Multi-mode collaborative corrosion state monitoring method and system

Abstract

The invention relates to the technical field of material analysis, in particular to a multi-mode collaborative corrosion state monitoring method and system. Through a self-supervised contrast learning method of acoustic emission and electrochemical noise dual-mode signals and in combination with an innovative network architecture of Transformer and a multi-layer perceptron, feature alignment of acoustic emission and electrochemical noise signals is realized, the problem of dependence on a large amount of labeled data in corrosion monitoring of chlor-alkali production equipment is overcome, and a corrosion state on-line monitoring model is formed. Under a self-supervised contrast learning framework, the model can utilize unlabeled acoustic emission and electrochemical noise data to carry out autonomous learning, and the demand for expensive and scarce labeled resources is reduced. Only a small amount of manual annotation acoustic emission data is needed for fine tuning training, and online monitoring of the corrosion state can be achieved. Meanwhile, the on-site rapid and non-destructive corrosion state monitoring analysis of the chlor-alkali production equipment can be completed only through acoustic emission data, and the corrosion state monitoring technology is promoted to develop towards the more intelligent and automatic direction.

Description

Technical Field

[0001] The present invention relates to the technical field of material analysis, and in particular to a multi-modal collaborative corrosion state monitoring method and system. Background Art

[0002] Chlor-alkali chemical production involves the use and treatment of multiple hazardous chemicals, such as hydrogen, chlorine, vinyl chloride, etc. These media are all classified as hazardous chemicals under national key supervision. Due to the complex and highly chemically active media in chlor-alkali chemical production, production equipment is usually exposed to corrosive environments of high chlorine (such as saturated brine, wet chlorine, hypochlorite) and strong acids and alkalis (such as hydrochloric acid and sodium hydroxide) for a long time. In order to cope with these harsh working conditions, corrosion-resistant materials such as 304, 316L stainless steel, TA2 titanium and Ni6 nickel are usually used. However, even these materials with strong corrosion resistance are still susceptible to corrosion by media such as chloride ions and alkali solutions during long-term service, resulting in damage phenomena such as pitting and stress corrosion. Corrosion damage not only reduces the service life of equipment and increases maintenance and replacement costs, but may also lead to safety accidents such as leakage and explosion, directly threatening the safety and stability of production. Therefore, corrosion has always been one of the major challenges facing the chlor-alkali chemical industry. In order to ensure the safe and efficient production of chlor-alkali, it is urgent to develop efficient and accurate monitoring technologies for corrosion problems so as to monitor and warn the corrosion status of equipment in real time. Through effective corrosion monitoring technology, timely measures can be taken before serious damage occurs to the equipment, thereby extending the service life of the equipment and reducing safety risks caused by corrosion, thereby ensuring the safety and continuity of production.

[0003] Electrochemical noise technology is based on real-time measurement of potential and current fluctuations that occur spontaneously during corrosion, thereby determining the corrosion rate and the corresponding mechanism. Its data contains rich information, and by analyzing the statistical characteristics of noise signals, rich corrosion information can be obtained. However, its signals are usually more complex, and its physical meaning is not easy to interpret directly. It requires complex data processing and analysis methods, which increases the difficulty of online monitoring of corrosion status and makes it difficult to arrange at the production site. Acoustic emission technology is based on the detection of elastic waves generated during mechanical damage to the surface of the specimen. The advantage is that the acoustic transducer is placed far away from the corrosion location, which can achieve non-destructive testing and is convenient for placement at the production site. The disadvantage is that the data interpretation of the sound signal is complex, and there may be a delay between the acoustic emission event and the expected corrosion process.

[0004] A single monitoring method can usually only reflect one aspect of the corrosion process and cannot fully and accurately assess the corrosion status. Combining electrochemical noise and acoustic emission signals can simultaneously obtain information on chemical reactions and physical changes, improving the overall understanding of the corrosion process. However, acoustic emission and electrochemical signals have different time-frequency characteristics. Currently, there is no method for the coordinated and accurate identification and monitoring of corrosion by acoustic emission and electrochemical noise, and there is a lack of correlation and mutual verification between the two signals. In addition, traditional corrosion analysis methods rely on a large amount of annotated data for supervised learning, which is not only time-consuming but also costly. The acquisition of annotated data is often limited to the experience and time of experts, which limits the ability to quickly and accurately perform corrosion analysis.

