature labels and aluminum leakage feature labels includes the following steps: Based on the MobileNet2 algorithm, set the number of label categories to 2; The non-aluminum leakage feature label corresponding to the casting voiceprint feature without aluminum leakage is set to 0; The aluminum leakage feature label corresponding to the casting voiceprint feature where aluminum leakage occurs is set to 1.

8. A method for detecting aluminum leakage, characterized in that: The aluminum leakage determination model generated according to the method described in any one of claims 1 to 7 is implemented, specifically comprising the following steps: Collect real-time voiceprint signals in the casting state; Calculating the real-time power spectrum density corresponding to the real-time voiceprint signal and multiplying it with the noise suppression gain function to obtain the real-time voiceprint feature corresponding to the real-time voiceprint signal; Inputting the real-time voiceprint feature into the aluminum leakage determination model and waiting for output of the determination result; If the determination result corresponds to the non-aluminum leakage characteristic label, it is determined that no aluminum leakage occurs; If the determination result corresponds to the aluminum leakage characteristic tag, it is determined to be suspected aluminum leakage, and after the output result is maintained for a preset time, the suspected aluminum leakage is converted into confirmed aluminum leakage and an alarm message is output.

9. An aluminum leakage detection device, characterized in that: include: A hydrophone (1), used for collecting real-time soundprint signals in a casting state; A voiceprint feature calculation module, used to calculate the real-time power spectrum density corresponding to the real-time voiceprint signal and multiply it by the noise suppression gain function to obtain the real-time voiceprint feature corresponding to the real-time voiceprint signal; A deep learning module, used to generate and train an aluminum leakage determination model, and also used to obtain the input real-time voiceprint features and output a determination result; The aluminum leakage determination module is used to detect aluminum leakage based on the determination result. Specifically, if the determination result corresponds to a non-aluminum leakage feature label, it is determined that no aluminum leakage occurs; if the determination result corresponds to an aluminum leakage feature label, it is determined to be suspected aluminum leakage, and after the output result is maintained for a preset time, the suspected aluminum leakage is converted into confirmed aluminum leakage and an alarm message is output.

10. The aluminum leakage detection device according to claim 9, characterized in that: The invention also comprises a float device, the float device comprising a float (2) and a guide tube (3), the float (2) and the hydrophone (1) being slidably arranged in the guide tube (3), a hanging ring (4) being arranged at the bottom end of the float (2), a cable of the hydrophone (1) passing through the hanging ring (4) to achieve connection, and a preset length of interval being present between the hydrophone (1) and the float (2).

Images