Method for detecting hydrogen content in copper melt based on ultrasonic wave dehydrogenation

Abstract

This invention relates to the field of hydrogen content detection technology in copper melts, and particularly to a method for detecting hydrogen content in copper melts based on ultrasonic dehydrogenation. The method includes: simultaneously acquiring acoustic response signals and spatial temperature distribution signals of the copper melt; extracting acoustic feature parameters related to hydrogen content based on the acoustic response signals, and dividing the melt space into multiple response regions based on the spatial temperature distribution signals; inputting the acoustic feature parameters and spatial information of the response regions into a pre-constructed mapping model to obtain the spatiotemporal distribution information of hydrogen content; extracting temporal evolution features and spatial change features based on the spatiotemporal distribution information; adjusting the acquisition parameters of the acoustic response signals and the sampling weights of each response region according to the temporal evolution features and spatial change features, and performing signal acquisition and updating of the spatiotemporal distribution information of hydrogen content in the next cycle. This invention enables dynamic and accurate detection of hydrogen content in copper melts, adapting to the spatiotemporal dynamic changes of hydrogen content during the dehydrogenation process.

Description

Technical Field

[0001] This invention relates to the technical field of hydrogen content detection in copper melt, and more particularly to a method for detecting hydrogen content in copper melt based on ultrasonic dehydrogenation. Background Technology

[0002] In the smelting and production of copper and copper alloys, the solubility of hydrogen in the copper melt decreases sharply with decreasing temperature. If the hydrogen content in the melt is too high during smelting, it will precipitate during the subsequent solidification of the ingot, forming defects such as pores and porosity, directly deteriorating the electrical conductivity, mechanical properties, and processing performance of the copper material. Ultrasonic dehydrogenation technology, with its advantages of being green and additive-free and having high dehydrogenation efficiency, is widely used in the melt refining of high-purity oxygen-free copper and high-end copper alloys. Its principle is that the cavitation effect generated by the propagation of high-energy ultrasound in the copper melt causes dissolved hydrogen to diffuse into cavitation bubbles and then float to the surface and escape.

[0003] The most commonly used method in the industry for detecting hydrogen content in copper melt is the reduced pressure solidification detection method. This method involves taking a sample from the smelting furnace and placing it in a low-pressure, closed environment for cooling and solidification. Pressure sensors or thermal conductivity detectors are used to collect pressure changes or gas concentration signals during the hydrogen evolution process, and the hydrogen content is calculated using a thermodynamic mathematical model.

[0004] Existing hydrogen content detection methods rely on fixed settings before operation for determining the detection frequency and sampling locations. However, the rate of change and spatial distribution of hydrogen content in the copper melt vary significantly at different stages of the dehydrogenation process. Fixed detection settings are insufficient to ensure timely detection during critical stages of rapid hydrogen content changes, while unnecessary repeated detection operations are likely to occur during stages of relatively stable hydrogen content. Summary of the Invention

[0005] This invention provides a method for detecting the hydrogen content of copper melt based on ultrasonic dehydrogenation, which can effectively solve the problems in the background art.

[0006] To achieve the above objectives, the technical solution adopted by the present invention is as follows: Methods for detecting hydrogen content in copper melts based on ultrasonic dehydrogenation include: Simultaneously acquire acoustic response signals and spatial temperature distribution signals of the copper melt; Acoustic characteristic parameters related to hydrogen content are extracted based on acoustic response signals, and the melt space is divided into multiple response regions based on spatial temperature distribution signals. The acoustic feature parameters and the spatial information of the response region are input into a pre-constructed mapping model to obtain the spatiotemporal distribution information of hydrogen content; Extracting temporal evolution features and spatial change features based on spatiotemporal distribution information; Based on the temporal evolution and spatial variation characteristics, the acquisition parameters of the acoustic response signal and the sampling weights of each response region are adjusted, and the signal acquisition and spatiotemporal distribution information of hydrogen content are updated for the next cycle.

[0007] Furthermore, the spatiotemporal distribution information of hydrogen content includes the temporal evolution curve and spatial distribution map of hydrogen content; Among them, the time evolution curve represents the trend of hydrogen content change over time, and the spatial distribution map represents the distribution of hydrogen content at various spatial locations in the melt.

[0008] Furthermore, temporal evolution features are extracted, including: Based on the time evolution curve, calculate at least one of the hydrogen content change rate or hydrogen content change acceleration as a time evolution characteristic.

[0009] Furthermore, spatial variation features are extracted, including: Based on the spatial distribution map, the hydrogen content gradient values ​​at different spatial locations are calculated, and the location boundaries where the hydrogen content gradient values ​​exceed a preset gradient threshold are identified as spatial variation characteristics.

[0010] Furthermore, the acoustic response signal is the time-domain waveform signal received after the ultrasonic wave propagates in the copper melt; acoustic characteristic parameters related to hydrogen content are extracted, including: Time-frequency analysis is performed on the time-domain waveform signal to extract at least one of the ultrasonic fundamental attenuation coefficient sequence and the nonlinear harmonic energy ratio sequence as acoustic characteristic parameters.

[0011] Furthermore, the ultrasonic fundamental frequency attenuation coefficient is determined based on the ratio of the ultrasonic transmission frequency to the fundamental frequency amplitude of the received signal; the nonlinear harmonic energy ratio is determined based on the ratio of the second harmonic energy to the fundamental frequency energy in the received signal.

