A data-driven corrosion monitoring and safety early warning system for offshore equipment

Abstract

The present invention discloses a data-driven offshore equipment corrosion monitoring and safety early warning system, including a hardware unit, a data storage unit and a software unit; the hardware unit includes a corrosion sensor and a vibration sensor, which are used to collect sensor signals related to corrosion of offshore equipment components; the data storage unit is used to store data, preprocess data and manage operation and maintenance safety logs for the collected sensor signals; the software unit is used to train and predict the input preprocessed data using a neural network model algorithm, output the predicted value of the corrosion trend of the monitored components and the classification result of the corrosion degree, and determine whether to issue a safety alarm. The present invention adopts multi-sensor fusion technology, compiles sensor signals into corrosion loss amounts, and based on convolutional neural networks and attention mechanisms, enables the corrosion conditions of important components of offshore equipment to be digitally displayed and the corrosion database to be accumulated, thereby improving the accuracy of corrosion safety monitoring and early warning.

Description

Technical Field

[0001] The present invention relates to the technical field of material corrosion failure monitoring, and in particular to a data-driven offshore equipment corrosion monitoring and safety early warning system. Background Art

[0002] Marine resource development, maritime transportation, marine infrastructure, oil drilling platforms, offshore wind energy, cross-sea bridges and other equipment and facilities are in an extremely corrosive marine atmosphere all year round. The equipment and facilities will suffer from severe corrosion failure, which will bring huge maintenance costs and service safety hazards. Therefore, accurate fault diagnosis and early safety inspection systems for offshore equipment can be used to arrange maintenance, extend service life and ensure service safety.

[0003] At present, the most common methods for implementing corrosion status monitoring are manual inspection or installing a single corrosion sensor on equipment and facilities. The former requires professional technicians to operate, which is time-consuming and laborious, but still cannot avoid misjudgment; the latter determines the corrosion status by judging whether a single measurement value or a simple calculation value is within a specified range, but in actual service, a single sensor often sends a large number of false alarms because it cannot efficiently and accurately identify complex and changing environments, resulting in false alarms that waste manpower or missed reports that cause damage. This is because the corrosion safety status of equipment and facilities cannot be reliably judged based on the analysis of a single measurement. The instability, weakness, and randomness of a single sensor fault signal make it easy for the fault signal in the early stage of a corrosion accident to be submerged in noise, which also loses the meaning of monitoring.

[0004] Corrosion diagnosis is not a simple one-to-one solution process, and modern multidisciplinary fusion technology should be used for comprehensive judgment. The occurrence of corrosion is strongly correlated with the temperature, humidity and chloride ions in the marine atmospheric environment, and the chloride ion concentration is strongly correlated with wind speed and direction. At the same time, the impact of corrosion on the monitored components will lead to a certain degree of mass loss and adhesion of corrosion products. The degree of corrosion and the vibration acceleration and amplitude of the components will have a potential relationship of mutual promotion. Therefore, there is an urgent need for a state monitoring and safety early warning system based on the fusion of multiple sensor devices and multiple algorithms to realize continuous monitoring of the corrosion safety status of offshore equipment, fault diagnosis, and prediction of future corrosion trends. Summary of the invention

[0005] The purpose of the present invention is to provide a data-driven offshore equipment corrosion monitoring and safety warning system to solve the technical problems that the existing offshore equipment fault monitoring technology leads to misjudgment and false alarm due to failure to consider material corrosion loss, and traditional corrosion monitoring methods are difficult to perform corrosion monitoring on special positions such as bearings, fasteners, bolts, etc. of offshore equipment.

[0006] To solve the above technical problems, the embodiments of the present invention provide the following solutions:

[0007] A data-driven offshore equipment corrosion monitoring and safety early warning system comprises a hardware unit, a data storage unit and a software unit, wherein the hardware unit is connected to the data storage unit, and the data storage unit is connected to the software unit;

[0008] The hardware unit includes a corrosion sensor and a vibration sensor for collecting sensor signals related to corrosion of offshore equipment components;

[0009] The data storage unit is used to store data, pre-process data and manage operation and maintenance safety logs of collected sensor signals;

[0010] The software unit is used to train and predict the input pre-processed data using a neural network model algorithm, output the corrosion prediction value of the monitored component and the classification result of the corrosion degree, and determine whether to issue a safety alarm.

[0011] Preferably, the corrosion sensor includes a built-in corrosion dual-electrode detector, an environmental temperature and humidity detector, a wind speed and direction detector, and a first data collector; each detector probe is fitted outside the environment where the monitored component is located, and is used to collect corrosion current signals and environmental parameters of the same material as the monitored component in the environment;

[0012] The vibration sensor includes a plurality of eddy current displacement sensor elements and a second data collector, and is used to collect vibration parameters of the monitored offshore equipment during operation, wherein the vibration parameters include vibration acceleration and amplitude.