[0005] In view of the above problems, a multimodal collaborative corrosion state monitoring method and system are proposed. This method can reflect the corrosion state more comprehensively and accurately by aligning the feature space representation of electrochemical noise and acoustic emission bimodal signals at multiple levels and comprehensively analyzing the various feature information generated during the corrosion process. A self-supervised pre-training algorithm is also adopted to allow the model to reduce its dependence on a large amount of labeled data by self-learning the intrinsic characteristics of corrosion signals. The electrochemical noise information is stored in the model through the model learning process. In on-site monitoring, only an acoustic emission device needs to be set up to collect signals, thereby realizing on-site non-destructive monitoring. Summary of the invention

[0006] In response to the problems raised in the above background technology, the present invention provides a corrosion status monitoring method and system based on multimodal collaboration of electrochemical noise signals and acoustic emission signals, so as to achieve on-site rapid, non-destructive, accurate and automated monitoring of corrosion damage in chlor-alkali equipment.

[0007] To achieve the above purpose, the specific method of the solution provided by the present invention is as follows:

[0008] S1: Signal acquisition: The sample material includes but is not limited to stainless steel and carbon steel of various materials. It is placed in a corrosive solution to cause corrosion. The solution used is but not limited to NaCl solution and other corrosive solutions. An electrochemical acquisition device and an acoustic emission acquisition device are set on the specimen to collect electrochemical noise signals and acoustic emission signals at the same time.

[0009] S2: Dataset construction: Using a large number of collected electrochemical noise signals and acoustic emission signals, we first align the two signals in time steps to construct a large-scale acoustic-electric alignment pre-training dataset that does not require manual labeling; a small number of acoustic emission signals with obvious corrosion characteristics are manually labeled to construct a labeled dataset for the acoustic emission corrosion stage.

[0010] S3: Model construction and training: Construct an improved two-stage network. The first stage is a contrastive learning structure based on Transformer, and the second stage is a corrosion stage monitoring network composed of a Transformer-based feature extractor and a multi-layer perceptron-based classifier. Use the two data sets constructed in step S2 to train the two-stage network in turn.

[0011] S4: Model testing: The unknown acoustic emission test signal is judged by the improved corrosion stage monitoring model, the corrosion stage corresponding to the test signal is output, and the test effect of the corrosion monitoring model is obtained.

[0012] S5: System construction: Create an interactive interface, package the software, and install the software on the server; connect the on-site acoustic emission signal acquisition device with the packaged software to realize the on-site corrosion status monitoring system.

[0013] As a further solution of the present invention: the described multi-modal collaborative corrosion status monitoring method and system, the described materials are commonly used materials for equipment in chlor-alkali production, such as 304 and 316L stainless steel and carbon steel.

[0014] As a further solution of the present invention: the signal acquisition in the step S1 uses 304 and 316L stainless steel and carbon steel, which are commonly used materials in chlor-alkali production equipment, and uses a large amount of rich data of various different materials to complete the training, which helps the model have better generalization ability.

[0015] As a further solution of the present invention: the data set construction in step S2 includes the following steps:

[0016] S21: Construct an acoustic-electric alignment pre-training dataset: There is no need for manual corrosion stage judgment, which greatly saves the cost of manual analysis. First, the collected electrochemical noise signals and acoustic emission signals are aligned in time steps and divided into smaller segments suitable for the time length to form a large number of acoustic-electric alignment data pairs, which are used as acoustic-electric alignment pre-training datasets.

[0017] S22: Construct a labeled dataset for the acoustic emission corrosion stage: For the acoustic emission data, a small number of acoustic emission signals with obvious corrosion characteristics are manually annotated to construct a labeled dataset for the acoustic emission corrosion stage. In addition, in addition to the above datasets, additional acoustic emission datasets are collected to form a test dataset for the subsequent evaluation of the model effect.

[0018] As a further solution of the present invention: the construction of the model in step S3 includes the following steps:

[0019] S31: Build a Transformer-based feature extractor: Utilize the self-attention mechanism to enable the model to focus on the timing information in the electrochemical noise and acoustic emission time series.