[0012] Furthermore, the spatial temperature distribution signal is acquired by thermocouples or infrared temperature measuring devices installed at multiple preset locations within the melting furnace; the molten space is dynamically divided into multiple response regions, including: A three-dimensional temperature field distribution of copper melt is constructed based on the spatial temperature distribution signal. Based on the comparison between the gradient change of the three-dimensional temperature field distribution and the preset temperature gradient threshold, the melt space is dynamically divided into multiple thermodynamic response zones, which serve as the response regions.

[0013] Furthermore, the mapping model is a neural network model pre-constructed based on the mapping relationship between historical acoustic feature parameters, historical response region spatial information and corresponding measured hydrogen content values.

[0014] Furthermore, adjust the acquisition parameters of the acoustic response signal, including: Adjust the acquisition window length or sampling frequency of the acoustic response signal according to the magnitude of the change in the time evolution characteristics; when the magnitude of the change in the time evolution characteristics increases, shorten the acquisition window length or increase the sampling frequency.

[0015] Furthermore, the sampling weights of each response region are adjusted, including: Based on the distribution of spatial variation features, increase the sampling weight of response regions where spatial variation features exceed a preset spatial feature threshold, and / or decrease the sampling weight of response regions where spatial variation features are below a preset spatial feature threshold. The sampling weights are used to adjust the acquisition density of acoustic response signals in each response region, or to adjust the input contribution of the acoustic feature parameters corresponding to the response region in the mapping model.

[0016] The technical solution of this invention can achieve the following technical effects: By simultaneously acquiring acoustic response signals and spatial temperature distribution signals from molten copper, and constructing a mapping model based on acoustic feature parameters and response region division to obtain spatiotemporal distribution information of hydrogen content, the acquisition parameters and sampling weights are dynamically adjusted through extracted temporal evolution features and spatial variation features, forming a complete closed-loop detection process. The combination of temporal evolution features and spatial variation features allows the adjustment of acquisition parameters to match the temporal change rate of hydrogen content, and the adjustment of sampling weights to adapt to the spatial distribution differences of hydrogen content, achieving real-time optimization of the detection strategy. At the same time, relying on acoustic feature parameters strongly correlated with hydrogen content and dynamically divided response regions, the targeting of hydrogen content detection is improved, making it suitable for ultrasonic dehydrogenation refining scenarios of high-purity oxygen-free copper and high-end copper alloys, providing reliable support for precise control of hydrogen content in molten copper.

[0017] The above description is only an overview of the technical solution of this application. In order to better understand the technical means of this application and to implement it in accordance with the contents of the specification, and to make the above and other objects, features and advantages of this application more obvious and understandable, specific embodiments of this application are given below. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 This is a schematic flowchart of the method for detecting hydrogen content in copper melt based on ultrasonic dehydrogenation according to the present invention. Detailed Implementation

[0020] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0021] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. The terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0022] like Figure 1 As shown, the method for detecting hydrogen content in copper melt based on ultrasonic dehydrogenation of the present invention is applied to ultrasonic dehydrogenation of copper melt and specifically includes the following steps: Step S100: Synchronously acquire the acoustic response signal and spatial temperature distribution signal of the copper melt; Step S200: Extract acoustic feature parameters related to hydrogen content based on acoustic response signals, and divide the melt space into multiple response regions based on spatial temperature distribution signals; Step S300: Input the acoustic feature parameters and the spatial information of the response region into the pre-constructed mapping model to obtain the spatiotemporal distribution information of hydrogen content; Step S400: Extract temporal evolution features and spatial change features based on spatiotemporal distribution information; Step S500: Based on the temporal evolution characteristics and spatial variation characteristics, adjust the acquisition parameters of the acoustic response signal and the sampling weight of each response region, and execute the signal acquisition and hydrogen content spatiotemporal distribution information update for the next cycle.

[0023] In this embodiment, by using the synchronously acquired acoustic response signal and spatial temperature distribution signal as the basis, combined with acoustic feature parameter extraction, melt spatial response region division, and hydrogen content spatiotemporal distribution information acquisition, and relying on the extracted time evolution characteristics and spatial change characteristics, the acquisition parameters and sampling weights are dynamically adjusted and a closed-loop update is formed. This allows the acquisition resources to be tilted towards the key stages of rapid hydrogen content change and key areas with uneven spatial distribution, adapting to the differences in the rate of change of hydrogen content over time and the differences in spatial distribution gradient during the dehydrogenation process, and achieving real-time matching between the detection strategy and the dynamic changes in hydrogen content.