[0013] Preferably, the first data collector comprises a first microcontroller, a first data storage card and a first receiving and transmitting device, the first data storage card and the first receiving and transmitting device are both connected to the first microcontroller, and the first data collector is used to transmit the data collected by the corrosion sensor to the data storage unit for preprocessing;

[0014] The second data collector includes a second microcontroller, a second data storage card and a second receiving and transmitting device. The second data storage card and the second receiving and transmitting device are both connected to the second microcontroller. The second data collector is used to transmit the data collected by the vibration sensor to the data storage unit for preprocessing.

[0015] Preferably, the data storage unit includes a storage module, a log management module and a preprocessing module; the storage module is used to store the corrosion-related sensor signals collected by the hardware unit, the log management module is used to perform log management on the corrosion safety operation and failure maintenance conditions of the components, and the preprocessing module is used to perform corrosion parameter quantization conversion, outlier elimination, and noise reduction preprocessing on the collected sensor signals, and complete data time alignment and space alignment, heterogeneous sensor signal feature extraction and fusion, so that each sensor signal can be used for diagnosis and identification of corrosion status information.

[0016] Preferably, the corrosion parameter quantitative conversion is expressed according to the corrosion current signal collected by the corrosion sensor, the environmental response coefficient and the stoichiometric coefficient of the current corrosion reaction:

[0017]

[0018] In the formula, v corr is the corrosion parameter, F is the Faraday constant, unit: C / mol; M j is the molar mass, unit: kg / mol; ρ j is the density of the species, unit: kg / m; i is the measured value of the corrosion sensor current signal, unit: A / cm 2 ; P is the environmental response coefficient; T is the stoichiometric coefficient of the corrosion reaction; n is the number of electrons participating in the reaction; j is the material coefficient; m is the electrode reaction coefficient;

[0019] The corrosion parameter quantization conversion is used to quantify the corrosion dissolution rate caused by the corrosion sensor electrode reaction before model training, so that the signal collected by the corrosion sensor becomes a physical quantity that directly reflects the degree of corrosion;

[0020] The environmental response coefficient is a characteristic coefficient established by collecting environmental parameters and corrosion parameters, which can quantitatively reflect the corrosion acceleration effect of the current environmental conditions on the monitored components. The mathematical expression is:

[0021]

[0022] Where T and RH are the actual temperature and humidity values ​​of the ambient temperature and humidity detectors, respectively. is the estimated value of chloride ions in the environment, which is positively correlated with wind speed and can be a constant; A, B, and C are the sensitivity coefficients of the monitored component materials to various environmental variables, which are used to quantify the importance of the environmental parameters on the corrosion of the monitored component materials, and their sum is 1; a, b, c, d, e, and f are the parameters in the corrosion acceleration formula for the monitored component materials by different environmental factors;

[0023] The environmental response coefficient is used to improve the accuracy of quantitative conversion of corrosion parameters.

[0024] Preferably, the software unit includes an algorithm module;

[0025] The algorithm module includes a GRU neural network model and a CTSAN neural network model; the GRU neural network model is used to predict the corrosion trend of the monitored components based on the GRU neural network and the parameter attention mechanism; the CTSAN neural network model is used to output the classification results of the transient corrosion degree of the monitored components.

[0026] Preferably, the software unit further comprises an output module, and the output module is connected to the algorithm module;

[0027] The output module includes a controller, a display and an alarm; the controller is used to set the upper limit of the residual between the actual value and the predicted value of the target modeling parameter of the monitored component under the normal corrosion state, and set the threshold of the over-limit time to identify the abnormal corrosion state; the display is used to display the instantaneous corrosion state and corrosion degree classification results of the monitored component; the alarm is used to send a fault alarm signal when the controller identifies that the continuous over-limit time exceeds the threshold.

[0028] Preferably, the GRU neural network model includes a parameter attention layer, a GRU layer and a fully connected layer; its working mode is: training the normal operating training data set and the pre-processed test data set input by the data storage unit; in the training stage, the normal operating training data set is input to train the normal corrosion behavior model of the monitored component; in the process of component corrosion status monitoring, the pre-processed test data set is input into the trained normal corrosion behavior model to obtain the predicted value of normal corrosion safety operation data.

[0029] Preferably, in the GRU neural network model, the input parameter vector x of a sequence single time step is (t) =[x1 (t) ,x2 (t) ...,x m (t) ] as the parameter attention mechanism input, according to g (t) =σ(Wx (t) +b) quantizes the weight of the current parameter, σ is the sigmoid activation function; W is the trainable coefficient matrix; b is the bias matrix; and the softmax function is used to normalize each attention weight so that it satisfies the probability distribution that the sum of the weights is 1. The normalized attention weight is multiplied by its corresponding parameter value to enhance or weaken the i-th input parameter, thereby obtaining the weighted input vector x' (t) .

[0030] Preferably, the GRU neural network model includes an update gate and a reset gate, wherein the update gate controls how much information of the hidden state at the previous moment is transferred to the current moment; the reset gate determines whether to ignore the previous hidden state; each GRU unit internally outputs the hidden state y based on the time t-1 (t-1) and the input time series x at time t (t) To calculate the hidden state y output at time t (t) , hidden state y (t) =f(y (t-1) ,x (t) ) represents the hidden timing information extracted by each unit, then the update gate at time t is expressed as: g t =σ(W g x (t) +U g h t-1 +b g ), the reset gate is represented by f t =σ(W f x (t) +U f h t-1 +b f ), where W and U are the coefficient transposed matrices.

