Corrosion quantification method, device and equipment for liquid lead bismuth loop pipeline

Abstract

The invention relates to a corrosion quantification method, device and equipment for a liquid lead-bismuth loop pipeline, and aims at quantitative evaluation of corrosion of the liquid lead-bismuth loop pipeline, multi-modal data such as dissolved oxygen, temperature, flow velocity, electrochemical impedance spectroscopy and eddy current signals of a monitoring position are obtained, and through time sequence alignment, noise reduction, equivalent circuit fitting and feature extraction, the corrosion of the liquid lead-bismuth loop pipeline is quantified. The method comprises the following steps: forming multi-dimensional features including an oxidation film, an electrochemical interface, local and uniform corrosion and environmental information, further constructing a multi-input deep learning model combining an attention mechanism and multi-modal coding, extracting time sequence dependence by using a long-short term memory artificial neural network, and realizing multi-modal feature weighted fusion; and respectively outputting the uniform corrosion rate, the local corrosion probability, the pit depth, the oxide film state grade and the recommended oxygen concentration interval. And meanwhile, ultrasonic data are introduced to calibrate an eddy current result, so that comprehensive quantitative evaluation on the corrosion state of the loop is realized.

Description

Technical Field

[0001] This application relates to the field of pipeline corrosion monitoring technology, and more specifically to a method, apparatus and equipment for quantifying corrosion in liquid lead-bismuth loop pipelines. Background Technology

[0002] In lead-cooled fast reactors (LFRs) and accelerator-driven subcritical systems (ADSs), liquid lead-bismuth eutectic (LBE) is widely used as a coolant and heat dissipation medium in the primary loop and the thermal-hydraulic systems of key components. It has advantages such as high boiling point, good thermal conductivity and low neutron absorption cross section, making it a highly promising coolant. However, the corrosion behavior of liquid lead-bismuth eutectic on structural materials exhibits complex and variable characteristics. Therefore, the complete monitoring and evaluation of liquid lead-bismuth eutectic materials is directly related to the reliable operation of the reactor core, system lifetime and irradiation safety.

[0003] Under the multi-physics coupling effect of high temperature, flow rate, temperature gradient and strong irradiation field, liquid lead-bismuth eutectic structural materials will undergo a variety of overlapping or competing corrosion and damage processes, such as oxide film growth, oxide film rupture and regeneration, dissolution corrosion, pitting corrosion, intergranular corrosion, and irradiation-induced corrosion acceleration. In particular, under the action of irradiation field, the enhanced diffusion caused by point defects, irradiation-assisted dissolution, and local disturbance of oxygen concentration will significantly change the interfacial reaction kinetics between the material surface and liquid lead-bismuth.

[0004] Currently, the corrosion monitoring methods commonly used in liquid lead-bismuth loop pipelines mainly include periodic sampling and single ultrasonic thickness measurement. These methods can obtain wall thickness changes or local surface damage to a certain extent, but they usually have problems such as long detection cycles, single information dimensions, inability to distinguish the state of the surface oxide film from the corrosion characteristics of the substrate, and inability to provide a basis for decision-making on oxygen concentration control in liquid lead-bismuth loop pipelines. Furthermore, existing monitoring methods cannot simultaneously obtain multimodal correlation information between environmental factors and material corrosion, making it difficult to analyze the corrosion mechanism and to conduct real-time quantitative assessment of corrosion evolution trends and local high-risk areas under multi-physics coupling. Summary of the Invention

[0005] This application provides a method, apparatus, and equipment for quantifying corrosion in liquid lead-bismuth loop pipelines. It can simultaneously collect multi-source data such as environmental parameters, electrochemical signals, eddy current signals, and ultrasonic signals under actual operating conditions of liquid lead-bismuth loop pipelines. By multimodal fusion and feature extraction, the corrosion state of the material is characterized and a nonlinear correlation between environmental factors and material corrosion response is established. This enables a quantitative assessment of the eutectic corrosion behavior of liquid lead-bismuth and provides a decision-making basis for oxygen concentration control in liquid lead-bismuth loop pipelines.

[0006] In a first aspect, this application provides a corrosion quantification method for liquid lead-bismuth loop pipelines. The method includes: acquiring environmental parameters characterizing corrosion driving factors and electrochemical and eddy current signals characterizing material response behavior at at least one monitoring location in the liquid lead-bismuth loop, generating multimodal data; performing data preprocessing and equivalent circuit fitting on the multimodal data to generate multidimensional feature data including surface and substrate corrosion characteristics, oxide film characteristics, interface reaction characteristics, and environmental characteristics; inputting the multidimensional feature data into a multi-input deep learning model employing an attention mechanism and establishing a nonlinear correlation between environmental parameters and material corrosion response, and outputting a quantitative index characterizing the corrosion state, wherein the quantitative index includes at least one of uniform corrosion rate, probability of local corrosion occurrence and pitting depth, oxide film state level, and recommended oxygen concentration range; evaluating and outputting the quantification result of the liquid lead-bismuth loop based on the quantitative index.

[0007] In one alternative embodiment of the first aspect, the method for generating multidimensional feature data including surface and substrate corrosion characteristics, oxide film characteristics, interface reaction characteristics, and environmental characteristics includes: acquiring dissolved oxygen concentration, temperature, flow rate, eddy current signal, and electrochemical impedance spectroscopy data of sensor clusters at high-temperature sections, bends, and areas with temperature gradients in the liquid lead-bismuth circuit; performing time-series alignment and normalization on the dissolved oxygen concentration, temperature, and flow rate; performing noise reduction processing on the eddy current signal and extracting amplitude and phase characteristics to generate surface oxide film thickness, local corrosion depth, and low-frequency wall thickness trend characteristics; and performing equivalent circuit fitting on the electrochemical impedance spectroscopy data to extract oxide film resistance, oxide film capacitance, and charge transfer resistance characteristics.

[0008] In one alternative embodiment of the first aspect, the method for fitting the equivalent circuit includes: selecting a preset equivalent circuit model to simulate the physical structure of the solution resistance, the outer oxide film, and the inner interface using the electrochemical impedance spectroscopy data, and performing a preset number of iterative fittings using a nonlinear least squares method to generate a fitting result; extracting the oxide film resistance and charge transfer resistance from the fitting result, and calculating the oxide film capacitance based on the oxide film resistance and charge transfer resistance. ,in For oxide film capacitors, For the amplitude parameters of the oxide film CPE, The exponential factor of CPE in oxide film, Oxide film resistance; wherein, the preset equivalent circuit model is an R(QRf)(QRct) model, and its circuit structure is as follows: , For the resistance of the solution, For oxide film CPE, It is a double-layer CPE. It is a charge transfer resistor.

[0009] In one alternative embodiment of the first aspect, the method for extracting amplitude and phase features includes: simultaneously or sequentially introducing an alternating current of a preset frequency into a high-temperature differential eddy current probe encapsulated on the outer wall of the pipe, and utilizing the skin effect penetration depth at different frequencies to obtain eddy current signals corresponding to the surface oxide film and substrate corrosion, wherein the preset frequency includes frequencies from 100kHz low frequency to 1MHz high frequency; performing DC removal and baseline correction on the eddy current signals, and applying narrowband bandpass filtering to the eddy current signals of different frequencies respectively, and then using wavelet denoising to remove transient random noise from the eddy current signals; inverting the conductivity change of the surface oxide film based on the amplitude and phase of the impedance plane of the high-frequency eddy current signal, and generating the average oxide film thickness according to the calibrated lookup table relationship; identifying the eddy current field distortion caused by local corrosion based on the amplitude and phase of the low-frequency eddy current signal, and matching the local corrosion size by the differential features and calibration curve, and then obtaining the remaining wall thickness change caused by uniform corrosion of the pipe wall based on the impedance change trend of the low-frequency eddy current signal, generating low-frequency wall thickness trend features.

[0010] In one alternative of the first aspect, the method for establishing a nonlinear correlation between environmental parameters and material corrosion response using a multi-input deep learning model employing an attention mechanism includes: retrieving historical data from a historical database of at least one monitoring location in a liquid lead-bismuth circuit within a preset time window and generating corresponding time-series feature data based on the historical data, wherein the time-series feature data includes oxide film resistance, interfacial charge transfer resistance, oxide film capacitance, eddy current phase, eddy current amplitude variation, dissolved oxygen concentration, temperature, and flow rate; setting a preset number of hidden units for a long short-term memory artificial neural network and inputting the time-series feature data into the long short-term memory artificial neural network. The network encodes the dynamic changes of temporal feature data through forget gates, input gates, and output gates to generate temporal features representing the corrosion state. The original temporal features of each modality are mapped to a common feature space of the same dimension using a preset encoder, and the similarity between the temporal features of each modality is calculated using an attention mechanism to generate attention weights. Based on the attention weights, the temporal features of different modalities are weighted and fused to generate fused temporal features reflecting the collaborative relationship of multiple modalities. The fused temporal features are then input into preset prediction branches, which output uniform corrosion rate, probability of localized corrosion, pitting depth, oxide film state level, and recommended oxygen concentration range.

[0011] In one alternative of the first aspect, the long short-term memory artificial neural network performs forget gate, input gate and output gate operations on the temporal feature data at each time step. When performing the forget gate operation, it decides whether to retain the previous state; when performing the input gate operation, it selectively writes new erosion-related features; when performing the output gate operation, it generates the corresponding hidden state, thereby gradually forming temporal features that characterize the erosion state.