[0020] S32: Construct the first stage model learning strategy based on the self-supervised contrastive learning architecture: train the feature extractor proposed in S31, use the acoustic and electrical data in the pre-training data set for contrastive learning pre-training, so as to train the model to achieve collaborative learning of two modal signals, acoustic emission signals and electrochemical noise signals, and store the effective information of electrochemical noise signals. Its loss function is distribution alignment loss and event alignment loss.

[0021] S33: Construct a model learning strategy for monitoring the second stage corrosion stage: extract the acoustic emission signal feature extractor trained in step S32, and combine it with a classifier built based on a multi-layer perceptron, and perform further supervised training on it based on the labeled data set of the acoustic emission corrosion stage in step S2 to train the model to obtain the ability to effectively monitor the corrosion stage.

[0022] As a further solution of the present invention: the model test in step S4 takes the acoustic emission data of the test data set as input, and based on the model output results, it is determined what kind of corrosion state each sample in the test set belongs to, and the model detection accuracy is calculated for evaluation.

[0023] As a further solution of the present invention: the system in step S5 is built, the above-mentioned corrosion monitoring model is deployed to the server, and connected to the acoustic emission signal acquisition module. After the acoustic emission signal acquisition module collects the signal, the signal preprocessing module preprocesses the data and then transmits it to the corrosion status monitoring algorithm of the server to calculate and obtain the online corrosion status monitoring result.

[0024] Beneficial effects of the present invention:

[0025] 1. Compared with the supervised deep learning method that relies on a large amount of labeled data, the present invention adopts a self-supervised learning algorithm, uses a large amount of acoustic and electrical data pairs that do not require manual annotation, and performs acoustic and electrical collaborative pre-training on the model, which greatly reduces the dependence on labeled data;

[0026] 2. By designing a variety of alignment objective functions in the alignment learning stage of acoustic emission and electrochemical noise signals, the model is promoted to achieve effective alignment of the two modal signals of acoustic emission and electrochemical noise, making full use of the relevant information of the two signals and improving the corrosion monitoring capability of the model;

[0027] 3. The present invention can carry out on-site, non-destructive and efficient corrosion status monitoring of chlor-alkali production equipment. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] Figure 1This is a network structure diagram of a multi-modal collaborative corrosion status monitoring method and system of the present invention.

[0029] Figure 2 The present invention is a flowchart of a multi-modal collaborative corrosion status monitoring method and system. DETAILED DESCRIPTION

[0030] The following examples or drawings are used to illustrate the present invention, but are not used to limit the scope of the present invention.

[0031] The specific embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0032] Reference Figure 2 , a multi-modal collaborative corrosion state monitoring method and system, including the specific method including steps S1-S5:

[0033] S1: Signal acquisition: Collect signals of two modes: electrochemical noise and acoustic emission.

[0034] For example, the experimental temperature is set to room temperature; the solution is a corrosion solution such as NaCl solution; the sample material is a variety of stainless steel and carbon steel. The electrochemical noise collection method refers to ISO 17093-2015Corrosion of metals and alloys-Guidelines for corrosion test by electrochemical noise measurements, the experimental equipment is Gamry Reference600; the acoustic emission collection device uses the PCI-2 acoustic emission detection system of the American Physical Acoustics Company (PAC), with a gain of 40dB, and a WD-type acoustic emission broadband sensor is arranged above the specimen.

[0035] S2: Dataset construction: First, the collected electrochemical noise signals and acoustic emission signals are aligned in time steps and divided into smaller segments of suitable time length to form a large number of acoustic-electric alignment data pairs, which are used as acoustic-electric alignment pre-training data sets. For acoustic emission data, a small number of acoustic emission signals with obvious corrosion characteristics are manually annotated to construct a labeled data set for the acoustic emission corrosion stage. The sample corrosion state samples mainly include no corrosion, metastable corrosion stage and steady-state corrosion stage. In addition, in addition to the above data sets, additional acoustic emission data sets are collected to form a test data set for subsequent evaluation of model effects. To ensure the test effect, the training set and test set samples should not overlap.