[0031] Preferably, the CTSAN neural network model includes a convolutional feature extraction layer, a spatial attention layer, a temporal attention layer and a classification layer; its working mode is: the multivariate time series X=[x1, x2..., x t ] Based on the convolutional feature extraction layer, the features of the time interval of interest are extracted, and the model focus is converted from a large number of time points to a small number of time interval sequences C = [C1, C2..., C t ]; then the time interval feature is sent to the spatial attention layer, and the global spatial feature C'=[C'1,C'2...,C' t ]; then the global spatial features of all intervals are sent to the temporal attention layer, and the global spatiotemporal features C" considering the weights of each interval are output; the global spatiotemporal features are then sent to the classification layer, and the predicted category corresponding to the sequence is output Finally, given a set of training examples, the coefficients of each layer in the model are iteratively updated to finally obtain the CTSAN corrosion fault classification model.

[0032] The beneficial effects brought about by the technical solution provided by the embodiment of the present invention include at least:

[0033] The data-driven offshore equipment corrosion status monitoring and safety early warning system provided by the embodiment of the present invention can provide timely early warning of corrosion damage to important components of offshore equipment and make more accurate judgments on future corrosion development trends. Compared with the traditional method of first causing a corrosion accident and then detecting a failure signal, the present invention realizes instantaneous monitoring of corrosion and judgment of development trends; compared with the disadvantage that a large number of false alarms will be sent in the traditional hard-coded alarm upper and lower limits, the present invention will trigger a corrosion safety alarm only after reaching a preset threshold for a certain period of time in algorithm control; compared with the traditional single sensor to obtain corrosion information, the present invention integrates multi-sensor signal characteristics, enriches the information source, broadens the channel, and predicts the occurrence of corrosion accidents on the basis of high automation and intelligence, which can assist maintenance personnel in replacing components in a timely manner and improve the level of offshore equipment corrosion safety monitoring. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0035] Figure 1 is a structural schematic diagram of a data-driven offshore equipment corrosion monitoring and safety early warning system provided by an embodiment of the present invention;

[0036] Figure 2 It is a diagram of the actual working process of the system provided by the embodiment of the present invention;

[0037] Figure 3 It is a schematic diagram of the structure of a corrosion dual-electrode sensor probe provided by an embodiment of the present invention;

[0038] Figure 4 is a schematic diagram of measured values ​​and predicted values ​​in a monitoring process provided by an embodiment of the present invention;

[0039] Figure 5 It is a schematic diagram of the corrosion state display interface provided by an embodiment of the present invention.

[0040] As shown in the figure, in order to clearly implement the structure of the embodiment of the present invention, specific structures and devices are marked in the figure, but this is only for illustrative purposes and is not intended to limit the present invention to the specific structure, device and environment. According to specific needs, ordinary technicians in this field can adjust or modify these devices and environments, and the adjustments or modifications made are still included in the scope of protection of the present invention. DETAILED DESCRIPTION

[0041] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0042] The embodiment of the present invention provides a data-driven offshore equipment corrosion monitoring and safety early warning system, such as Figure 1 As shown, the system includes a hardware unit 1, a data storage unit 2 and a software unit 3, the hardware unit 1 is connected to the data storage unit 2, and the data storage unit 2 is connected to the software unit 3;

[0043] The hardware unit 1 includes a corrosion sensor 11 and a vibration sensor 12, which are used to collect sensor signals related to corrosion of offshore equipment components;

[0044] The data storage unit 2 is used for data storage, data preprocessing and operation and maintenance safety log management of the collected sensor signals;

[0045] The software unit 3 is used to train and predict the input pre-processed data using a neural network model algorithm, output the corrosion prediction value of the monitored component and the classification result of the corrosion degree, and determine whether to issue a safety alarm.

[0046] Furthermore, the corrosion sensor 11 includes a built-in corrosion dual-electrode detector, an environmental temperature and humidity detector, a wind speed and direction detector, and a first data collector; each detector probe is installed on the outside of the environment where the monitored component is located, and is used to collect corrosion current signals and environmental parameters of the same material as the monitored component in the environment.

[0047] The vibration sensor 12 includes a plurality of eddy current displacement sensor elements and a second data collector, and is used to collect vibration parameters during the operation of the monitored offshore equipment, wherein the vibration parameters include vibration acceleration and amplitude.

[0048] Wherein, the first data collector includes a first microcontroller, a first data storage card and a first receiving and transmitting device, the first data storage card and the first receiving and transmitting device are both connected to the first microcontroller, and the first data collector is used to transmit the data collected by the corrosion sensor 11 to the data storage unit 2 for preprocessing;

[0049] The second data collector includes a second microcontroller, a second data storage card and a second receiving and transmitting device. The second data storage card and the second receiving and transmitting device are both connected to the second microcontroller. The second data collector is used to transmit the data collected by the vibration sensor 12 to the data storage unit 2 for preprocessing.