[0012] In one alternative of the first aspect, the preset encoder includes an electrochemical mode encoder, an eddy current detection mode encoder, and an environmental mode encoder. The electrochemical mode encoder, the eddy current detection mode encoder, and the environmental mode encoder are trained separately to map the original features of electrochemical modes, eddy current detection modes, and environmental modes with different dimensions and different physical meanings to a common feature space of the same dimension.

[0013] In one alternative to the first aspect, the method for generating attention weights includes: posing a question: ,in, To query the weight matrix, For environmental modal temporal features; collection options: , , ,in, The value of electrochemical mode timing characteristics; This is the key-value weight matrix; These are the temporal characteristics of electrochemical modes, such as oxide film resistance, interfacial charge transfer resistance, and oxide film capacitance. The value of eddy current mode time series characteristics These are the temporal characteristics of the eddy current modes, such as eddy current phase and amplitude variations; The value of temporal features of the environmental modality is calculated; the dot product similarity between each question and each collection option is calculated to generate an initial score. , , ,in The initial score for the time-series characteristics of electrochemical modes. The initial score for the temporal characteristics of the eddy current modes. The initial scores for the temporal features of the environmental modality are used; these initial scores are then normalized into attention weights using the Softmax function. , , ,in Attention weights for the temporal characteristics of electrochemical modes. Attention weights for the temporal features of eddy current modes. Attention weights are assigned to the temporal features of different modalities; based on these attention weights, a weighted fusion of the temporal features of different modalities is performed to generate fused temporal features that reflect the collaborative relationships among multiple modalities. .

[0014] In one optional embodiment of the first aspect, the preset prediction branch includes a corrosion rate regression branch, a localized corrosion risk hybrid branch, an oxide film state classification branch, and a recommended oxygen concentration regression branch. The corrosion rate regression branch processes the fused temporal features through a linear regression layer to output the current uniform corrosion rate. The localized corrosion risk hybrid branch outputs the probability of localized corrosion occurrence through a first linear layer and a Sigmoid activation function, and outputs the predicted value of localized pitting depth through a second linear layer. The oxide film state classification branch outputs the state levels of oxide film stability, degradation, and failure through a linear layer and a Softmax activation function. The recommended oxygen concentration regression branch generates endpoint values ​​for the oxygen concentration range through a linear layer and a Tanh activation function, and scales these endpoint values ​​to a preset physical range to output the recommended oxygen concentration range.

[0015] In one alternative embodiment of the first aspect, the generation of multidimensional feature data further includes: acquiring ultrasonic signals from probes that penetrate at least partially into the high-temperature section, bends, and areas with temperature gradients in the liquid lead-bismuth circuit during planned shutdowns or preset maintenance cycles; calculating the oxide film thickness, local corrosion depth, and wall thickness trend characteristics at the corresponding locations in the liquid lead-bismuth circuit based on the acoustic time of the ultrasonic signals; and calibrating the quantization results of the eddy current signals based on the oxide film thickness, local corrosion depth, and wall thickness trend characteristics of the ultrasonic signals.

[0016] In a second aspect, this application provides a corrosion quantification device for a liquid lead-bismuth loop pipeline, comprising at least one processor; at least one memory; the at least one memory being coupled to the at least one processor and used to store instructions executed by the at least one processor, the instructions, when executed by the at least one processor, causing the corrosion quantification device to perform the method according to any one of the first aspects.

[0017] Thirdly, this application provides a corrosion quantification device for liquid lead-bismuth loop pipelines, including the aforementioned corrosion quantification apparatus; a sensing cluster, a three-electrode probe system, a high-temperature differential eddy current probe, a telescopic in-situ multi-parameter probe, a solid oxygen sensor, a temperature sensor, and a flow meter. The three-electrode probe system includes a working electrode, a reference electrode, and a counter electrode. The working electrode is made of the same material as the liquid lead-bismuth loop pipeline. The reference electrode uses a yttrium-stabilized zirconia electrolyte tube encapsulated with Bi / Bi₂O₃ or Pb / PbO as a reference material. The counter electrode is made of platinum or tungsten. This allows for the acquisition of electrochemical values ​​at predetermined locations within the liquid lead-bismuth loop pipeline. Impedance spectroscopy data; the high-temperature differential eddy current probe is encapsulated at a preset position in the liquid lead-bismuth loop pipe to acquire a preset frequency eddy current signal at the corresponding position; the telescopic in-situ multi-parameter probe remains closed during normal operation and, during operation, at least partially extends, inserts, and is fixed at a preset position in the liquid lead-bismuth loop pipe to acquire ultrasonic, acoustic emission, and frictional force signals at the corresponding position; the solid oxygen sensor is used to acquire dissolved oxygen concentration data at the preset position in the liquid lead-bismuth loop pipe; the temperature sensor is used to acquire temperature data at the preset position in the liquid lead-bismuth loop pipe; and the flow meter is used to acquire flow velocity data at the preset position in the liquid lead-bismuth loop pipe.

[0018] It should be understood that the above general description and the following detailed description are merely exemplary and do not limit this application. Attached Figure Description

[0019] Figure 1 This is a schematic diagram illustrating the overall trend of corrosion evolution in an existing exemplary liquid lead-bismuth loop pipeline.

[0020] Figure 2 This is a schematic flowchart of an exemplary corrosion quantification method according to some embodiments of this application.

[0021] Figure 3 This is an exemplary electrochemical impedance Nyquist plot of a liquid lead-bismuth loop pipe material under different operating time conditions according to some embodiments of this application.

[0022] Figure 4 This is an exemplary correspondence to some embodiments of this application. Figure 3 Electrochemical impedance Bode amplitude-frequency relationship graph.

[0023] Figure 5 This is an exemplary correspondence to some embodiments of this application. Figure 3 Electrochemical impedance Bode phase angle-frequency relationship graph.

[0024] Figure 6This is a schematic diagram of the calibration relationship curve between the differential amplitude change of a low-frequency eddy current detection signal and the local pitting depth, according to some embodiments of this application.

[0025] Figure 7 This is an exemplary schematic diagram of the correspondence between the phase shift of a low-frequency eddy current detection signal and the size of a local corrosion feature, according to some embodiments of this application.

[0026] Figure 8 This is a schematic flowchart of an exemplary amplitude and phase feature extraction method according to some embodiments of this application.

[0027] Figure 9 This is a schematic flowchart of an exemplary equivalent circuit fitting method according to some embodiments of this application.

[0028] Figure 10 This is an exemplary equivalent circuit model schematic diagram according to some embodiments of this application.

[0029] Figure 11 This is a schematic diagram comparing exemplary electrochemical impedance spectroscopy experimental data with equivalent circuit fitting results according to some embodiments of this application.

[0030] Figure 12 A schematic diagram illustrating the relationship between oxide film resistance and operating time, obtained by equivalent circuit fitting according to some embodiments of this application.

[0031] Figure 13 A schematic diagram illustrating the relationship between charge transfer resistance and operating time, obtained by equivalent circuit fitting, according to some embodiments of this application.

[0032] Figure 14 A schematic diagram illustrating the relationship between the equivalent capacitance of the oxide film obtained by equivalent circuit fitting and the operating time, according to some embodiments of this application.

[0033] Figure 15 This is a flowchart illustrating an exemplary multi-input deep learning model construction method according to some embodiments of this application.

[0034] Figure 16 This is a flowchart illustrating an exemplary method for generating fused temporal features according to some embodiments of this application.

[0035] Figure 17 This is a schematic diagram illustrating the evolution of corrosion rate of a pipeline material over operating time, according to some embodiments of this application.

[0036] Figure 18This is a schematic diagram illustrating the time-series variation of temperature parameters of an exemplary operating pipeline according to some embodiments of this application.

[0037] Figure 19 This is a time series diagram illustrating the change of oxygen concentration over time in an exemplary operating pipeline system according to some embodiments of this application.

[0038] Figure 20 This is a time series diagram illustrating the change of fluid velocity over time in an exemplary operating pipe according to some embodiments of this application.

[0039] Figure 21 This is an exemplary schematic diagram showing how the attention weights of different modal input features in a model change over runtime, according to some embodiments of this application.

[0040] Figure 22 This is a schematic diagram comparing an exemplary corrosion rate prediction result based on some embodiments of this application with a calibration result obtained experimentally or by calibration over time.

[0041] Figure 23 This is a schematic flowchart of an exemplary ultrasonic signal processing method according to some embodiments of this application.

[0042] Figure 24 This is a schematic diagram of the connection of an exemplary corrosion quantification device according to some embodiments of this application.

[0043] Figure 25 This is a connection diagram of an exemplary corrosion quantification device according to some embodiments of this application. Detailed Implementation

[0044] The exemplary implementation will now be described more fully with reference to the accompanying drawings.

[0045] refer to Figure 1 As shown, Figure 1 This paper illustrates an exemplary, existing, continuous, and describable evolution trend of pipeline corrosion over operating time in a liquid lead-bismuth eutectic environment. Under typical operating conditions, liquid lead-bismuth eutectic exists in an environment with high temperature (400–550°C), high-speed flow, strong temperature gradient, and high irradiation flux. The surface of the structural material undergoes various degradation processes, including oxide film growth and rupture, dissolution corrosion, localized corrosion (pitting corrosion, intergranular corrosion), and liquid metal embrittlement (LME). Furthermore, the accumulation of lattice defects caused by the irradiation field and the irradiation-induced diffusion acceleration effect are significantly coupled with the corrosion process, forming an irradiation-corrosion coupling effect, which further reduces the service life of the material.