[0036] Exemplarily, the electrochemical noise and acoustic emission signals collected in step S1 are first aligned in time steps and divided into small fragments of time series data with a length of 10 minutes to form a large number of acoustic and electrical data pairs. Among them, a small number of acoustic emission signals with obvious corrosion characteristics are manually annotated to construct a labeled data set for the acoustic emission corrosion stage. The sample corrosion state samples mainly include no corrosion, metastable corrosion stage and steady-state corrosion stage. In addition, in addition to the above data sets, an additional part of the acoustic emission data set is collected to form a test data set for the subsequent evaluation of the model effect.

[0037] S3: Model construction and training: reference Figure 1 , to build an improved network, firstly, a temporal self-attention feature extractor was built based on Transformer, and a two-stage learning strategy was designed. In the first stage, a model learning strategy based on a self-supervised contrastive learning architecture was built, and in the second stage, a model learning strategy with supervised training in the erosion stage was built.

[0038] The network is trained in two stages using the training dataset constructed in step S2.

[0039] S31: A temporal self-attention feature extractor is built based on Transformer, which performs attention calculation between temporal samples in the time dimension and extracts temporal information.

[0040] S32: Construct the first stage model learning strategy based on self-supervised contrastive learning architecture: as shown in the attached Figure 1 As shown, the self-supervised comparison structure has multiple parallel branches for two modal data, each branch includes the feature extractor described in S31, and is connected to a multi-layer perceptron structure to perform various collaborative alignment loss function calculations on the feature representation vectors output by each branch to optimize the model parameters. Sample event alignment calculation, with q evt express The extracted acoustic emission signal features, k evt express The extracted acoustic emission signal features, y represents h EN The extracted electrochemical noise signal feature is expressed as: evt =L nce (ζ(q evt ,y),{ζ(k evt ,y)}), where The two-modal distribution alignment calculation is as follows: in

[0041] S33: Constructing the model learning strategy for the second stage of supervised corrosion monitoring: Extracting the acoustic emission feature extractor trained in the first stage Then a classifier based on a multi-layer perceptron is added. The labeled dataset samples of the acoustic emission corrosion stage are input into the model, and the classification cross entropy loss is calculated to train the model to obtain the corrosion stage classification ability. The formula is:

[0042] S4: Corrosion monitoring capability evaluation: The corrosion stage of the test data set is classified through the trained model to obtain the evaluation results of the model detection performance.

[0043] Exemplarily, based on the acoustic emission data of the test set, the prediction results of the calculation model are The accuracy between the true label y is used to evaluate the model's predictive ability. The formula is: Among them, Accuracy represents accuracy, TP represents true positive examples, TN represents true negative examples, FP represents false positive examples, and FN represents false negative examples.

[0044] S5: System construction: Create an interactive interface, package the software, and install the software on the server; connect the on-site acoustic emission signal acquisition device with the packaged software to realize the on-site corrosion status monitoring system.

[0045] The system includes a signal acquisition device, a processor and a memory, wherein the memory stores computer program instructions, and when the processor executes the computer program instructions, a corrosion state monitoring method based on acoustic emission signals is implemented. The system obtains the on-site acoustic emission signal through the signal acquisition device, and transmits the signal to the processor through the communication interface. The processor preprocesses and extracts features of the signal, and then analyzes the corrosion state of the equipment. Other components of the system, such as communication protocols and data storage methods, are well known to those skilled in the art, so they will not be repeated here.

[0046] Compared with the supervised deep learning method that relies on a large amount of labeled data, the self-supervised learning algorithm is used to pre-train the model using a large amount of acoustic emission and electrochemical noise training data that does not require manual annotation, which significantly reduces the model's demand for labeled data. The Transformer-based feature extractor is combined with a multi-layer perceptron-based classifier to fully learn the effective time series information of time series data and effectively improve the performance of image segmentation tasks. At the same time, this method can complete the task of corrosion status monitoring using only acoustic emission data, which is particularly suitable for on-site non-destructive corrosion status monitoring of chlor-alkali production equipment.