[0050] In the embodiment of the present invention, the corrosion current signals of different intensities measured by the corrosion dual-electrode detector can directly reflect the intensity of corrosion. However, due to the polarization of the probe and environmental factors, the noise problem of the monitoring signal is caused. Therefore, an environmental temperature and humidity detector and a wind speed and direction detector are inserted inside the corrosion sensor 11 to improve the detection accuracy.

[0051] Preferably, the corrosion dual-electrode detector includes: a uniform dual-electrode corrosion detector, a galvanic dual-electrode corrosion detector, a resistive dual-electrode corrosion detector, a galvanic dual-electrode potential corrosion detector, etc.

[0052] Furthermore, the data storage unit 2 includes a storage module 21, a log management module 22 and a preprocessing module 23; wherein the storage module 21 is used to store the corrosion-related sensor signals collected by the hardware unit 1, the log management module 22 is used to perform log management on the corrosion safety operation and failure maintenance conditions of the components, and the preprocessing module 23 is used to perform corrosion parameter quantization conversion, outlier elimination, and noise reduction preprocessing on the collected sensor signals, and complete data time alignment and space alignment, heterogeneous sensor signal feature extraction and fusion, so that each sensor signal can be used for diagnosis and identification of corrosion status information.

[0053] In the embodiment of the present invention, the corrosion parameter quantitative conversion is expressed according to the corrosion current signal collected by the corrosion sensor, the environmental response coefficient and the stoichiometric coefficient of the current corrosion reaction:

[0054]

[0055] In the formula, v corr is the corrosion parameter, F is the Faraday constant, unit: C / mol; M j is the molar mass, unit: kg / mol; ρ j is the density of the species, unit: kg / m; i is the measured value of the corrosion sensor current signal, unit: A / cm 2 ; P is the environmental response coefficient; T is the stoichiometric coefficient of the corrosion reaction; n is the number of electrons participating in the reaction; j is the material coefficient; m is the electrode reaction coefficient.

[0056] The corrosion parameter quantization conversion is used to quantify the corrosion dissolution rate caused by the corrosion sensor electrode reaction before model training, so that the signal collected by the corrosion sensor becomes a physical quantity that directly reflects the degree of corrosion.

[0057] The environmental response coefficient is a characteristic coefficient established by collecting environmental parameters and corrosion parameters, which can quantitatively reflect the corrosion acceleration effect of the current environmental conditions on the monitored components. The mathematical expression is:

[0058]

[0059] Where T and RH are the actual temperature and humidity values ​​of the ambient temperature and humidity detectors, respectively. is the estimated value of chloride ions in the environment, which is positively correlated with the wind speed and can be a constant; A, B, and C are the sensitivity coefficients of the monitored component materials to various environmental variables, which are used to quantify the importance of the environmental parameters on the corrosion of the monitored component materials, and their sum is 1. They are calculated by using statistical methods, regression algorithms, random forests, and other methods based on the initial corrosion current data monitored by the corrosion sensor; a, b, c, d, e, and f are the parameters in the corrosion acceleration formula for the monitored component materials by different environmental factors, which are calculated by using the nonlinear regression method based on the initial corrosion current data and environmental parameter data monitored by the corrosion sensor; the environmental response coefficient is used to improve the accuracy of the quantitative conversion of corrosion parameters.

[0060] In the embodiment of the present invention, the data denoising method includes mean wavelet packet denoising, fast Fourier transform denoising, etc.; the extraction of eigenvalues ​​includes statistical means such as mean, autocorrelation, filter, power spectral density and variance, etc.

[0061] Furthermore, the software unit 3 includes an algorithm module 31; the algorithm module 31 includes two neural network models for training historical data and applying the trained model to real-time data to predict and classify corrosion signals.

[0062] Specifically, the algorithm module 31 includes a GRU neural network model 311 and a CTSAN neural network model 312 (convolutional spatiotemporal attention neural network model); the GRU neural network model 311 is used to predict the corrosion trend of the monitored components based on the GRU neural network and parameter attention mechanism; the CTSAN neural network model 312 is used to output the classification results of the transient corrosion degree of the monitored components.

[0063] In an embodiment of the present invention, the GRU neural network model 311 includes a parameter attention layer, a GRU layer and a fully connected layer; its working mode is: training the normal operating training data set and the pre-processed test data set input by the data storage unit 2; in the training stage, the normal operating training data set is input to train the normal corrosion behavior model of the monitored component; in the process of component corrosion status monitoring, the pre-processed test data set is input into the trained normal corrosion behavior model to obtain the predicted value of normal corrosion safety operation data.

[0064] In the GRU neural network model 311, the input parameter vector x of a sequence single time step is (t) =[x1 (t) ,x2 (t)...,x m (t) ] as the parameter attention mechanism input, according to g (t) =σ(Wx (t) +b) quantizes the weight of the current parameter, σ is the sigmoid activation function; W is the trainable coefficient matrix; b is the bias matrix; and the softmax function is used to normalize each attention weight so that it satisfies the probability distribution that the sum of the weights is 1. The normalized attention weight is multiplied by its corresponding parameter value to enhance or weaken the i-th input parameter, thereby obtaining the weighted input vector x' (t) .