[0024] In a specific implementation, as one example, given the dynamic differences in the rate of change and spatial distribution of hydrogen content during the ultrasonic dehydrogenation process of copper melt, it is necessary to collect multi-dimensional signals related to hydrogen content in order to comprehensively capture the melt state. Hydrogen bubbles will change the propagation characteristics of ultrasonic waves, and the time-domain waveform signal can reflect this change and thus correlate it with hydrogen content. Temperature distribution directly determines hydrogen solubility, and temperature differences in different spaces lead to uneven hydrogen content, while signal asynchrony can also cause errors in hydrogen content calculation. This embodiment achieves synchronous acquisition of the acoustic response signal and spatial temperature distribution signal of copper melt by deploying suitable acquisition devices, setting reasonable acquisition parameters, and implementing a synchronous triggering mechanism, as detailed below: Step S110: Install ultrasonic transmitting transducers and ultrasonic receiving transducers symmetrically at preset positions on the furnace wall of the copper melt smelting furnace. The placement positions are selected where the melt flow is relatively stable and can cover the core area of ​​the melt to reduce the interference of flow disturbance on the propagation of ultrasonic waves. The ultrasonic transmitting transducers are made of piezoelectric ceramic material, which utilizes its stable piezoelectric effect to continuously emit high-energy ultrasonic waves. It is suitable for the high-temperature environment of copper melt and is not prone to performance degradation, meeting the requirements for long-term continuous acquisition. The ultrasonic receiving transducers are matched with the transmitting transducers to ensure accurate reception of ultrasonic signals propagated through the copper melt and conversion into electrical signals. The sampling frequency selection must be sufficient to fully capture the fundamental wave of the ultrasonic wave and any possible second harmonic signals, and avoid signal distortion or redundancy. The acquisition window length is determined based on the propagation distance and speed of the ultrasonic wave in the copper melt. The propagation distance is determined by the furnace size and the transducer placement, and the propagation speed is determined by the temperature and composition of the copper melt. This ensures that the complete waveform of one ultrasonic wave propagation is received, and avoids waveform truncation affecting the extraction of acoustic feature parameters. Step S120: Install thermocouples or infrared temperature measuring devices at multiple preset locations within the smelting furnace. These locations cover the upper, middle, and lower parts of the furnace body, as well as the center of the melt. The number of thermocouples is determined by the furnace volume; a larger volume requires more locations to ensure accurate capture of temperature gradients between adjacent points and avoid distortion of the temperature distribution signal. The thermocouples are contact-type temperature measuring devices, inserted into the copper melt through preset mounting holes in the furnace wall. The insertion depth is determined by the furnace dimensions and the height of the melt, ensuring the measuring end is completely immersed in the melt and does not contact the furnace wall or bottom, preventing interference from furnace wall temperature on the measurement results. The infrared temperature measuring device is a non-contact temperature measuring device, fixed to a preset bracket on the furnace wall. The temperature probe is aligned with a preset area on the surface of the copper melt, and the probe angle is adjusted to ensure coverage of the preset surface temperature measuring points, avoiding interference from high-temperature radiation from the furnace wall. The combination of these two methods achieves comprehensive temperature acquisition of the entire copper melt space, avoiding the coverage limitations of a single temperature measuring device. The acquisition frequency is kept consistent with the acoustic response signal acquisition period to ensure synchronous correspondence between the two signals; the temperature measurement accuracy is determined according to the hydrogen content detection requirements, and the accuracy of the infrared temperature measurement device is matched with the thermocouple to ensure the consistency of the temperature distribution signal; Step S130: Select a pulse generator as the synchronous trigger control module. The pulse generator generates a synchronous pulse signal, such as a square wave pulse, that matches the acquisition cycle. The parameters of the pulse generator must meet the acquisition requirements. The pulse width is set to a reasonable range that ensures stable response of the acquisition components. The pulse amplitude is matched to the trigger thresholds of the ultrasonic transmitting transducer, thermocouple, and infrared temperature measuring device to ensure reliable and effective trigger signals. The synchronous pulse signals output by the pulse generator are respectively assigned to ultrasonic transmitting and receiving actions and temperature acquisition actions. The pulse generator continuously sends synchronous pulses according to the preset acquisition cycle. Each time a pulse is sent, the trigger signal of the pulse generator acts synchronously on each acquisition component. The ultrasonic transmitting transducer responds immediately and emits ultrasonic waves. The ultrasonic receiving transducer synchronously starts receiving action, receiving the ultrasonic signal propagated through the copper melt in real time and converting it into a time-domain waveform signal. At the same time, the thermocouples and infrared temperature measuring devices at each preset position synchronously start temperature acquisition action to obtain the copper melt temperature data at the corresponding position. After the two signals are acquired, the time-domain waveform signals within the same pulse trigger cycle are bound to the temperature data at each position at the corresponding time to achieve spatiotemporal correspondence between the two signals.

[0025] In this embodiment, considering the dynamic changes in hydrogen content in the copper melt during ultrasonic dehydrogenation, acoustic response signals and spatial temperature distribution signals directly related to hydrogen content are simultaneously acquired. A multi-location, multi-device acquisition method adapted to the melt is adopted to achieve signal coverage of the entire melt space and all time periods. Furthermore, a synchronous triggering mechanism ensures the spatiotemporal correspondence of the two signals, providing a basis for the detection of the spatiotemporal distribution of hydrogen content and the dynamic adjustment of acquisition parameters.