[0046] The liquid lead-bismuth eutectic (LBE) refers to a low-melting-point alloy formed by mixing lead and bismuth in a certain proportion (e.g., 44.5% Pb, 55.5% Bi), which is usually used as a coolant or spallation target; the irradiation-corrosion coupling effect refers to the phenomenon that when a material is simultaneously subjected to irradiation by particles such as neutrons and LBE corrosion, the two promote each other and accelerate the degradation of material properties.

[0047] Currently, the corrosion monitoring methods used in liquid lead-bismuth loop pipelines in engineering mainly include periodic disassembly and sampling and external inspection (such as metallography, SEM, EDS) and online ultrasonic thickness measurement. Although periodic disassembly and sampling and external inspection can provide microscopic information, they require shutdown of the reactor or system, which has drawbacks such as long cycle, strong lag, and interference with components. Online ultrasonic thickness measurement can monitor the wall thickness changes caused by uniform corrosion to a certain extent, but its signal only reflects geometric information and cannot distinguish key factors such as corrosion type, oxide film quality, and corrosion rate change mechanism.

[0048] Therefore, for reference Figure 2 As shown, Figure 2 A schematic flowchart of an exemplary corrosion quantification method according to some embodiments of this application is shown. This application relates to a corrosion quantification method for liquid lead-bismuth loop pipelines, the method comprising:

[0049] 11: Obtain environmental parameters characterizing corrosion drivers and electrochemical and eddy current signals characterizing material response behavior at at least one monitoring location in the liquid lead-bismuth loop pipeline to generate multimodal data.

[0050] Specifically, the monitoring locations include, but are not limited to, high-temperature sections, bends, and areas with temperature gradients in the liquid lead-bismuth loop pipeline. Multiple sensors are integrated and deployed at these locations, such as a three-electrode probe, a high-temperature differential eddy current probe, a telescopic in-situ multi-parameter probe, a solid oxygen sensor, a temperature sensor, and a flow meter, to construct a sensor cluster and synchronize the data from all sensors in the cluster over time. The three-electrode probe includes a working electrode, a reference electrode, and a counter electrode. The working electrode is made of the same material as the liquid lead-bismuth loop pipeline. The reference electrode uses a yttrium-stabilized zirconia electrolyte tube encapsulated with Bi / Bi₂O₃ or Pb / PbO as the reference material. The counter electrode is made of platinum or tungsten. This three-electrode probe acquires electrochemical impedance spectroscopy data at the location within the liquid lead-bismuth loop pipeline. The high-temperature differential eddy current probe is used to acquire eddy current signals at a preset frequency at the corresponding location. The telescopic in-situ multi-parameter probe remains closed during normal operation and, during operation, extends at least partially to be inserted into and fixed at the location in the liquid lead-bismuth loop pipeline. It is used to acquire ultrasonic, acoustic emission, and frictional signals at the corresponding location. The solid-state oxygen sensor is used to acquire dissolved oxygen concentration data at a preset location in the liquid lead-bismuth loop pipeline. The temperature sensor is used to acquire temperature data at a preset location in the liquid lead-bismuth loop pipeline. The flow meter is used to acquire flow velocity data at a preset location in the liquid lead-bismuth loop pipeline.

[0051] In this study, during the operation of the liquid lead-bismuth loop pipeline, impedance spectroscopy was performed using a three-electrode probe system positioned on the inner wall of the pipeline to obtain electrochemical impedance data under different operating time conditions. (Reference) Figure 3 As shown, the electrochemical impedance Nyquist plots at different operating times exhibit significant differences, indicating that the oxide film structure and corrosion behavior on the pipe material surface evolve over time. By comparing the changes in impedance spectrum radius at different times, the evolution process of oxide film formation and corrosion behavior can be characterized. The electrochemical impedance Nyquist plots at different operating times exhibit significant differences, indicating that the oxide film structure and corrosion behavior on the pipe material surface evolve over time.

[0052] Furthermore, regarding Figure 3 Frequency domain analysis was performed on the impedance data shown, and the following results were obtained: Figure 4 and Figure 5 The Bode amplitude-frequency and phase angle-frequency graphs shown are as follows. Figure 4 It reflects the changes in the system impedance characteristics in different frequency ranges, and is used to distinguish between oxide film-controlled processes and interfacial charge transfer processes; Figure 5 The existence of different time constants in the characterization system and their evolution characteristics over time. Figure 4 and Figure 5It can be seen that the system exhibits different impedance response characteristics in different frequency ranges, indicating that the corrosion process is simultaneously affected by the oxide film characteristics and the interfacial electrochemical reaction.

[0053] 12: Perform data preprocessing and equivalent circuit fitting on the multimodal data to generate multidimensional feature data including surface and substrate corrosion characteristics, oxide film characteristics, interface reaction characteristics, and environmental characteristics.

[0054] Specifically, after collecting dissolved oxygen concentration, temperature, flow rate, eddy current signal, and electrochemical impedance spectroscopy data at high-temperature sections, bends, and temperature gradient areas in the liquid lead-bismuth loop pipeline using the aforementioned three-electrode probe system, high-temperature differential eddy current probe, telescopic in-situ multi-parameter probe, solid oxygen sensor, temperature sensor, and flow meter, the collected dissolved oxygen concentration, temperature, and flow rate are time-series aligned and normalized. The eddy current signal is denoised and amplitude-phase features are extracted to generate surface oxide film thickness, local corrosion depth, and low-frequency wall thickness trend features. The electrochemical impedance spectroscopy data are fitted with an equivalent circuit to extract oxide film resistance, oxide film capacitance, and charge transfer resistance features.

[0055] To achieve a quantitative assessment of the degree of localized corrosion, the detection signals acquired by a high-temperature differential eddy current probe deployed outside the pipeline were analyzed. (Reference) Figures 6 to 7 As shown, Figure 6 The differential amplitude variation of the low-frequency eddy current detection signal shows a clear correspondence with the local pitting depth, which can be used to achieve quantitative estimation of local corrosion depth. Figure 7 This curve represents the correlation between the phase shift of the low-frequency eddy current detection signal and the characteristic dimensions of localized corrosion, used to assist in determining the degree and morphological characteristics of localized corrosion. By establishing a calibration relationship between the low-frequency eddy current detection signal and the depth of localized corrosion, a quantitative estimation of the depth of localized corrosion in pipelines can be achieved.

[0056] Specifically, refer to Figure 8 As shown, Figure 8 A schematic flowchart of an exemplary amplitude and phase feature extraction method according to some embodiments of this application is shown. The method for extracting amplitude and phase features (specifically step 12) includes the following steps 12A to 12D.

[0057] 12A: Simultaneously or sequentially, an alternating current of a preset frequency is applied to a high-temperature differential eddy current probe encapsulated on the outer wall of the pipe. The eddy current signal corresponding to the surface oxide film and substrate corrosion is obtained by utilizing the skin effect's varying penetration depth at different frequencies. The preset frequency ranges from a low frequency of 100kHz to a high frequency of 1MHz.

[0058] 12B: The eddy current signal is de-DC and baseline corrected, and narrowband bandpass filtering is applied to eddy current signals of different frequencies. Then, wavelet denoising method is used to remove transient random noise from the eddy current signal.

[0059] Specifically, the original eddy current signal is subjected to DC bias removal and baseline correction based on the initial calibration data, so that the acquisition impedance under the non-corrosion state is used as the reference origin. Then, a narrowband bandpass filter with the corresponding center frequency is applied to each frequency channel. The filter bandwidth is preferably ±3% of the center frequency, while retaining the target frequency components and suppressing harmonics and electromagnetic noise. Then, Daubechies (db4 or db6) wavelet basis is used for multi-scale decomposition to remove transient random noise. The signal is reconstructed by using 3-5 layers of wavelet decomposition and applying a soft thresholding strategy, while retaining trend characteristics and amplitude and phase information. As a result, the eddy current signal after the above processing has a higher signal-to-noise ratio and is suitable for subsequent impedance analysis.

[0060] 12C: The conductivity change of the surface oxide film is inverted by monitoring the amplitude and phase of the impedance plane of the high-frequency eddy current signal, and the average thickness of the oxide film is generated according to the calibrated lookup table relationship.

[0061] Specifically, for high-frequency eddy current signals, i.e., for 1 MHz high-frequency signals, the changes in impedance amplitude, phase, and impedance plane trajectory are calculated. That is, the formation of oxide film changes the surface conductivity and permeability, causing the impedance to shift towards higher reactance. Then, a lookup table relationship between oxide film thickness and impedance characteristics is obtained, thereby obtaining the average oxide film thickness. The calibrated lookup table relationship can be obtained by using existing measurement methods (such as cross-sectional metallographic / SEM instruments) to obtain the actual thickness of the oxide film of the same material as the liquid lead-bismuth loop pipe of this application. At the same time, the corresponding impedance characteristics (such as the amplitude and phase changes of the impedance plane) are measured by existing high-temperature differential eddy current probes. This process is repeated multiple times, and the average value is taken to establish the corresponding relationship, generating the calibrated lookup table relationship.

[0062] 12D: By analyzing the amplitude and phase of low-frequency eddy current signals, the distortion of the eddy current field caused by local corrosion is identified. The size of local corrosion is matched with the differential characteristics and calibration curves. Then, the change in the remaining wall thickness caused by uniform corrosion of the pipe wall is obtained by analyzing the trend of the impedance change of low-frequency eddy current signals, and low-frequency wall thickness trend characteristics are generated.