[0065] The GRU neural network model 311 includes an update gate and a reset gate, wherein the update gate controls how much information of the hidden state at the previous moment is transferred to the current moment; the reset gate determines whether to ignore the previous hidden state; each GRU unit has a hidden state y output at time t-1. (t-1) and the input time series x at time t (t) To calculate the hidden state y output at time t (t) , hidden state y (t) =f(y (t-1) ,x (t) ) represents the hidden timing information extracted by each unit, then the update gate at time t is expressed as: g t =σ(W g x (t) +U g h t-1 +b g ), the reset gate is represented by f t =σ(W f x (t) +U f h t-1 +b f ), where W and U are the coefficient transposed matrices.

[0066] In the embodiment of the present invention, the CTSAN neural network model 312 includes a convolutional feature extraction layer, a spatial attention layer, a temporal attention layer and a classification layer; its working mode is: the multivariate time series X=[x1, x2..., x t ] Based on the convolutional feature extraction layer, the features of the time interval of interest are extracted, and the model focus is converted from a large number of time points to a small number of time interval sequences C = [C1, C2..., C t ]; then the time interval feature is sent to the spatial attention layer, based on Output the global spatial features of all intervals C' = [C'1, C'2..., C' t]; then the global spatial features of all intervals are sent to the temporal attention layer, and the global spatiotemporal features C" considering the weights of each interval are output; the global spatiotemporal features are then sent to the classification layer, and the predicted category corresponding to the sequence is output Finally, given a set of training examples, the coefficients of each layer in the model are iteratively updated to finally obtain the CTSAN corrosion fault classification model.

[0067] Furthermore, the software unit 3 also includes an output module 32, which is connected to the algorithm module 31; the output module 32 includes a controller 321, a display 322 and an alarm 323; the controller 321 is used to set the upper limit of the residual between the actual value and the predicted value of the target modeling parameter under the normal corrosion state of the monitored component, and set the threshold of the over-limit time to identify the abnormal corrosion state; the display 322 is used to display the instantaneous corrosion state and corrosion degree classification results of the monitored component; the alarm 323 is used to send a fault alarm signal when the controller identifies that the continuous over-limit time exceeds the threshold.

[0068] Preferably, 60 minutes can be set as the threshold of the over-limit duration, that is, when the over-limit duration exceeds 60 minutes continuously, the alarm 323 will send out a fault alarm signal.

[0069] In the data-driven offshore equipment corrosion status monitoring and safety warning system described in the embodiment of the present invention, in the working state, the signal collected by the sensor of the hardware unit 1 is sent to the data storage unit 2 through the internal receiving and transmitting device; the data storage unit 2 sends the pre-processed sensor signal to the software unit 3 for model training and outputs the corrosion trend prediction result and the corrosion degree classification result. Because this embodiment adopts multi-sensor fusion technology and is based on convolutional neural network and attention mechanism, the corrosion condition of important parts of offshore equipment is digitally displayed, which improves the accuracy of corrosion safety warning while realizing continuous corrosion monitoring, and presents the spatiotemporal characteristics extracted by the model in an understandable form, further improving the reliability of the diagnosis result.

[0070] Figure 2 It is the actual working flow diagram of the system of the present invention, which specifically includes:

[0071] The characteristic parameters of the known corrosion loss state are set as the target modeling parameters, and other operating parameters and environmental parameters constitute the input sequence at the corresponding time;

[0072] The parameter attention mechanism is introduced to focus on the focal area of ​​corrosion state prediction, reduce or even ignore the attention to other areas, dig out the correlation between the input parameters of the data set and the target modeling parameters, assign attention weights to each input parameter, and avoid the subjective threshold limitation of traditional parameter selection methods;

[0073] According to the attention weight, the key parameter expression is enhanced and the redundant parameters are weakened, so as to avoid the loss of relevant information caused by the existing parameter selection method due to the selection of part of the input parameters by the preset threshold;

[0074] The normalized attention weights and their corresponding parameter values ​​are enhanced or weakened to obtain a weighted input vector, which is then input into the GRU neural network model for model training to extract hidden temporal information.

[0075] Finally, the hidden time series information output by the terminal GRU unit is sent to the fully connected layer to output the predicted value of the corrosion loss of the target modeling parameter corresponding to the sequence.

[0076] At the same time, the data recorded in the corrosion safety operation and maintenance log is transferred from the time point to the time interval in the convolutional feature extraction layer of the convolutional spatiotemporal attention network (CTSAN);

[0077] Next, the weights of each spatial parameter in each time interval are determined through the spatial attention layer, and then the weights of each time interval are determined through the temporal attention layer. Finally, the features that comprehensively consider the weights of the above two parts are input into the classification layer to output the predicted corrosion category.