[0026] In some embodiments of the present invention, existing methods for processing acoustic response signals of molten copper and dividing response regions typically employ static homogenization. For example, a fixed time window is used to directly calculate the fundamental or harmonic amplitude. However, during the dehydrogenation process of molten copper, the cavitation bubble density and hydrogen atom concentration along the ultrasonic propagation path continuously change. The fixed time window cannot adapt to the non-stationary characteristics of the acoustic response signal, and the calculation results of the fundamental attenuation coefficient and harmonic energy ratio fluctuate due to the start and end positions of the window, resulting in significant deviations in repeated measurements. Based on fixed geometric grid partitioning, the fixed geometric partitioning cannot reflect the true temperature gradient boundaries formed by local heating due to cavitation effects and heat dissipation from the furnace wall during the dehydrogenation process. After merging molten regions with significant temperature differences into the same partition, the correlation between acoustic characteristic parameters and hydrogen content within this partition is averaged, and the mapping model cannot distinguish the differences in hydrogen solubility at different temperatures, leading to a significant amplification of estimation errors at locations with large temperature gradients. This embodiment extracts the fundamental attenuation coefficient sequence and the nonlinear harmonic energy ratio sequence through time-frequency decomposition and combines it with the three-dimensional temperature field gradient to divide the thermodynamic response region, achieving non-stationary adaptation of acoustic characteristic parameters and temperature-adaptive partitioning of the molten space. Specifically, the following operations are performed: Step S210: Denoise the time-domain waveform signal to remove noise such as furnace vibration and electromagnetic interference, and retain the effective signal related to the ultrasonic propagation characteristics. Step S220: Perform time-frequency analysis on the processed time-domain waveform signal to convert the time-domain signal into a two-dimensional time-frequency distribution, so as to present the energy change law of signals at different frequencies; Step S230: Extract at least one of the ultrasonic fundamental wave attenuation coefficient sequence and the nonlinear harmonic energy ratio sequence as acoustic characteristic parameters. The ultrasonic fundamental wave attenuation coefficient sequence is used because hydrogen bubbles scatter and absorb the ultrasonic fundamental wave when it propagates in the copper melt. The higher the hydrogen content, the more obvious the scattering and absorption of the fundamental wave, and the smaller the fundamental wave amplitude at the receiving end. Since the ultrasonic transmission frequency is fixed, the ratio of the transmission frequency to the received fundamental wave amplitude can directly quantify the degree of fundamental wave attenuation, and thus correlate with the hydrogen content. Using a sequence format can completely record the fundamental wave attenuation at different times, adapting to the time dynamic changes of hydrogen content during the dehydrogenation process. The nonlinear harmonic energy ratio sequence is used because pure copper melt is a linear medium, and ultrasonic propagation only generates the fundamental wave. When hydrogen bubbles are present, the hydrogen bubbles act as a nonlinear medium, causing nonlinear distortion in ultrasonic propagation and generating second harmonics. The higher the hydrogen content, the more obvious the nonlinear distortion, and the higher the proportion of second harmonic energy. The ratio of second harmonic energy to fundamental wave energy can quantify the degree of distortion, directly correlate with the hydrogen content, and the sequence format can record the time dynamic changes of hydrogen content. To extract the ultrasonic fundamental wave attenuation coefficient sequence, firstly, the fundamental wave signal is separated and its amplitude is extracted. The ratio of the transmitted frequency to the received fundamental frequency amplitude is calculated as the ultrasonic fundamental wave attenuation coefficient for a single acquisition. The above operation is repeated according to a preset acquisition period to obtain a continuous ultrasonic fundamental wave attenuation coefficient sequence. To extract the nonlinear harmonic energy ratio sequence, firstly, the fundamental wave and second harmonic signals are separated. The energy ratio between the two is calculated as the nonlinear harmonic energy ratio for a single acquisition. The operation is repeated according to a preset acquisition period to obtain a continuous nonlinear harmonic energy ratio sequence. Depending on the required accuracy of hydrogen content detection, one of the above sequences or two sequences can be extracted simultaneously. Extracting both sequences simultaneously can correlate hydrogen content from different perspectives, reducing detection bias caused by a single feature parameter. Step S240: Remove and calibrate the temperature data at each preset location to eliminate abnormal temperature data caused by temperature measuring device failure or melt flow disturbance. Use temperature data from adjacent locations for interpolation to supplement the data and avoid distortion of the temperature field construction due to missing locations. At the same time, correct the temperature data according to the system error characteristics of thermocouples and infrared temperature measuring devices to ensure that the temperature data collected by different devices are consistent. Step S250: Based on the processed temperature data, using the spatial coordinates of the smelting furnace as a basis, the temperature data of each preset position is mapped to its spatial coordinates to form a three-dimensional temperature data set covering the entire copper melt space, presenting the temperature differences at different spatial locations of the melt. Step S260: Using the spatial coordinates of the three-dimensional temperature field as a reference, divide the space into several small spatial units, calculate the ratio of the temperature difference between each small unit and its adjacent positions to the spatial distance, and take the value with the largest absolute value of the ratio as the temperature gradient value of the small unit to clarify the gradient change of the entire three-dimensional temperature field. Step S270: Traverse each micro-unit and compare its temperature gradient value with a preset threshold. The preset threshold is set according to the hydrogen solubility characteristics of the copper melt and the requirements of the dehydrogenation process. It can be adjusted according to the actual working conditions such as the copper melt composition and smelting temperature. It is used to distinguish between regions with drastic temperature changes and relatively stable regions. A connected component labeling algorithm is used for region merging: Starting from any unlabeled micro-unit, if the gradient value of the micro-unit is less than or equal to the threshold, it is added to the current region. The micro-units in its adjacent directions are recursively checked. Micro-units in adjacent directions with gradient values ​​also less than or equal to the threshold and which are not labeled are added to the same region until no further expansion is possible. If the gradient value of the micro-unit is greater than the threshold, it is marked as an independent connected region and not included in the aforementioned region to avoid confusion between regions with different temperature change characteristics. The above process is repeated until all micro-units are labeled as a certain region. Each labeled region is a thermodynamic response zone, and adjacent thermodynamic response zones are naturally separated by units with significant differences in gradient values.