[0063] Specifically, since the eddy currents corresponding to low-frequency excitation (e.g., 100 kHz) have a large electromagnetic penetration depth, their induced eddy currents can penetrate the surface oxide film and couple with the base metal at a greater depth, thus maintaining high sensitivity to the heterogeneity and localized corrosion configuration of the material. Furthermore, when pitting corrosion, intergranular corrosion, or localized thinning occurs in the pipe matrix, the conductivity and permeability of the local area will change, causing the original eddy current streamlines to deflect. This results in a significant deviation of the complex impedance signal measured by the differential eddy current probe from the baseline of the non-corrosion state. This deviation is usually manifested as changes in the amplitude and phase of the low-frequency eddy current signal, as well as the differential output (including differential amplitude and differential phase).

[0064] Therefore, this application extracts characteristic parameters such as amplitude change, phase change, and differential amount from the low-frequency eddy current channel; then, these characteristic parameters are matched with the calibration relationship obtained in advance through standard samples; the calibration relationship includes the correspondence curve between differential amplitude and corrosion depth, the correspondence curve between phase shift and corrosion depth, and an empirical database used to describe the variation law of differential output under different corrosion morphologies; by looking up the table, the approximate depth range and characteristic size of local corrosion can be inverted, thereby realizing the qualitative identification and quantitative estimation of local corrosion.

[0065] Furthermore, for scenarios where there are no significant local distortion features or where the differential output does not match the local corrosion model, the main characteristic is the slow change in the overall electromagnetic equivalent distance of the pipe wall. This application estimates the remaining wall thickness reduction caused by uniform corrosion by analyzing the trend change of low-frequency complex impedance in the time series. That is, since the wall thickness reduction will lead to an increase in the equivalent magnetic circuit distance between the probe and the effective conductor, the real and imaginary parts of the low-frequency impedance will show a trend of gradually increasing or monotonically changing phase. Therefore, by constructing the curve of low-frequency impedance change over time and combining it with the calibration relationship between wall thickness and impedance trend, the relative change amount or rate of change of the remaining wall thickness can be obtained, thereby generating the low-frequency wall thickness trend characteristics under uniform corrosion.

[0066] In an exemplary application of this application, for example: at a high frequency of 1 MHz, the impedance phase angle of the probe coil is measured to be -12.5°. At a low frequency of 100 kHz, the change in signal amplitude relative to the baseline is calculated to be +2.1%. Environmental parameters (from the corresponding solid-state oxygen sensor, temperature sensor, and flow meter): the dissolved oxygen concentration, after conversion using the Nernst formula, is... wt%. The temperature measured by the thermocouple was 452°C. The flow rate measured by the flow meter was 1.98 m / s.

[0067] refer to Figure 9 As shown, Figure 9A flowchart illustrating an exemplary equivalent circuit fitting method according to some embodiments of this application is shown. Specifically, the method for performing equivalent circuit fitting (specifically step 12) includes steps 12E to 12F.

[0068] For the above electrochemical impedance spectroscopy data, the following method was used: Figure 10 The equivalent circuit model shown is used to fit and analyze the electrochemical impedance spectroscopy data, wherein the equivalent circuit model includes solution resistance. Parallel branches related to oxide films and parallel branches related to interfacial reactions. For example... Figure 10 As shown, the equivalent circuit model exhibits good consistency with the experimentally measured electrochemical impedance data, thus the oxide film resistance can be obtained through the fitting process. Charge transfer resistance and oxide film equivalent capacitance Characteristic parameters, etc.

[0069] 12E: A preset equivalent circuit model is selected to simulate the solution resistance, the physical structure of the outer oxide film and the inner interface of the electrochemical impedance spectroscopy data, and a preset number of iterations are performed using the nonlinear least squares method to generate the fitting results.

[0070] The preset equivalent circuit model adopts the R(QRf)(QRct) model, and its circuit structure is as follows: The preset number of attempts is set by the designer according to actual needs, so that the error no longer decreases significantly.

[0071] in, The resistance of the solution is the resistance of the liquid lead-bismuth eutectic itself, which is closely related to temperature, impurity content, oxygen concentration, etc. CPE is the oxide film, which represents the capacitive behavior of the passivation film. Oxide film resistance represents the resistance encountered by ions as they pass through the film layer. It corresponds to the high-frequency semicircular fitting result of the oxide film impedance and reflects the ionic conductivity and density of the oxide film. The larger the oxide film, the denser it is and the stronger its protective effect. A rapid decline can cause the oxide film to rupture or dissolve, increasing the risk of corrosion. The double-layer oxide film CPE reflects the double-electrolytic layer capacitance behavior at the interface, that is, the double-layer capacitance behavior formed at the interface between the metal substrate and the liquid lead-bismuth eutectic. Charge transfer resistance (CRT) represents the ease with which electrons / ions can be exchanged between a metal and a liquid lead-bismuth eutectic. The larger the value, the lower the corrosion rate, and vice versa. The smaller the value, the faster the corrosion reaction.

[0072] 12F: Extract the oxide film resistance and charge transfer resistance from the fitting results, and calculate the oxide film capacitance based on these resistances. .

[0073] in, The oxide film capacitance is a key physical quantity characterizing the thickness and dielectric properties of the oxide film. It corresponds to the actual capacitance value equivalent to the constant phase element (CPE) representing the oxide film in the equivalent circuit. It is inversely proportional to the oxide film thickness and is therefore used to quantify whether the oxide film has become thicker, thinner, or degraded. For the amplitude parameters of CPE of oxide film; The exponential factor for the CPE of the oxide film typically ranges from 0.5 to 1.0. When n=1, the oxide film CPE is completely equivalent to an ideal capacitor; when n<1, it indicates that the oxide film surface or interface has roughness, porous structure, or uneven thickness. (This is from the embodiments of this application.) This is to convert the oxide film CPE into an equivalent ideal capacitor.

[0074] Therefore, for reference Figure 11 As shown, Figure 11 A comparison graph of electrochemical impedance spectroscopy experimental data and equivalent circuit fitting results is shown, where scatter points represent experimental measurement data and solid lines represent fitting results obtained based on the equivalent circuit model.

[0075] In an exemplary application of this application, for example: the three-electrode system probe completes a frequency scan, measuring electrochemical impedance spectroscopy data at 56 frequency points. Example data points for the high-frequency portion are: at 10000 Hz, the measured impedance is Z = 0.82 + j0.15 Ω (where j is the imaginary unit); at 1000 Hz, the measured impedance is Z = 15.6 + j22.8 Ω. Example data points for the low-frequency portion are: at 0.01 Hz, the measured impedance is Z = 218.5 + j185.2 Ω. Then, the R(QRf)(QRct) model is selected to simulate the physical structure of the solution resistance, the outer oxide film, and the inner interface, i.e., the circuit structure is as follows: Then, based on the shape of the spectrum, manually set the initial values: , , , , , , .

[0076] Then, a nonlinear least squares method is used for iterative fitting. The specific iterative fitting process is as follows: In the first iteration, a theoretical curve is calculated using the initial parameters and compared with 56 frequency points. The total error (sum of squared residuals) is calculated to be 1050. In the second iteration, the parameters are adjusted, such as... Increase to 120Ω Increasing the Ω to 1200, the error was recalculated and reduced to 620. Iteration continued, gradually descending the parameter space to find the minimum error. In the 15th iteration, the error no longer decreased significantly, the fit converged, and the final error (chi-square value) was... This indicates that the theoretical curve almost coincides with the experimental point; therefore, the final parameters can be read from the fitting results:

[0077] , , , , , , Then, the oxide film capacitance is calculated:

[0078] F. Thus, we have obtained multidimensional feature data for the current moment containing eight specific values: [ , , 1.12e F, eddy phase = -12.5°, eddy amplitude change = +2.1%, oxygen concentration = (wt%, temperature = 452°C, flow rate = 1.98 m / s).

[0079] 13: Input the multidimensional feature data into a multi-input deep learning model. This multi-input deep learning model adopts an attention mechanism and establishes a nonlinear correlation between environmental parameters and material corrosion response, thereby outputting a quantitative index characterizing the corrosion state.

[0080] The quantitative indicators include at least one of the following: uniform corrosion rate, probability of localized corrosion and pitting depth, oxide film condition level and recommended oxygen concentration range.

[0081] refer to Figures 12 to 14 As shown, Figure 12 The curve showing the relationship between oxide film resistance and operating time, obtained by fitting an equivalent circuit, is used to reflect the densification and stability evolution characteristics of the oxide film. Figure 13 The curve showing the relationship between charge transfer resistance and operating time, obtained by fitting an equivalent circuit, is used to characterize the changing trend of the corrosion reaction kinetics process. Figure 14 The curve showing the relationship between the equivalent capacitance of the oxide film and operating time, obtained by fitting the equivalent circuit, is used to reflect the changes in oxide film thickness and dielectric properties.

[0082] refer to Figure 15 As shown, Figure 15The diagram illustrates a flowchart of an exemplary multi-input deep learning model construction method according to some embodiments of this application. Specifically, the method for building the multi-input deep learning model (specifically step 13) includes the following steps 13A to 13E.

[0083] 13A: Retrieve historical data of at least one monitoring location in the liquid lead-bismuth loop pipeline within a preset time window from the historical database and generate corresponding time-series characteristic data based on the historical data.