[0078] In a more specific embodiment, the same steel type as the gearbox material of the No. 1 wind turbine of the Donghai Bridge Wind Farm is selected, and 14 rectangular steel sheets with a length of 20 mm, a width of 10 mm, and a thickness of 1 mm are cut, and a number of polytetrafluoroethylene insulating gaskets with a length of 20 mm, a width of 10 mm, and a thickness of 0.1 mm are prepared. A polytetrafluoroethylene insulating gasket is sandwiched between each two adjacent steel sheets. The odd-numbered steel sheets are connected by wires as anodes, and the even-numbered steel sheets are connected by wires as cathodes. After all the test pieces are compacted without leaving any gaps between the adjacent pieces, they are encapsulated with epoxy resin, leaving only one test surface exposed. After air drying, the cathode wire is connected to the positive electrode of the data collector, and the anode wire is connected to the negative electrode of the data collector to make a corrosion dual-electrode sensor probe structure as shown in FIG. Figure 3 shown.

[0079] The corrosion dual-electrode detector probe and the environmental temperature and humidity detector probe are fixed on the surface of the gearbox, and the eddy current displacement sensor element is fitted inside the gearbox. After the installation is completed, the sensor signal is collected and the corrosion operation data of the gearbox of the No. 1 wind turbine of the Donghai Bridge Wind Farm is obtained. The obtained instantaneous sensor data and corrosion safety operation log are sent to the data storage unit. The corrosion parameter quantification conversion of the sensor signal is performed in the data storage unit through the collected sensor data (as shown in Table 1). At this time, only the iron dissolution reaction in the material is considered, that is, At this time, the stoichiometric coefficient is 1; the number of participating electrons is 2; the density of the monitored steel is 7850kg / m3; the relative molecular mass is 0.056kg / mol. The instantaneous loss value of the corrosion thickness of the monitored component material in the service environment is calculated by collecting the initial sensor signal data, where A is 0.4, B is 0.35, and C is 0.25; the values ​​of a, b, c, d, e, and f are 1.3e-11, 7.1, 21.75, 75.5, 0.26, and 0.73, respectively. The instantaneous corrosion loss value after the quantitative conversion of the corrosion parameters can be expressed as:

[0080]

[0081] After that, it is pre-processed for noise reduction and outliers are removed before being sent to the software unit for model training.

[0082] Table 1

[0083]

[0084] During the monitoring process, the newly input sensor signal participates in the calculation within the trained model unit and outputs the classification result and prediction value. Figure 4 The controller displays the measured and predicted values ​​of the instantaneous data after preprocessing and noise removal. If the preset upper limit of the residual error is reached and the over-limit time threshold is exceeded, a safety alarm is triggered, indicating that the gearbox of the No. 1 wind turbine of the Donghai Bridge Wind Farm is in an abnormal corrosion state, the chassis material is severely corroded, and the risk of corrosion fracture may occur at any time; if the preset upper limit of the residual error is not reached, it is a normal corrosion behavior state. At this time, the instantaneous corrosion state and corrosion degree classification results can be viewed on the display, such as Figure 5 As shown, the newly output non-abnormal instantaneous corrosion data is stored in the data storage unit in the form of an operation and maintenance safety log, thereby achieving the effect of improving the offshore wind turbine corrosion database and optimizing the neural network model.

[0085] In summary, the data-driven offshore equipment corrosion status monitoring and safety early warning system described in this embodiment can continuously monitor the corrosion safety service status of important components of offshore equipment, compile sensor signals into intuitive corrosion information by constructing corrosion parameter quantification conversion equations and environmental response coefficients, and integrate multiple algorithm models to classify and evaluate the transient corrosion degree of the monitored components based on big data technology, and predict future corrosion trends. While improving the accuracy of corrosion monitoring, it changes the "occur first, then detect" phenomenon of traditional corrosion failure monitoring, and provides early warning for important components of offshore equipment before corrosion failure occurs.

[0086] In addition, the data-driven offshore equipment corrosion status monitoring and safety warning system described in this embodiment adopts multi-sensor fusion technology, which eliminates the disadvantages of frequent false alarms and missed alarms of a single sensor, enriches the information source, broadens the channel, and realizes the prediction of corrosion accidents on the basis of high automation and intelligence, and assists maintenance personnel to replace equipment components in time. At the same time, it can accumulate a large amount of corrosion and corrosion-related data, enrich the corrosion database for the safe service of offshore equipment, realize a virtuous cycle of offshore equipment corrosion safety monitoring, and provide theoretical support for the selection of materials for offshore equipment.

[0087] It should be noted that the references to "one embodiment", "embodiment", "exemplary embodiment", "some embodiments" and the like in the specification indicate that the embodiments described may include specific features, structures or characteristics, but not every embodiment may include the specific features, structures or characteristics. In addition, when a specific feature, structure or characteristic is described in conjunction with an embodiment, it should be within the knowledge of a person skilled in the art to implement such feature, structure or characteristic in conjunction with other embodiments (whether or not explicitly described).