[0027] In this embodiment, through acoustic response signal preprocessing and reasonable time-frequency analysis, the ultrasonic fundamental attenuation coefficient sequence and the nonlinear harmonic energy ratio sequence are clearly extracted as acoustic characteristic parameters. Both are directly related to hydrogen content. The fundamental attenuation coefficient reflects the scattering and absorption of ultrasonic waves by hydrogen bubbles, while the nonlinear harmonic energy ratio reflects the nonlinear distortion of ultrasonic waves caused by hydrogen bubbles. The sequence form can completely record the dynamic changes of hydrogen content over time. Through temperature signal preprocessing, three-dimensional temperature field construction, temperature gradient calculation, and dynamic region division, it is ensured that the response region matches the temperature distribution and spatial distribution characteristics of the copper melt and hydrogen content, and that the temperature gradient characteristics of each thermodynamic response region are consistent.

[0028] In practical implementation, as one example, existing methods for establishing the mapping relationship between parameters and target values ​​typically employ multiple linear regression or shallow fully connected neural networks for fitting. These models treat the values ​​of each detection parameter at different sampling times as independent input dimensions, or concatenate all input parameters into a single long vector for unified processing. This approach fails to adequately address the nonlinear correlation between hydrogen content and acoustic feature parameters, as well as the spatial information of the response region. Acoustic feature parameters exhibit significant time-series correlation; their values ​​at different sampling times are intrinsically linked, directly reflecting the dynamic changes in hydrogen content over time. Conversely, the spatial information of the response region possesses significant spatial distribution characteristics. The parameter values ​​in different response regions correspond to spatial differences in hydrogen content. The attributes of these two types of parameters are significantly different. Treating them as independent dimensions or concatenating them into a single long vector would lose crucial temporal and spatial distribution features, failing to accurately capture the complex intrinsic relationship between hydrogen content and these two types of parameters. This leads to a significant deviation between the model's predictions and the actual hydrogen content, and the inability to output spatiotemporal distribution information of hydrogen content. This embodiment uses a neural network model to fit the complex mapping relationship between the two types of input parameters and the measured hydrogen content, simultaneously outputting temporal evolution curves and spatial distribution maps to obtain the spatiotemporal distribution of hydrogen content and improve prediction accuracy. The specific implementation steps are as follows: Step S310: Pre-construct a mapping model; First, prepare training data, which includes historical acoustic feature parameters, historical response region spatial information, and corresponding measured hydrogen content values, with each of the three corresponding to the others. The acquisition of historical acoustic feature parameters and historical response region spatial information is achieved through step S200, that is, processing the acoustic response signals of copper melt collected under different batches and different dehydrogenation conditions, and extracting at least one of the ultrasonic fundamental attenuation coefficient sequence and nonlinear harmonic energy ratio sequence to form a set of historical acoustic feature parameters; processing the spatial temperature distribution signals collected under the corresponding batches and conditions to obtain spatial information such as the spatial coordinates, region range, and temperature gradient characteristics of the historical response region, and forming a set of historical response region spatial information; the measured hydrogen content values ​​are obtained using the inert gas melting-infrared absorption method or the thermal conductivity method, obtaining the actual hydrogen content values ​​of the corresponding time and corresponding response region, which are used as label data for model training. The training data is preprocessed, including data normalization and data alignment. Data normalization uses conventional normalization methods to convert the three types of data into the same magnitude range. Data alignment matches the acquisition times of the three types of data to ensure that the three types of data at the same time and in the same response region are matched one by one. Subsequently, a deep learning network structure is selected as the neural network model. The selection can be based on the scale and complexity of the training data. For example, a structure combining convolutional neural networks and recurrent neural networks can be used, where the recurrent neural network is used to capture the temporal correlation of the acoustic feature parameter sequence, and the convolutional neural network is used to extract the spatial features of the response region spatial information. Finally, the preprocessed training data is divided into a training set and a validation set according to a preset ratio. The training set is used for model parameter learning, and the validation set is used to verify the model prediction accuracy. The model weight parameters are adjusted through the backpropagation algorithm until the deviation between the model prediction value and the measured value reaches a preset range. The model parameters are saved to form a mapping model that can be directly called. Step S320: Obtain the acoustic feature parameters and spatial information of the response region under the current working condition and preprocess them. The processing method is consistent with the data preprocessing method in the model training process. Input the preprocessed input parameters into the pre-built mapping model and start the model operation. The mapping model extracts the temporal correlation features of the acoustic feature parameter sequence through the corresponding network structure to capture the change law of hydrogen content over time. At the same time, it extracts the spatial features of the response region spatial information to capture the spatial distribution law of hydrogen content. Through the fully connected layer of the model, the temporal correlation features and spatial features are fused. Combined with the weight parameters of the model training, the hydrogen content values ​​of different times and different response regions are calculated and integrated to form the spatiotemporal distribution information of hydrogen content. Step S330: Generate the time evolution curve and spatial distribution map of hydrogen content. Both together constitute the spatiotemporal distribution information of hydrogen content. The time evolution curve of hydrogen content is generated based on the hydrogen content values ​​of different acquisition cycles output by the mapping model. With the acquisition time as the horizontal axis and the hydrogen content value at the corresponding moment as the vertical axis, the hydrogen content values ​​of each acquisition cycle are connected sequentially to form a continuous curve. This curve represents the trend of hydrogen content change over time and presents the changes in hydrogen content at different dehydrogenation stages. The spatial distribution map of hydrogen content is generated based on the hydrogen content values ​​of different response regions output by the mapping model. Based on the three-dimensional spatial coordinates of the copper melt, the hydrogen content values ​​of each response region are mapped to their spatial locations. Different colors or grayscale values ​​are used to represent the distribution of different hydrogen content values ​​to form a three-dimensional spatial distribution map. This map represents the distribution state of hydrogen content at various spatial locations in the melt and presents regions with high or low hydrogen content.