[0047] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features thereof may be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A multi-modal collaborative corrosion state monitoring method and system, characterized by: The specific methods are as follows: S1: Signal acquisition: The sample materials include stainless steel and carbon steel, which are placed in a corrosive solution to cause corrosion. The corrosive solution used includes a NaCl solution. An electrochemical acquisition device and an acoustic emission acquisition device are set on the sample to collect electrochemical noise signals and acoustic emission signals at the same time; S2: Dataset construction: Using the collected electrochemical noise signals and acoustic emission signals, first align the two signals in time steps to construct a large-scale acoustic-electric alignment pre-training dataset that does not require manual annotation; take acoustic emission signals with obvious corrosion characteristics for manual annotation to construct a labeled dataset for the acoustic emission corrosion stage; S3: Model construction and training: Construct an improved two-stage network. The first stage is a contrastive learning structure based on Transformer, and the second stage is a corrosion stage monitoring network composed of a Transformer-based feature extractor and a multi-layer perceptron-based classifier. Use the two data sets constructed in step S2 to train the two-stage networks in turn. S4: Model test: The unknown acoustic emission test signal is judged by the improved corrosion stage monitoring model, and the corrosion stage corresponding to the test signal is output; S5: System construction: create an interactive interface, package the software, and install the software on the server; The on-site acoustic emission signal acquisition device is connected to the packaging software to realize the on-site corrosion status monitoring system.

2. A multi-modal coordinated corrosion state monitoring method and system according to claim 1, characterized in that: The signal acquisition in step S1 uses 304 and 316L stainless steel and carbon steel, which are commonly used materials in chlor-alkali production equipment, and uses a large amount of rich data of various materials to complete the training, helping the model to have better generalization ability.

3. A multi-modal coordinated corrosion state monitoring method and system according to claim 1, characterized in that: The data set construction in step S2 includes the following steps: S21: Constructing an acoustic-electric alignment pre-training dataset: No manual judgment of the corrosion stage is required, saving the cost of manual analysis. First, the collected electrochemical noise signals and acoustic emission signals are aligned in time steps and divided into smaller segments of suitable time length to form a large number of acoustic-electric alignment data pairs, which are used as acoustic-electric alignment pre-training datasets. S22: Construct a labeled dataset for the acoustic emission corrosion stage: For the acoustic emission data, the acoustic emission signals with obvious corrosion characteristics are manually annotated to construct a labeled dataset for the acoustic emission corrosion stage; in addition, in addition to the above datasets, some additional acoustic emission datasets are collected to form a test dataset for the subsequent evaluation of the model effect.

4. A multi-modal coordinated corrosion state monitoring method and system according to claim 1, characterized in that: The construction of the model in step S3 includes the following steps: S31: Build a Transformer-based feature extractor: Using the self-attention mechanism, the model can focus on the timing information in the electrochemical noise and acoustic emission time series; S32: Construct the first stage model learning strategy based on the self-supervised contrastive learning architecture: train the feature extractor proposed in S31, use the acoustic and electrical data in the pre-training data set for contrastive learning pre-training, so as to train the model to achieve the collaborative learning of the two modal signals of acoustic emission signal and electrochemical noise signal, and store the effective information of electrochemical noise signal. Its loss function is distribution alignment loss and event alignment loss. S33: Construct a model learning strategy for monitoring the second stage corrosion stage: extract the acoustic emission signal feature extractor trained in step S32, and combine it with a classifier built based on a multi-layer perceptron, and perform further supervised training on it based on the labeled data set of the acoustic emission corrosion stage in step S2 to train the model to obtain the ability to effectively monitor the corrosion stage.

5. A multi-modal coordinated corrosion state monitoring method and system according to claim 1, characterized in that: The model test in step S4 takes the acoustic emission data of the test data set as input, obtains the corrosion state of each sample in the test set based on the model output result, and calculates the model detection accuracy for evaluation.

6. A multi-modal coordinated corrosion state monitoring method and system according to claim 1, characterized in that: The system in step S5 is built, and the corrosion monitoring model trained in step S3 is deployed to the server, and connected to the acoustic emission signal acquisition module. After the acoustic emission signal acquisition module collects the signal, the signal preprocessing module preprocesses the data and transmits it to the corrosion state monitoring algorithm of the server to calculate and obtain the online corrosion state monitoring result.

7. A multi-modal coordinated corrosion state monitoring method and system according to claim 1, characterized in that: The sample material is not limited to a variety of materials, and the corrosion solution is not limited to a variety of solutions.

Images