[0084] The preset time window is set by the designer according to actual needs; the timing characteristic data includes oxide film resistance, interface charge transfer resistance, oxide film capacitance, eddy current phase, eddy current amplitude change, dissolved oxygen concentration, temperature and flow rate.

[0085] 13B: A Long Short-Term Memory (LSTM) artificial neural network is used. A preset number of hidden units are set for the LSTM artificial neural network, and the time series feature data is input into the LSTM artificial neural network. The dynamic changes of the time series are encoded through the forget gate, input gate and output gate to generate time series features that characterize the corrosion state.

[0086] The number of hidden units in an LSTM is a crucial design parameter. Too few units prevent the model from learning complex temporal dynamics, while too many units lead to overfitting and wasted computational resources. Therefore, based on the dimensionality and problem complexity of the input time-series feature data, grid search or random search was performed within typical value ranges such as [16, 32, 64, 128]. Using historical datasets, models with different numbers of hidden units were trained, and their performance was evaluated on the retained validation set. The main evaluation metrics were mean squared error (MSE) and prediction accuracy. Through comparison, it was found that when the number of hidden units was 32, the model achieved the lowest prediction error for corrosion rate and oxide film state on the validation set, and the model complexity and generalization ability were optimally balanced. When the number of units increased to 64, the validation set error no longer decreased significantly and even showed a slight increase (indicating overfitting). Therefore, in this embodiment, the preset number is 32.

[0087] In this long short-term memory artificial neural network, forget gate, input gate and output gate operations are performed on the temporal feature data at each time step. The forget gate determines whether to retain the previous state, the input gate selectively writes new corrosion-related features, and the output gate generates the corresponding hidden state, thereby gradually forming temporal features that represent the corrosion state.

[0088] 13C: The original temporal features of the corresponding modalities are mapped to a common feature space of the same dimension through a preset encoder, and the similarity between the temporal features of each modality is calculated through an attention mechanism to generate attention weights.

[0089] The preset encoder includes an electrochemical mode encoder, an eddy current detection mode encoder, and an environmental mode encoder. The electrochemical mode encoder, the eddy current detection mode encoder, and the environmental mode encoder are trained separately to map the original features of electrochemical modes, eddy current detection modes, and environmental modes with different dimensions and different physical meanings to a common feature space of the same dimension.

[0090] The encoder refers to a feedforward neural network that converts the original modal feature vectors into high-order feature representations required by the deep learning model. Its core idea is to design customized encoding paths for modal data with different physical properties. In the embodiments of this application, the three encoders, namely the electrochemical modal encoder, the eddy current detection modal encoder, and the environmental modal encoder, are trained separately. They are part of the entire end-to-end model. Their function is to map all the original features with different dimensions and different physical meanings to the same 32-dimensional common feature space. In this common feature space, the features from different modalities have comparable characteristics, thus serving as the basis for the weighted fusion of the subsequent attention mechanism.

[0091] 13D: Perform weighted fusion of temporal features of different modalities according to the attention weights to generate fused temporal features that reflect the collaborative relationship of multiple modalities.

[0092] 13E: Input the fused temporal features into the preset prediction branch and output the uniform corrosion rate, the probability of local corrosion and the pitting depth, the oxide film state level and the recommended oxygen concentration range.

[0093] Specifically, the preset prediction branches include a corrosion rate regression branch, a localized corrosion risk hybrid branch, an oxide film state classification branch, and a recommended oxygen concentration regression branch. The corrosion rate regression branch processes the fused temporal features through a linear regression layer to output the current uniform corrosion rate. The localized corrosion risk hybrid branch outputs the probability of localized corrosion occurrence through a first linear layer and a Sigmoid activation function, and outputs the predicted value of localized pitting depth through a second linear layer. The oxide film state classification branch outputs the state levels of oxide film stability, degradation, and failure through a linear layer and a Softmax activation function. The recommended oxygen concentration regression branch generates endpoint values ​​for the oxygen concentration range through a linear layer and a Tanh activation function, and scales these endpoint values ​​to a preset physical range to output the recommended oxygen concentration range.

[0094] refer to Figure 16 As shown, Figure 16A flowchart illustrating an exemplary method for generating fusion temporal features according to some embodiments of this application is shown. In some examples of this application, the method for generating fusion temporal features reflecting multimodal cooperative relationships (specifically steps 13C and 13D) includes steps 13F to 13J.

[0095] 13F: Asking Questions .in, To query the weight matrix, This represents temporal features of environmental modalities, such as dissolved oxygen concentration, temperature, and flow rate. The query weight matrix is ​​included. Used for analyzing environmental modal temporal features Feature transformation is performed to extract key information related to the evolution of pipeline corrosion. Therefore, through... The mapping transforms the original time-series features characterizing changes in operating temperature, oxygen concentration, and flow conditions within the environmental modal into query vectors. This allows us to pinpoint the key concerns of current operating conditions in assessing corrosion status.

[0096] For example, a question (Query, Q) is posed: "What are the temporal characteristics of the current environment modality (...)?" Under what conditions is the modality most important? (13G: Collection Options:) , , .in, The value of electrochemical mode timing characteristics; To query the weight matrix; These are the temporal characteristics of electrochemical modes, such as oxide film resistance, interfacial charge transfer resistance, and oxide film capacitance. The value of eddy current mode time series characteristics These are the temporal characteristics of the eddy current modes, such as eddy current phase and amplitude variations; The value of environmental modal temporal features. Among them, the query weight matrix... This is used to map the temporal features of different modes to a unified feature correlation space, so as to characterize the degree of response of each mode feature to corrosion assessment under the current operating state. This is achieved through... Mapping, electrochemical mode temporal characteristics eddy current mode timing characteristics and environmental modal temporal characteristics Generate the corresponding key vectors respectively , and This allows features with different physical meanings and dimensions to be correlated within the same space.

[0097] For example, when collecting options (Key, K), each modal temporal feature states its own value.

[0098] 13H: Calculate the dot product similarity between each question and each collection option to generate an initial score: , , .in The initial score for the time-series characteristics of electrochemical modes. The initial score for the temporal characteristics of the eddy current modes. This is the initial score for the temporal characteristics of the environmental modality.

[0099] 13I: The initial scores are normalized into attention weights using the Softmax function: , , .in, Attention weights for the temporal characteristics of electrochemical modes. Attention weights for the temporal features of eddy current modes. The attention weights are for the temporal features of the environmental modality.

[0100] 13J: Based on the attention weights, perform weighted fusion of temporal features from different modalities to generate fused temporal features reflecting multimodal collaborative relationships: .

[0101] 14: Evaluate and output the quantitative results of the liquid lead-bismuth loop pipeline based on this quantitative index.

[0102] Specifically, when outputting quantitative results, the results can be sent to the terminal device for interface visualization, providing corrosion evolution trend charts, risk warnings, and operation guidance. That is, the terminal device interface will display "Good condition, low corrosion rate, stable oxide film, it is recommended to maintain the current operating parameters or fine-tune the oxygen concentration to the recommended percentage to optimize energy consumption".

[0103] Therefore, this application specifically selects and combines a three-electrode probe, a high-temperature differential eddy current probe, a solid oxygen sensor, a temperature sensor, and a flow meter for online synchronous monitoring of the corrosion mechanism of liquid lead-bismuth eutectic. Then, for training in the liquid lead-bismuth eutectic environment, a multi-input deep learning model is developed that can deeply integrate environmental parameters such as oxygen concentration and temperature gradient with material response signals such as electrochemical impedance and eddy current signals. This multi-input deep learning model not only outputs the uniform corrosion rate, but also outputs the probability of local corrosion, pitting depth, and oxide film state level, and provides multi-dimensional quantitative results for oxygen concentration control suggestions.

[0104] Therefore, this application can explain corrosion behavior from a mechanistic perspective, with higher quantification accuracy than single wall thickness measurement methods. By fusing multimodal data, it can offset the performance fluctuations of a single sensor under high temperature and irradiation, improving operational stability. Furthermore, it can provide early warning of localized corrosion and oxide film failure, realizing the transformation from post-measurement to pre-prediction. The directly output oxygen concentration control suggestions close the monitoring and control loop, achieving proactive corrosion protection.

[0105] In an exemplary application of this application, for example: historical data of the monitoring point over the past 5 hours (from hour 995 to hour 999) is retrieved from a historical database, and corresponding time-series feature data is generated based on the historical data. The specific time-series feature data is shown in the table below:

[0106] Table: Time Series Data Matrix (6 Time Steps × 8 Features):

[0107]

[0108] The Long Short-Term Memory (LSTM) artificial neural network is configured with 32 hidden units, meaning it contains 32 memory cells. The processing time step t = 995: Input: X = [190.1, 1700.5, ..., 2.2]. (8-dimensional), initial hidden state h_994 and cell state c_994 are 0. Forget gate: Check X, decide to clear most of the initial memory. Input gate: Discover This is an important starting point, writing the new cell state c_995. Output gate: Based on the new state, output the first hidden state h_995 (a 32-dimensional vector). Processing time step t=996: Input: New data X=[195.5, 1725.8, ...] and the previous state h_995, c_995. The Long Short-Term Memory artificial neural network notices... The increase from 190.1 to 195.5 is a positive sign, reinforcing the idea that... The memory is increased and the cell state is updated to c_996, output h_996. This process continues from t=997 to t=999. Processing the current time step t=1000: Input: X=[215.4, 1850.7, ...] and h_999, c_999. At this point, the cell state c_1000 has been strongly memorized. The trend pattern shows a steady increase over the past 5 hours. The Long Short-Term Memory (LSTM) artificial neural network outputs the final hidden state h_1000, which is a 32-dimensional feature vector V_lstm containing the dynamic trend over the past 6 hours.