[0029] In this embodiment, the acquisition and processing of training data ensure the reliability of model training. The selection of the neural network model is adapted to the fitting requirements of the nonlinear mapping relationship, which can fully fit the complex nonlinear mapping relationship between hydrogen content and acoustic feature parameters and response region spatial information. Input parameter preprocessing avoids model operation deviations caused by input parameter misalignment and incompatible formats, ensuring the accuracy of model operation. The generation methods of time evolution curves and spatial distribution maps are clear, which can accurately characterize the spatiotemporal variation characteristics of hydrogen content and adapt to the spatiotemporal dynamic changes of hydrogen content in the dehydrogenation process.

[0030] In a specific implementation, as one example, when extracting relevant features based on the spatiotemporal distribution information of hydrogen content, since hydrogen content has significant temporal evolution and spatial distribution characteristics, the rate of change of hydrogen content varies in different dehydrogenation stages, and the distribution of hydrogen content differs in different spatial locations of the melt. Therefore, simple numerical statistics or single feature extraction methods cannot capture the temporal correlation and spatial distribution gradient of hydrogen content. Temporal evolution features and spatial variation features should be extracted separately for the temporal evolution and spatial distribution characteristics of hydrogen content. The temporal evolution features should reflect the rate of change and / or the trend of the rate of change of hydrogen content over time, while the spatial variation features should reflect the gradient distribution of hydrogen content in spatial locations and the location of supergradient boundaries. This embodiment achieves accurate extraction of both types of features by specifically extracting temporal evolution features and spatial variation features. The specific implementation steps are as follows: Step S410: Based on the time evolution curve, extract the time evolution features. The time evolution features include at least one of the hydrogen content change rate or the hydrogen content change acceleration. The two reflect the time evolution characteristics of hydrogen content from different perspectives. The rate of change of hydrogen content directly characterizes how fast the hydrogen content changes over time, reflecting the efficiency of the dehydrogenation process. The rate of change of hydrogen content varies in different dehydrogenation stages; the larger the rate of change, the more drastic the dynamic change of hydrogen content. The method for obtaining the rate of change of hydrogen content is as follows: based on the hydrogen content values ​​of two adjacent collection cycles in the time evolution curve and the collection cycle duration, the ratio of the difference between the two values ​​to the collection cycle duration is calculated to obtain the rate of change of hydrogen content within that time period; the calculation is performed sequentially according to the collection cycle to obtain a continuous set of hydrogen content change rates. The acceleration of hydrogen content change characterizes the trend of the rate of change of hydrogen content and reflects the dynamic changes in the dehydrogenation process. For example, when the acceleration is positive, it indicates that the rate of change of hydrogen content is accelerating and the dehydrogenation efficiency is improving; when the acceleration is negative, it indicates that the rate of change of hydrogen content is slowing down and the dehydrogenation process is stabilizing. The acceleration of hydrogen content change is obtained as follows: based on the continuous rate of change of hydrogen content calculated above, the ratio of the difference between two adjacent rates of change to the duration of the collection period is calculated to obtain the acceleration of hydrogen content change within that time period; the continuous set of accelerations of hydrogen content change is obtained by calculating sequentially according to the collection period. When choosing to extract one feature, the corresponding feature can be selected based on the basic requirements of parameter adjustment. For example, if only the rate of change of hydrogen content needs to be known, the rate of change of hydrogen content can be extracted; if fine adjustment of the acquisition parameters is required, the acceleration of the change of hydrogen content can be extracted; when both features are extracted simultaneously, more comprehensive and accurate support can be provided for the adjustment of acquisition parameters, avoiding adjustment deviations caused by a single feature, and ensuring that the acquisition parameters are adapted to the dynamic changes of hydrogen content over time in the dehydrogenation process. Step S420: Based on the spatial distribution map, extract spatial variation features, including hydrogen content gradient values ​​at different spatial locations and the location boundaries where the hydrogen content gradient values ​​exceed a preset gradient threshold. The hydrogen content gradient value characterizes the rate of change of hydrogen content between different spatial locations, reflecting the degree of spatial unevenness of hydrogen content distribution. The larger the gradient value, the greater the difference in hydrogen content between adjacent spatial locations, and the more uneven the spatial distribution of hydrogen content in that region; the smaller the gradient value, the smaller the difference in hydrogen content between adjacent spatial locations, and the more uniform the hydrogen content distribution. The hydrogen content gradient value is obtained as follows: based on the three-dimensional spatial coordinates of the spatial distribution map, divide it into several small spatial units, each corresponding to a hydrogen content value; calculate the ratio of the difference in hydrogen content values ​​between adjacent small spatial units to the spatial distance, and this ratio is the hydrogen content gradient value at the corresponding spatial location; traverse all small spatial units of the entire spatial distribution map to calculate the hydrogen content gradient value at each spatial location, forming a set of hydrogen content gradient values. Location boundaries are critical regions where the spatial distribution of hydrogen content differs significantly. The hydrogen content distribution on either side of the boundary shows a marked difference, and identifying this boundary clarifies the spatial range requiring focused attention. The location boundary is identified by: traversing the hydrogen content gradient values ​​at all spatial locations and marking small spatial units with gradient values ​​exceeding a preset gradient threshold as boundary units; connecting adjacent boundary units to form a continuous spatial boundary line or boundary surface, which represents the location boundary where the hydrogen content gradient value exceeds the preset threshold. The hydrogen content gradient threshold is determined based on the requirements of the copper melt dehydrogenation process and the accuracy requirements for hydrogen content detection, combined with the normal distribution range of hydrogen content in the copper melt. It can be adjusted according to the actual operating conditions such as the composition of the copper melt, smelting temperature, and dehydrogenation stage, and is used to distinguish between regions with uniform and non-uniform spatial distribution of hydrogen content.