[0088] In general, a term can be understood, at least in part, from its use in context. For example, depending, at least in part, on the context, the term "one or more" as used herein can be used to describe any feature, structure, or characteristic in the singular sense, or can be used to describe a combination of features, structures, or characteristics in the plural sense. Additionally, the term "based on" can be understood as not necessarily intended to convey an exclusive set of factors, but can instead, depending, at least in part, on the context, allow for the presence of other factors that are not necessarily explicitly described.

[0089] As used herein, the term "nominal / nominal" refers to an expected or target value for a characteristic or parameter of a component or process operation set during the design phase of a production or manufacturing process, as well as a range of values ​​above and / or below the expected value. The range of values ​​may be due to slight variations in the manufacturing process or tolerances. As used herein, the term "approximately" indicates a value of a given quantity that may vary based on a particular technology node associated with the subject semiconductor device. Based on a particular technology node, the term "approximately" may indicate a value of a given quantity that varies, for example, within 5%-15% of the value (e.g., ±5%, ±10%, or ±15% of the value).

[0090] It will be understood that the meaning of “on,” “over,” and “above” in this disclosure should be interpreted in the broadest manner, so that “on” means not only “directly on” something, but also includes the meaning of being “on” something with intervening features or layers therebetween, and “on” or “over” means not only “on” or “above” something, but also includes the meaning of being “on” or “above” something with no intervening features or layers therebetween.

[0091] Additionally, spatially relative terms such as "under," "beneath," "lower," "above," "upper," and the like may be used herein for descriptive convenience to describe the relationship of one element or feature to another element or features, as shown in the accompanying drawings. Spatially relative terms are intended to encompass different orientations of the device in use or operation in addition to the orientation depicted in the accompanying drawings. The device may be oriented in other ways, and the spatially relative descriptors used herein may be similarly interpreted accordingly.

[0092] The present invention covers any substitution, modification, equivalent method and scheme made on the essence and scope of the present invention. In order to make the public have a thorough understanding of the present invention, specific details are described in detail in the following preferred embodiments of the present invention, but those skilled in the art can fully understand the present invention without the description of these details. In addition, in order to avoid unnecessary confusion about the essence of the present invention, well-known methods, processes, procedures, components and circuits are not described in detail.

[0093] A person skilled in the art will appreciate that all or part of the steps in the above-mentioned embodiment method can be completed by instructing related hardware through a program, and the program can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc.

[0094] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A data-driven offshore equipment corrosion monitoring and safety early warning system, characterized in that: It comprises a hardware unit, a data storage unit and a software unit, wherein the hardware unit is connected to the data storage unit, and the data storage unit is connected to the software unit; The hardware unit includes a corrosion sensor and a vibration sensor for collecting sensor signals related to corrosion of offshore equipment components; The data storage unit is used to store data, pre-process data and manage operation and maintenance safety logs of collected sensor signals; The data storage unit includes a storage module, a log management module and a preprocessing module; the storage module is used to store the corrosion-related sensor signals collected by the hardware unit, the log management module is used to log the corrosion safety operation and failure maintenance of the components, and the preprocessing module is used to perform corrosion parameter quantization conversion, outlier removal, and noise reduction preprocessing on the collected sensor signals, and complete data time registration and space registration, heterogeneous sensor signal feature extraction and fusion, so that each sensor signal can be used for diagnosis and identification of corrosion status information; The corrosion parameter quantitative conversion is expressed according to the corrosion current signal collected by the corrosion sensor, the environmental response coefficient and the stoichiometric coefficient of the current corrosion reaction: In the formula, v corr is the corrosion parameter, F is the Faraday constant, unit: C / mol; M j is the molar mass, unit: kg / mol; ρ j is the density of the species, unit: kg / m; i is the measured value of the corrosion sensor current signal, unit: A / cm 2 ; P is the environmental response coefficient; T is the stoichiometric coefficient of the corrosion reaction; n is the number of electrons participating in the reaction; j is the material coefficient; m is the electrode reaction coefficient; The corrosion parameter quantization conversion is used to quantify the corrosion dissolution rate caused by the corrosion sensor electrode reaction before model training, so that the signal collected by the corrosion sensor becomes a physical quantity that directly reflects the degree of corrosion; The environmental response coefficient is a characteristic coefficient established by collecting environmental parameters and corrosion parameters, which can quantitatively reflect the corrosion acceleration effect of the current environmental conditions on the monitored components. The mathematical expression is: Where T and RH are the actual temperature and humidity values ​​of the ambient temperature and humidity detectors, respectively. is the estimated value of chloride ions in the environment, which is positively correlated with wind speed and can be a constant; A, B, and C are the sensitivity coefficients of the monitored component materials to various environmental variables, which are used to quantify the importance of the environmental parameters on the corrosion of the monitored component materials, and their sum is 1; a, b, c, d, e, and f are the parameters in the corrosion acceleration formula for the monitored component materials by different environmental factors; The environmental response coefficient is used to improve the accuracy of quantitative conversion of corrosion parameters; The software unit is used to train and predict the input pre-processed data using a neural network model algorithm, output the corrosion prediction value of the monitored component and the classification result of the corrosion degree, and determine whether to issue a safety alarm.