[0109] In the exemplary application of this application, the workflow of the attention fusion layer of the Long Short-Term Memory artificial neural network is as follows: It obtains the following results through different encoders, such as electrochemical mode encoders, eddy current detection mode encoders, and environmental mode encoders: =[0.12, 0.85, -0.03, ..., 0.45] (32 dimensions, representing electrochemical characteristics); =[-0.22, 0.15, 0.08, ..., -0.11] (32 dimensions, representing eddy current characteristics); =[0.05, 0.92, 0.01, ..., 0.33] (32 dimensions, representing environmental characteristics); Propose a question (Query, Q), "Under the current operating conditions (...)..." Under what conditions is the mode most important? After computation, Q is a vector. Collect options (Key, K), each modality stating its own value. , , Initial score, calculate the similarity (dot product) between Q and each K. ; ; =1.7; Through a normalization (Softmax) operation, the score is converted into a percentage-based attention weight:

[0110] ; ;

[0111] Therefore, the attention mechanism concludes that, under current stable operating conditions, 50% should be relied upon for the electrochemical signal, 35% for environmental parameters, and 15% for the eddy current signal. The final decision-making basis (fusion vector) is: This is a smart summary that integrates all the information but highlights the most important information at present.

[0112] This intelligent summary is fed into four dedicated output layers: the overall erosion rate branch (regression), and the execution: output = LinearLayer( The result outputs a scalar value of 0.83, meaning that after unit conversion, the model predicts the current uniform corrosion rate to be 0.83 μm / year. For the localized corrosion risk branch (mixed), execute: [probability, depth] = [Sigmoid(LinearLayer1( LinearLayer2 The result is [0.15, 5.2], indicating a 15% probability of localized corrosion. If it occurs, the predicted maximum pit depth is approximately 5.2 μm. The oxide film state rating branch (classification) is executed as follows: [P_Stable, P_Degraded, P_Failed] = Softmax(LinearLayer( The result is an output probability [0.85, 0.15, 0.0]. Because the probability of stability (85%) is higher than others, the final rating is stable. The recommended branch is oxygen concentration (regression), executed as follows: [min, max] = Tanh(LinearLayer( Then scale it to a reasonable range; the result is the output [1.8]. 2.5 This suggests maintaining the oxygen concentration at 1.8. Up to 2.5 Within the range of wt%.

[0113] Therefore, the quantitative results are visualized, providing data such as corrosion evolution trend charts, risk warnings, and operational guidance. The interface displays: "Good condition, low corrosion rate, stable oxide film. It is recommended to maintain the current operating parameters, or the oxygen concentration can be fine-tuned to 2.0." wt% to optimize energy consumption.

[0114] Therefore, for reference Figure 17 It can be seen that this application can achieve continuous quantification of the evolution trend of corrosion rate with operating time, that is... Figure 17 This is an evolution curve of the corrosion rate of pipeline materials in a liquid lead-bismuth environment as a function of operating time. It is used to solve the problem that existing technologies can only obtain the instantaneous corrosion state and cannot reflect the overall evolution trend of the corrosion process, thereby realizing a continuous quantitative characterization of the degree of pipeline corrosion. Figure 17 The trend is as follows: the initial rate is high (the oxide film is not stable), the rate decreases and stabilizes in the middle stage (the protective film is formed), and the rate slowly recovers in the later stage (the film degrades / local damage), which is the safest trend in liquid lead-bismuth eutectic corrosion.

[0115] refer to Figures 18 to 20 It is evident that this application comprehensively considers various operating parameters such as operating temperature, oxygen concentration, and fluid flow rate. Figure 18 This is a time series graph showing the temperature parameters of the pipeline during the operation of a liquid lead-bismuth circuit. It is used to address the problem that environmental conditions were not fully considered during corrosion quantification, leading to deviations between corrosion assessment results and actual operating conditions. Figure 19 This is a time series diagram showing the change of oxygen concentration in the system over time during the operation of a liquid lead-bismuth circuit. It is used to address the difficulty in quantifying the impact of oxygen content fluctuations in liquid metal environments on oxide film stability and corrosion behavior. Figure 20This is a time series diagram showing the change of fluid velocity in the pipeline over time during the operation of a liquid lead-bismuth circuit. It is used to address the problem that the impact of changes in flow conditions on corrosion behavior is difficult to reflect in traditional corrosion monitoring methods.

[0116] refer to Figure 21 It can be seen that the contribution weights of electrochemical characteristics, local corrosion response characteristics, and operating condition parameters in the model change dynamically in different corrosion stages, thereby enhancing the interpretability of the model. Figure 21 This diagram illustrates how the attention weights of different modal input features in the model change over time. It demonstrates that the model's attention to electrochemical, eddy current, and environmental parameters dynamically changes at different operational stages, supported by the algorithm and not a black box. This approach addresses the issues of unclear contribution of multi-source monitoring data to corrosion assessment and insufficient model interpretability.

[0117] refer to Figure 22 It can be seen that the corrosion prediction results obtained by the method of this application have good consistency with the calibration results, which verifies the effectiveness of the method. Figure 22 This is a comparison chart of corrosion rate prediction results obtained based on multimodal input features and calibration results obtained through experiments or calibration over time, used to address the problems of insufficient reliability of corrosion prediction results and difficulty in verifying the prediction results with the actual corrosion state. Figure 22 This means that the overall trend is highly consistent, there is noise, but it is not divergent.

[0118] refer to Figure 23 As shown, Figure 23 A schematic flowchart of an exemplary ultrasonic signal processing method according to some embodiments of this application is shown. In some embodiments of this application, the steps 12G to 13J are included when generating multidimensional feature data (specifically step 12).

[0119] 12G: Acquire ultrasonic signals from probes that penetrate at least partially into high-temperature sections, bends, and areas with temperature gradients in the liquid lead-bismuth loop piping during planned shutdowns or preset maintenance cycles.

[0120] 12H: Calculate the oxide film thickness, local corrosion depth, and wall thickness trend characteristics at the corresponding locations in the liquid lead-bismuth circuit pipeline based on the acoustic time of the ultrasonic signal.

[0121] 12I: The quantization results of the eddy current signal are calibrated based on the oxide film thickness, local corrosion depth, and wall thickness trend characteristics of the ultrasonic signal.

[0122] refer to Figure 24 As shown, Figure 24A schematic diagram of the connection of an exemplary corrosion quantification device according to some embodiments of this application is shown. Therefore, this application also provides a corrosion quantification device for a liquid lead-bismuth loop pipeline, comprising: a corrosion quantification device and a sensor cluster.

[0123] The sensing cluster includes a three-electrode probe system, a high-temperature differential eddy current probe, a telescopic in-situ multi-parameter probe, a solid-state oxygen sensor, a temperature sensor, and a flow meter. The three-electrode probe system comprises a working electrode, a reference electrode, and a counter electrode. The working electrode is made of the same material as the liquid lead-bismuth circuit pipe. The reference electrode uses a yttrium-stabilized zirconia electrolyte tube encapsulated with Bi / Bi₂O₃ or Pb / PbO as the reference material. The counter electrode is made of platinum or tungsten, thereby acquiring electrochemical impedance spectroscopy data at a predetermined location in the liquid lead-bismuth circuit pipe. The high-temperature differential eddy current probe is encapsulated at a predetermined location in the liquid lead-bismuth circuit pipe to acquire eddy current signals at a predetermined frequency at that location. The telescopic in-situ multi-parameter probe remains closed during normal operation and, during operation, extends at least partially to be inserted into and fixed at a predetermined location in the liquid lead-bismuth circuit pipe to acquire ultrasonic, acoustic emission, and frictional signals at that location. The solid-state oxygen sensor acquires dissolved oxygen concentration data at a predetermined location in the liquid lead-bismuth circuit pipe. The temperature sensor acquires temperature data at a predetermined location in the liquid lead-bismuth circuit pipe. This flow meter is used to obtain flow velocity data at a preset location in a liquid lead-bismuth loop pipeline.

[0124] In the exemplary application of this application, the three-electrode probe system is, for example, in-situ integrated into the inner wall or bypass test section of the high-temperature section, bend, or temperature gradient region of the liquid lead-bismuth loop pipe. The working electrode is made of the same material as the liquid lead-bismuth loop pipe (e.g., ferritic / martensitic steel T91) and formed into a small probe shape; its surface condition represents the inner wall condition of the liquid lead-bismuth loop pipe. The reference electrode is a solid-state reference electrode suitable for high-temperature liquid metals, specifically encapsulating a reference material (e.g., Bi / Bi₂O₃ or Pb / PbO mixture) with a specific oxygen partial pressure in a stable electrolyte (e.g., yttrium-stabilized zirconium oxide, YSZ) tube. This electrolyte tube is in eutectic contact with the liquid lead-bismuth to form a stable electrochemical potential reference. For the counter electrode, an inert material (e.g., platinum or tungsten) wire or rod is used.