[0031] In this embodiment, by extracting temporal evolution features and spatial variation features respectively, the two different types of features in the spatiotemporal distribution information of hydrogen content are decoupled. The rate of change in the temporal evolution features reflects the speed of the dehydrogenation process, and the acceleration of change reflects the trend of the dehydrogenation rate, thus constituting a description of the temporal dynamics of hydrogen content. The hydrogen content gradient value in the spatial variation features quantifies the difference in hydrogen content between adjacent spatial locations, and the location boundary marks the critical region where the gradient exceeds the threshold. Both together locate the spatial range of uneven hydrogen content distribution in the melt. The temporal evolution features and spatial variation features correspond to the adjustment of the acquisition window length and sampling frequency, as well as the adjustment of the sampling weight of the response region, respectively, forming a direct physical correspondence between feature extraction and control actions.

[0032] In a specific implementation, as one example, when adjusting the acquisition parameters and sampling weights, if a longer acquisition window or a lower sampling frequency is used during the rapid hydrogen content change phase, the intermediate process information of hydrogen content fluctuations will be lost, resulting in insufficient temporal resolution of the input data to the subsequent mapping model to distinguish the details of the changes. If a shorter acquisition window or a higher sampling frequency is used during the stable hydrogen content phase, redundant data will be generated, increasing the burden of signal acquisition and model calculation. Similarly, if the same sampling weight is used in the boundary region with a large spatial gradient of hydrogen content as in the uniform region, the difference in hydrogen content on both sides of the boundary will not be accurately captured due to insufficient sampling points. If an excessively high sampling weight is used in the uniform region, it will waste sampling resources. Therefore, it is necessary to adjust the acquisition window length and sampling frequency according to the change amplitude of the time evolution characteristics, and adjust the sampling weight of each response region according to the distribution of spatial change characteristics. This embodiment completes the next cycle of acquisition and information update by matching acquisition parameters with time evolution characteristics and allocating sampling weights with spatial change characteristics. The specific implementation steps are as follows: Step S510: Adjust the acquisition window length or sampling frequency of the acoustic response signal according to the change amplitude of the time evolution characteristics. The change amplitude of the time evolution characteristics is determined by the difference between the current period characteristic value and the previous period characteristic value. An increase in the change amplitude indicates that the hydrogen content change tends to be drastic. At this time, shorten the acquisition window length to reduce signal analysis lag, or increase the sampling frequency to improve the ability to capture details. A decrease in the change amplitude indicates that the hydrogen content change tends to be gradual. At this time, extend the acquisition window length or reduce the sampling frequency to reduce redundant data acquisition. Step S520: Adjust the sampling weight of each response region according to the distribution of spatial variation characteristics. The sampling weight is used to adjust the acoustic response signal acquisition density in the corresponding region, or to adjust the input contribution of the acoustic feature parameters in the mapping model. The spatial variation characteristics of the response region exceed the preset spatial feature threshold, and the content of the content varies greatly in space. It is necessary to increase the detection attention. Therefore, the sampling weight is increased to increase the acoustic response signal acquisition density in the region or to increase the input contribution of the acoustic feature parameters in the mapping model. The hydrogen content of the response region is evenly distributed and does not need to be repeatedly acquired. Therefore, the sampling weight is reduced to decrease the acoustic response signal acquisition density in the region or to reduce the input contribution of the acoustic feature parameters in the mapping model. Step S530: Using the adjusted acquisition parameters and sampling weights, perform synchronous acquisition of the acoustic response signal and spatial temperature distribution signal of the copper melt in the next cycle, and transmit the acquired signals sequentially to steps S200 to S400 to complete acoustic feature extraction, response region division, acquisition of hydrogen content spatiotemporal distribution information, and updating of time evolution characteristics and spatial change characteristics.

[0033] In this embodiment, by establishing a direct correspondence between the magnitude of changes in temporal evolution characteristics and the length of the acquisition window and the sampling frequency, the acquisition parameters are adjusted synchronously with the degree of change in hydrogen content: when the change intensifies, the time window is shortened and the frequency is increased to improve temporal resolution; when the change intensifies, the time window is extended and the frequency is reduced to reduce redundant data. By establishing a direct correspondence between spatial change characteristics and sampling weights, sampling resources are tilted towards areas with large spatial differences in hydrogen content: in areas where the gradient exceeds the threshold, the sampling weight is increased to densify acquisition or enhance model attention; in areas where the gradient is below the threshold, the sampling weight is reduced to avoid wasting resources. After the next cycle is executed, the adjustment results of the acquisition parameters and sampling weights are updated through steps S200 to S400, forming a closed-loop link of feature extraction, parameter adjustment, signal acquisition, and information update, so that the detection behavior is synchronized with the real-time status of the dehydrogenation process.