2. The data-driven offshore equipment corrosion monitoring and safety early warning system according to claim 1 is characterized in that: The corrosion sensor includes a built-in corrosion dual-electrode detector, an environmental temperature and humidity detector, a wind speed and direction detector, and a first data collector; each detector probe is installed outside the environment where the monitored component is located, and is used to collect corrosion current signals and environmental parameters of the same material as the monitored component in the environment; The vibration sensor includes a plurality of eddy current displacement sensor elements and a second data collector, and is used to collect vibration parameters during the operation of the monitored offshore equipment, wherein the vibration parameters include vibration acceleration and amplitude.

3. The data-driven offshore equipment corrosion monitoring and safety early warning system according to claim 2 is characterized in that: The first data collector includes a first microcontroller, a first data storage card and a first receiving and transmitting device, the first data storage card and the first receiving and transmitting device are both connected to the first microcontroller, and the first data collector is used to transmit the data collected by the corrosion sensor to the data storage unit for preprocessing; The second data collector includes a second microcontroller, a second data storage card and a second receiving and transmitting device. The second data storage card and the second receiving and transmitting device are both connected to the second microcontroller. The second data collector is used to transmit the data collected by the vibration sensor to the data storage unit for preprocessing.

4. The data-driven offshore equipment corrosion monitoring and safety early warning system according to claim 1 is characterized in that: The software unit includes an algorithm module; The algorithm module includes a GRU neural network model and a CTSAN neural network model; the GRU neural network model is used to predict the corrosion trend of the monitored components based on the GRU neural network and the parameter attention mechanism; the CTSAN neural network model is used to output the classification results of the transient corrosion degree of the monitored components; The software unit also includes an output module, and the output module is connected to the algorithm module; The output module includes a controller, a display and an alarm; the controller is used to set the upper limit of the residual between the actual value and the predicted value of the target modeling parameter of the monitored component under the normal corrosion state, and set the threshold of the over-limit time to identify the abnormal corrosion state; the display is used to display the instantaneous corrosion state and corrosion degree classification results of the monitored component; the alarm is used to send a fault alarm signal when the controller identifies that the continuous over-limit time exceeds the threshold.

5. The data-driven offshore equipment corrosion monitoring and safety early warning system according to claim 4 is characterized in that: The GRU neural network model includes a parameter attention layer, a GRU layer and a fully connected layer; its working mode is: training the normal operation training data set and the pre-processed test data set input by the data storage unit; In the training phase, the normal operation training data set is input to train the normal corrosion behavior model of the monitored components; In the process of component corrosion status monitoring, the preprocessed test data set is input into the trained normal corrosion behavior model to obtain the predicted value of normal corrosion safety operation data.

6. The data-driven offshore equipment corrosion monitoring and safety early warning system according to claim 5 is characterized in that: In the GRU neural network model, the input parameter vector x of a sequence single time step is (t) =[x1 (t) ,x2 (t) ...,x m (t) ] as the parameter attention mechanism input, according to g (t) =σ(Wx (t) +b) quantizes the weight of the current parameter, σ is the sigmoid activation function; W is the trainable coefficient matrix; b is the bias matrix; and the softmax function is used to normalize each attention weight so that it satisfies the probability distribution that the sum of the weights is 1. The normalized attention weight is multiplied by its corresponding parameter value to enhance or weaken the i-th input parameter, thereby obtaining the weighted input vector x' (t) .

7. The data-driven offshore equipment corrosion monitoring and safety early warning system according to claim 6 is characterized in that: The GRU neural network model includes an update gate and a reset gate, wherein the update gate controls how much information of the hidden state at the previous moment is transferred to the current moment; the reset gate determines whether to ignore the previous hidden state; each GRU unit has a hidden state y output at time t-1. (t-1) and the input time series x at time t (t) To calculate the hidden state y output at time t (t) , hidden state y (t) =f(y (t-1) ,x (t) ) represents the hidden timing information extracted by each unit, then the update gate at time t is expressed as: g t =σ(W g x (t) +U g h t-1 +b g ), the reset gate is represented by f t =σ(W f x (t) +U f h t-1 +b f ), where W and U are the coefficient transposed matrices.

8. The data-driven offshore equipment corrosion monitoring and safety early warning system according to claim 4 is characterized in that: The CTSAN neural network model includes a convolutional feature extraction layer, a spatial attention layer, a temporal attention layer and a classification layer; its working mode is: the multivariate time series X=[x1, x2..., x t ] Based on the convolutional feature extraction layer, the features of the time interval of interest are extracted, and the model focus is converted from a large number of time points to a small number of time interval sequences C = [C1, C2..., C t ]; then the time interval feature is sent to the spatial attention layer, and the global spatial feature C'=[C'1,C'2...,C' t ]; then the global spatial features of all intervals are sent to the temporal attention layer, and the global spatiotemporal features C" considering the weights of each interval are output; the global spatiotemporal features are then sent to the classification layer, and the predicted category y corresponding to the sequence is output (t) =softmax(W F C"+b F ); Finally, given a set of training examples, the coefficients of each layer in the model are iteratively updated, and finally the CTSAN corrosion fault classification model is obtained.

Images