[0125] The measurement process of this three-electrode probe involves periodically (e.g., every 4 hours) or continuously applying a small sinusoidal potential perturbation (e.g., ±10 mV), typically in the frequency range of 10 kHz to 10 mHz, to the working electrode during operation. The current signal in response of the working electrode is then measured. The resulting operational characteristics include oxide film resistance, oxide film capacitance, and charge transfer resistance. Specifically, the oxide film resistance is determined by fitting the obtained electrochemical impedance spectroscopy data using an equivalent circuit model of R(QRf)(QRct). This allows for the separation of the inner dense oxide film resistance corresponding to high-frequency capacitive arcs and the outer loose oxide film resistance corresponding to mid-frequency capacitive arcs. This resistance value is directly inversely proportional to the ionic conductivity of the oxide film, i.e., its protective capability. An increase in resistance indicates a dense oxide film with good protection; a sharp decrease in resistance indicates oxide film rupture or dissolution. The oxide film capacitance can be obtained from the same equivalent circuit fitting, which yields the capacitance value of the constant phase angle element corresponding to the oxide layer. The capacitance value is inversely proportional to the thickness of the oxide film (C∝1 / d). Therefore, the continuous increase in capacitance value can quantitatively characterize the process of oxide film thinning due to corrosion. The charge transfer resistance represents the ease with which charge transfer (i.e., corrosion reaction) occurs between the metal substrate and the liquid lead-bismuth eutectic. This resistance value is directly inversely proportional to the uniform corrosion rate and is a key parameter for quantifying the overall corrosion rate.

[0126] In the exemplary application of this application, the high-temperature differential eddy current probe is, for example, permanently encapsulated on the outer wall of the pipe in the high-temperature section, bend, and temperature gradient region of the liquid lead-bismuth loop pipeline (non-contact, avoiding direct contact with the liquid lead-bismuth eutectic); the probe coil needs to be made of a high-temperature radiation-resistant insulating material (such as ceramic encapsulation). The measurement process of this high-temperature differential eddy current probe involves simultaneously or sequentially passing multiple alternating currents of various frequencies (e.g., 100 kHz, 400 kHz, 1 MHz) onto the probe. Eddy currents of different frequencies have different skin effect penetration depths (δ∝1 / √f) in the material. Therefore, the operable features obtained include oxide film thickness estimation and substrate corrosion and defect detection. Specifically, in oxide film thickness estimation, high-frequency eddy currents (e.g., 1 MHz) have shallow penetration depth and are primarily sensitive to changes in the conductivity / permeability of the surface oxide film. By monitoring changes in the impedance plane (amplitude and phase) of the high-frequency signal, the average thickness of the oxide film can be inferred, which can then be cross-validated with the capacitance value measured by electrochemical impedance spectroscopy. In substrate corrosion and defect detection, low-frequency eddy currents (e.g., 100 kHz) have deep penetration depth, capable of penetrating the oxide film to detect the substrate metal. Localized corrosion (pitting, intergranular corrosion) leads to inhomogeneity in the substrate material, causing distortion of the eddy current field. By analyzing the amplitude and phase characteristics of the differential signal, the presence of localized corrosion can be identified, and its approximate depth can be estimated using calibration curves. Although the quantification accuracy is not as high as that of ultrasound, the overall impedance change of the low-frequency eddy current signal is related to the effective distance from the probe to the conductive substrate (i.e., the remaining wall thickness), which can be used for monitoring and supplementing verification of wall thickness reduction trends, thereby obtaining data on wall thickness reduction caused by uniform corrosion.

[0127] In the exemplary application of this application, the telescopic in-situ multi-parameter probe is, for example, pre-installed with a hydraulically or electrically driven telescopic locking mechanism at high-temperature sections, bends, and areas with temperature gradients in the liquid lead-bismuth loop pipeline. The measurement process of this telescopic in-situ multi-parameter probe involves maintaining a closed loop during normal operation to ensure a sealed loop; during planned shutdowns or specific maintenance cycles, a probe rod integrating multiple micro-sensors is inserted and fixed by an operator, causing its sensor head to contact the inner wall of the pipeline or the surface of a pre-placed standard sample. This telescopic in-situ multi-parameter probe integrates a high-frequency ultrasonic thickness measurement module, a micro-force scratching module, and a contact resistance probe. The high-frequency ultrasonic thickness measurement module uses a miniature delay block probe to directly emit and receive ultrasonic waves, accurately calculating the local remaining wall thickness and the depth of corrosion pits through acoustic time. The micro-force scratching module integrates a diamond indenter in the probe head, which scratches the surface with a constant or increasing load while monitoring the acoustic emission signal and frictional force. The critical load corresponds to the point where the oxide film begins to peel off, used to quantify the bonding force between the oxide film and the substrate. The contact resistance probe uses the probe head itself as a tiny resistive element, whose resistance increases with corrosion consumption, providing corrosion depth data with micron-level accuracy.

[0128] Thus, the operational features obtained include absolute quantification of geometric dimensions, oxide film mechanical properties, and high-precision corrosion data. Specifically: in acquiring absolute quantification of geometric dimensions, accurate wall thickness reduction and localized corrosion depth, unaffected by material properties, are provided for calibrating online eddy current and ultrasonic measurements. In acquiring oxide film mechanical properties, the critical load value is a key mechanical indicator for evaluating the protective performance of the oxide film and its susceptibility to peeling under thermal stress. In acquiring high-precision corrosion data, contact resistance probes provide near real-time and accurate corrosion rate data; although acquired periodically, this data can be used to correct and verify the accuracy of continuous online monitoring.

[0129] In some embodiments, reference Figure 25 As shown, Figure 25 A schematic diagram of the connection of an exemplary corrosion quantification apparatus according to some embodiments of this application is shown. The corrosion quantification apparatus 20 includes a memory 21 and a processor 22. The memory 21 stores a computer program that can run on the processor 22. When the processor 22 executes the computer program, it implements the methods described in the above embodiments. The number of memories 21 and processors 22 can be one or more.

[0130] The electronic device 3 also includes a communication interface 23 for communicating with external devices and transmitting data.

[0131] If the memory 21, processor 22, and communication interface 23 are implemented independently, they can be interconnected via a bus to communicate with each other. This bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus, etc. This bus can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 25 The bus is represented by a single thick line, but this does not mean that there is only one bus or one type of bus.

[0132] Optionally, in a specific implementation, if the memory 21, processor 22 and communication interface 23 are integrated on a single chip, the memory 21, processor 22 and communication interface 23 can communicate with each other through an internal interface.

[0133] This application provides a computer-readable storage medium storing a computer program that, when executed by a processor 302, implements the method provided in this application.

[0134] This application also provides a chip, which includes a processor 22 for calling and running instructions stored in a memory 21, so that a communication device equipped with the chip executes the method provided in this application.

[0135] This application also provides a chip, including: an input interface, an output interface, a processor 22 and a memory 21. The input interface, the output interface, the processor 22 and the memory 21 are connected through an internal connection path. The processor 22 is used to execute code in the memory 21. When the code is executed, the processor 22 is used to execute the method provided in the application embodiment.

[0136] It should be understood that the processor 22 mentioned above can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. General-purpose processors can be microprocessors or any conventional processor. It is worth noting that processor 22 can be a processor 22 supporting Advanced Reduced Instruction Set Machines (ARM) architecture.

[0137] Furthermore, the aforementioned memory 21 may include read-only memory and random access memory, and may also include non-volatile random access memory. The memory may be volatile or non-volatile, or may include both. Non-volatile memory may include read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), or flash memory. Volatile memory may include random access memory (RAM), which serves as an external cache. Many forms of RAM are available by way of example, but not limitation. Examples include Static Random Access Memory (SRAM), Dynamic Random Access Memory (DRAM), Synchronous DRAM (SDRAM), Double Data Rate SDRAM (DDR SDRAM), Enhanced Synchronous DRAM (ESDRAM), Synchlink DRAM (SLDRAM), and Direct Rambus RAM (DR RAM).

[0138] In the above embodiments, implementation can be achieved, in whole or in part, through software, hardware, firmware, or any combination thereof. When implemented in software, it can be implemented, in whole or in part, as a computer program product. A computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the flow or function according to this application is generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transferred from one computer-readable storage medium to another.

[0139] Other embodiments of this application will readily conceive of by those skilled in the art upon consideration of the specification and practice of the disclosure herein. This application is intended to cover any variations, uses, or adaptations of this application that follow the general principles of this application and include common knowledge or customary techniques in the art not disclosed herein. The specification and embodiments are to be considered exemplary only, and this application is not limited to the precise structures described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope.

Claims

1. A method for quantifying corrosion in liquid lead-bismuth loop pipelines, characterized in that, The method includes: Environmental parameters characterizing corrosion drivers and electrochemical and eddy current signals characterizing material response behavior are acquired at at least one monitoring location in the liquid lead-bismuth circuit to generate multimodal data. The multimodal data is preprocessed and fitted with equivalent circuits to generate multidimensional feature data including surface and substrate corrosion characteristics, oxide film characteristics, interface reaction characteristics, and environmental characteristics. The multidimensional feature data is input into a multi-input deep learning model employing an attention mechanism to establish a nonlinear correlation between environmental parameters and material corrosion response. The model outputs quantitative indicators characterizing the corrosion state, wherein these quantitative indicators include at least one of uniform corrosion rate, probability of localized corrosion and pitting depth, oxide film state level, and recommended oxygen concentration range; and The quantitative results of the liquid lead-bismuth circuit are evaluated and output based on the quantitative indicators.