[0034] Although this application has been described in conjunction with specific features and embodiments, it is obvious that various modifications and combinations can be made thereto without departing from the spirit and scope of this application. Accordingly, this specification and drawings are merely exemplary illustrations of the application as defined herein, and are to be considered as covering any and all modifications, variations, combinations, or equivalents within the scope of this application. Clearly, those skilled in the art can make various alterations and modifications to this application without departing from its scope. Thus, if such modifications and modifications fall within the scope of this application and its equivalents, this application intends to include such modifications and modifications.

Claims

1. A method for detecting hydrogen content in a copper melt based on ultrasonic dehydrogenation, the method being applied to ultrasonic dehydrogenation of a copper melt, characterized by, include: Simultaneously acquire acoustic response signals and spatial temperature distribution signals of the copper melt; Based on the acoustic response signal, acoustic feature parameters related to hydrogen content are extracted, and the melt space is divided into multiple response regions based on the spatial temperature distribution signal. The acoustic feature parameters and the spatial information of the response region are input into a pre-constructed mapping model to obtain the spatiotemporal distribution information of hydrogen content; Based on the aforementioned spatiotemporal distribution information, temporal evolution features and spatial change features are extracted; Based on the temporal evolution characteristics and spatial variation characteristics, the acquisition parameters of the acoustic response signal and the sampling weights of each response region are adjusted, and the signal acquisition and spatiotemporal distribution information update of hydrogen content are performed in the next cycle.

2. The method for detecting hydrogen content in copper melt based on ultrasonic dehydrogenation according to claim 1, characterized in that, The spatiotemporal distribution information of hydrogen content includes the time evolution curve and spatial distribution map of hydrogen content; The time evolution curve represents the trend of hydrogen content change over time, and the spatial distribution map represents the distribution of hydrogen content at various spatial locations in the melt.

3. The method for detecting hydrogen content in copper melt based on ultrasonic dehydrogenation according to claim 2, characterized in that, The extracted time evolution features include: Based on the time evolution curve, at least one of the hydrogen content change rate or hydrogen content change acceleration is calculated as the time evolution feature.

4. The method for detecting hydrogen content in copper melt based on ultrasonic dehydrogenation according to claim 2, characterized in that, The extraction of spatial variation features includes: Based on the spatial distribution map, the hydrogen content gradient values ​​at different spatial locations are calculated, and the location boundaries where the hydrogen content gradient values ​​exceed a preset gradient threshold are identified as the spatial variation features.

5. The method for detecting hydrogen content in copper melt based on ultrasonic dehydrogenation according to claim 1, characterized in that, The acoustic response signal is a time-domain waveform signal received after the ultrasonic wave propagates in the copper melt; The extraction of acoustic feature parameters related to hydrogen content includes: Time-frequency analysis is performed on the time-domain waveform signal to extract at least one of the ultrasonic fundamental attenuation coefficient sequence and the nonlinear harmonic energy ratio sequence as the acoustic characteristic parameter.

6. The method for detecting hydrogen content in copper melt based on ultrasonic dehydrogenation according to claim 5, characterized in that, The ultrasonic fundamental frequency attenuation coefficient is determined based on the ratio of the ultrasonic transmission frequency to the fundamental frequency amplitude of the received signal; the nonlinear harmonic energy ratio is determined based on the ratio of the second harmonic energy to the fundamental frequency energy in the received signal.

7. The method for detecting hydrogen content in copper melt based on ultrasonic dehydrogenation according to claim 1, characterized in that, The spatial temperature distribution signal is acquired by thermocouples or infrared temperature measuring devices installed at multiple preset locations within the melting furnace; the dynamic division of the melt space into multiple response regions includes: Based on the spatial temperature distribution signal, a three-dimensional temperature field distribution of copper melt is constructed, and according to the comparison result of the gradient change of the three-dimensional temperature field distribution with the preset temperature gradient threshold, the melt space is dynamically divided into multiple thermodynamic response zones, which are referred to as the response regions.

8. The method for detecting hydrogen content in copper melt based on ultrasonic dehydrogenation according to claim 1, characterized in that, The mapping model is a neural network model pre-constructed based on the mapping relationship between historical acoustic feature parameters, historical response region spatial information and corresponding measured hydrogen content values.

9. The method for detecting hydrogen content in copper melt based on ultrasonic dehydrogenation according to claim 1, characterized in that, The adjustment of the acquisition parameters of the acoustic response signal includes: Based on the magnitude of the change in the time evolution characteristics, the acquisition window length or sampling frequency of the acoustic response signal is adjusted; wherein, when the magnitude of the change in the time evolution characteristics increases, the acquisition window length is shortened or the sampling frequency is increased.

10. The method for detecting hydrogen content in copper melt based on ultrasonic dehydrogenation according to claim 1, characterized in that, The adjustment of the sampling weights of each of the response regions includes: Based on the distribution of the spatial change features, increase the sampling weight of the response region where the spatial change features exceed a preset spatial feature threshold, and / or decrease the sampling weight of the response region where the spatial change features are below the preset spatial feature threshold. The sampling weights are used to adjust the acquisition density of acoustic response signals in each response region, or to adjust the input contribution of the acoustic feature parameters corresponding to the response region in the mapping model.

Images