2. The method according to claim 1, characterized in that, The method for generating multidimensional feature data including surface and substrate corrosion characteristics, oxide film characteristics, interface reaction characteristics, and environmental characteristics includes: Acquire dissolved oxygen concentration, temperature, flow rate, eddy current signal, and electrochemical impedance spectroscopy data from sensor clusters located in high-temperature sections, bends, and areas with temperature gradients in the liquid lead-bismuth circuit; The dissolved oxygen concentration, temperature, and flow rate were time-aligned and normalized. The eddy current signal is denoised and its amplitude and phase features are extracted to generate surface oxide film thickness, local corrosion depth, and low-frequency wall thickness trend features; and The electrochemical impedance spectroscopy data were fitted with an equivalent circuit to extract the characteristics of oxide film resistance, oxide film capacitance, and charge transfer resistance.

3. The method according to claim 1 or 2, characterized in that, The equivalent circuit fitting includes: A preset equivalent circuit model is selected for the electrochemical impedance spectroscopy data to simulate the physical structure of the solution resistance, outer oxide film and inner interface, and a preset number of iterations are performed using the nonlinear least squares method to generate fitting results. Extract the oxide film resistance and charge transfer resistance from the fitting results, and calculate the oxide film capacitance based on the oxide film resistance and charge transfer resistance: ， in For oxide film capacitors, For the amplitude parameters of the oxide film CPE, The exponential factor of CPE in oxide film, Oxide film resistance; where, The preset equivalent circuit model is the R(QRf)(QRct) model, and its circuit structure is as follows: ， in For the resistance of the solution, For oxide film CPE, It is a double-layer CPE. It is a charge transfer resistor.

4. The method according to claim 2, characterized in that, The method for extracting amplitude and phase features includes: Simultaneously or sequentially, an alternating current of a preset frequency is passed into a high-temperature differential eddy current probe encapsulated on the outer wall of the pipe, and the skin effect penetration depth at different frequencies is used to obtain eddy current signals corresponding to the surface oxide film and substrate corrosion. The preset frequency includes frequencies from 100kHz low frequency to 1MHz high frequency. The eddy current signal is subjected to DC removal and baseline correction, and narrowband bandpass filtering is applied to eddy current signals of different frequencies. Then, wavelet denoising method is used to remove transient random noise from the eddy current signal. The conductivity variation of the surface oxide film is inverted based on the amplitude and phase of the high-frequency eddy current signal impedance plane, and the average oxide film thickness is generated according to the calibrated lookup table relationship; and Based on the amplitude and phase of the low-frequency eddy current signal, the distortion of the eddy current field caused by local corrosion is identified. By matching the local corrosion size with the differential characteristics and calibration curve, the remaining wall thickness change caused by uniform corrosion of the pipe wall is obtained based on the impedance change trend of the low-frequency eddy current signal, and low-frequency wall thickness trend characteristics are generated.

5. The method according to claim 2 or 4, characterized in that, The method for establishing the nonlinear correlation between environmental parameters and material corrosion response using a multi-input deep learning model with an attention mechanism includes: Historical data of at least one monitoring location in the liquid lead-bismuth circuit within a preset time window are retrieved from the historical database, and corresponding time-series characteristic data are generated based on the historical data. The time-series characteristic data includes oxide film resistance, interface charge transfer resistance, oxide film capacitance, eddy current phase, eddy current amplitude change, dissolved oxygen concentration, temperature, and flow rate. A preset number of hidden units are set for the long short-term memory artificial neural network, and the temporal feature data is input into the long short-term memory artificial neural network. The dynamic changes of the temporal feature data are encoded through the forget gate, input gate and output gate to generate temporal features representing the erosion state. The original temporal features of each modality are mapped to a common feature space of the same dimension through a preset encoder, and the similarity between the temporal features of each modality is calculated through an attention mechanism to generate attention weights; Based on the attention weights, a weighted fusion of temporal features from different modalities is performed to generate fused temporal features reflecting multimodal collaborative relationships; and The fused temporal features are input into preset prediction branches and output uniform corrosion rate, probability of local corrosion occurrence and pitting depth, oxide film state level and recommended oxygen concentration range.

6. The method according to claim 5, characterized in that, The Long Short-Term Memory (LSTM) artificial neural network performs forgetting gate, input gate, and output gate operations on the temporal feature data at each time step. When performing the forget gate operation, decide whether to retain the previous state; When performing an input gate operation, new erosion-related features are selectively written; When performing output gate operations, corresponding hidden states are generated, thereby gradually forming temporal characteristics that characterize the erosion state.

7. The method according to claim 5, characterized in that, The preset encoder includes an electrochemical mode encoder, an eddy current detection mode encoder, and an environmental mode encoder. The electrochemical mode encoder, the eddy current detection mode encoder, and the environmental mode encoder are trained separately to map the original features of electrochemical modes, eddy current detection modes, and environmental modes with different dimensions and different physical meanings to a common feature space of the same dimension.

8. The method according to claim 7, characterized in that, The method for generating attention weights includes: Question: ， in, To query the weight matrix, Temporal characteristics of environmental modes; Collection options: ， ， ， in, The value of electrochemical mode timing characteristics; This is the key-value weight matrix; These are the temporal characteristics of electrochemical modes, such as oxide film resistance, interfacial charge transfer resistance, and oxide film capacitance. The value of eddy current mode time series characteristics These are the temporal characteristics of the eddy current modes, such as eddy current phase and amplitude variations; The value of environmental modal temporal characteristics; Calculate the dot product similarity between each question and each collection option to generate an initial score: ， ， ， in The initial score for the time-series characteristics of electrochemical modes. The initial score for the temporal characteristics of the eddy current modes. Initial scores for the temporal characteristics of environmental modalities; The initial scores are normalized into attention weights using the Softmax function: ， ， ， in Attention weights for the temporal characteristics of electrochemical modes. Attention weights for the temporal features of eddy current modes. Attention weights for temporal features of environmental modalities; and Based on the attention weights, the temporal features of different modalities are weighted and fused to generate fused temporal features that reflect the collaborative relationships of multiple modalities: 。 9. The method according to claim 5, characterized in that, The preset prediction branches include a corrosion rate regression branch, a localized corrosion risk mixed branch, an oxide film state classification branch, and a recommended oxygen concentration regression branch, wherein... The corrosion rate regression branch processes the fused temporal features through a linear regression layer to output the current uniform corrosion rate; The localized corrosion risk hybrid branch outputs the probability of localized corrosion occurrence through the first linear layer and the Sigmoid activation function, and outputs the predicted value of localized pitting depth through the second linear layer. The oxide film state classification branch outputs the state levels of oxide film stability, degradation, and failure through a linear layer and a Softmax activation function; and The recommended oxygen concentration regression branch generates endpoint values ​​for the oxygen concentration range through a linear layer and a Tanh activation function, and scales these endpoint values ​​to a preset physical range to output the recommended oxygen concentration range.

10. The method according to claim 1, characterized in that, The generation of the multidimensional feature data also includes: When planning a reactor shutdown or a pre-set maintenance cycle, acquire ultrasonic signals from probes that penetrate at least partially into the high-temperature section, bends, and areas with temperature gradients in the liquid lead-bismuth loop. The oxide film thickness, local corrosion depth, and wall thickness trend characteristics at corresponding locations in the liquid lead-bismuth circuit are calculated based on the acoustic time of the ultrasonic signal; and The quantization results of the eddy current signal are calibrated based on the oxide film thickness, local corrosion depth, and wall thickness trend characteristics of the ultrasonic signal.

11. A corrosion quantification device for liquid lead-bismuth loop pipelines, characterized in that, include: At least one processor; At least one memory; The at least one memory is coupled to the at least one processor and is used to store instructions executed by the at least one processor, which, when executed by the at least one processor, cause the corrosion quantification device to perform the method according to any one of claims 1 to 10.

12. A corrosion quantification device for liquid lead-bismuth loop pipelines, characterized in that, include: The corrosion quantification device according to claim 11; The sensor cluster includes a three-electrode probe, a high-temperature differential eddy current probe, a telescopic in-situ multi-parameter probe, a solid-state oxygen sensor, a temperature sensor, and a flow meter. The three-electrode system probe includes a working electrode, a reference electrode, and a counter electrode. The working electrode is made of the same material as the liquid lead-bismuth circuit pipe. The reference electrode is encapsulated in a yttrium-stabilized zirconium oxide electrolyte tube using Bi / Bi2O3 or Pb / PbO as a reference material. The counter electrode is made of platinum or tungsten. This allows for the acquisition of electrochemical impedance spectroscopy data at a predetermined location in the liquid lead-bismuth circuit pipe. The high-temperature differential eddy current probe is encapsulated at a preset position in the liquid lead-bismuth circuit pipeline to obtain a preset frequency eddy current signal at the corresponding position. The telescopic in-situ multi-parameter probe remains closed during normal operation and, during operation, extends at least partially to be inserted into and fixed at a preset position in the liquid lead-bismuth circuit pipeline to acquire ultrasonic, acoustic emission, and frictional signals at the corresponding position. The solid-state oxygen sensor is used to acquire dissolved oxygen concentration data at a preset location in the liquid lead-bismuth circuit pipeline; The temperature sensor is used to acquire temperature data at a preset location in the liquid lead-bismuth loop pipeline. The flow meter is used to obtain flow velocity data at a preset location in the liquid lead-bismuth loop pipeline.

Images