Intelligent automatic diagnosis probe detection system

Abstract

The invention relates to the technical field of on-line detection probes, in particular to an intelligent detection system for automatically diagnosing probes, which is based on an on-line thickness measuring probe, detects equipment or pipeline thinning alarm, automatically grabs parameters including temperature, pressure, medium components, flow velocity and corrosion mechanism judgment, and can grab a chemical plant platform. And after grabbing, calculating the parameters by an algorithm on the basis of a corrosion mechanism model, calculating the corrosion mechanism of the thinned position, and giving a corresponding monitoring and detecting plan according to the calculated corrosion mechanism. According to the intelligent detection system for automatically diagnosing the probe, the system can monitor the thickness change of equipment or a pipeline on line in real time, an alarm is triggered immediately once a thinning phenomenon is found, and related parameters are automatically grabbed for analysis. The instant response capability greatly shortens the time from fault discovery to fault processing, and improves the working efficiency.

Description

Technical Field

[0001] The present invention relates to the technical field of online detection probes, and in particular to a detection system of intelligent automatic diagnosis probes. Background Art

[0002] Online thickness measurement probes are widely used in many industries, including steel, automobiles, machinery, light industry, etc. Various transparent and non-transparent pipes, plates, sheets, coatings, solder pastes, paint films, coatings, films, metal plates and other products in these industries all require real-time online thickness measurement. The application of probes is not limited to thickness measurement.

[0003] At present, the online detection probes on the market are mainly based on thickness measurement, and the result is the thinning rate (corrosion rate). It is impossible to form a corrosion mechanism model and give a targeted monitoring and detection plan. Corrosion is a complex physical and chemical process involving many factors such as medium properties, temperature, pressure, flow rate, etc. Evaluating the corrosion situation only by thickness measurement cannot fully reflect the mechanism and process of corrosion. The inability to form a corrosion mechanism model means that it is impossible to deeply understand the essential cause of corrosion, making it difficult to formulate effective anti-corrosion measures. For this reason, we propose an intelligent automatic diagnostic probe detection system. Summary of the invention

[0004] The purpose of the present invention is to solve the problems existing in the prior art and to propose an intelligent automatic diagnosis probe detection system.

[0005] In order to achieve the above-mentioned purpose, the present invention adopts the following technical scheme: an intelligent automatic diagnostic probe detection system, based on an online thickness measuring probe, detects thinning alarms of equipment or pipelines, automatically captures parameters for on-site and corrosion mechanism judgment, including temperature, pressure, medium composition, flow rate, and can be docked with a chemical plant platform for capture. After capture, the parameters are calculated through an algorithm based on a corrosion mechanism model (the accuracy of the corrosion mechanism model is improved through continuous calculation and improvement), and the corrosion mechanism existing at the thinning position is calculated, and a corresponding monitoring and detection plan is given according to the calculated corrosion mechanism.

[0006] As a further technical solution of the present invention, it includes: a data collection module, a data preprocessing module, a feature extraction module and a model building and training module;

[0007] The data collection module is used for data collection, and collects data through an electrochemical sensor, an environmental sensor and an image capture device installed on the diagnostic probe;

[0008] The data preprocessing module is used to clean and normalize the data collected by the data collection module, by cleaning the collected data, removing outliers and erroneous data, and normalizing data of different units and magnitudes to make the data on the same scale;

[0009] The feature extraction module is used for corrosion feature extraction, and extracts features directly related to corrosion from the collected data, including corrosion rate and corrosion type. The feature extraction module can calculate the change in wall thickness of the device to be detected within a time period to determine the corrosion rate, and determine the corrosion type based on the corrosion morphology image analysis of the device to be detected, and extract basic information of the device to be detected at the same time;

[0010] The model building and training module adopts a decision tree artificial intelligence model, and uses the data in the feature extraction module to train the model to evaluate the corrosion status of the equipment to be detected. The model building and training module evaluates the model performance through a cross-validation method, optimizes the model according to the evaluation results, and inputs the newly collected equipment data into the trained model. The model building and training module evaluates the current corrosion status of the equipment and predicts the best time for the next maintenance of the equipment.

[0011] As a further technical solution of the present invention, the electrochemical sensor is used to obtain corrosion potential and corrosion current density, the environmental sensor is used to obtain temperature, humidity, pH and salinity data, the image capture device is used to capture the morphology of the equipment to be detected, and the data collection module is also used to collect basic information of the equipment to be detected, including material composition, service life and past maintenance records; the data collection and preprocessing module is also used to collect basic information of the equipment to be detected, including material composition, service life and past maintenance records of the equipment to be detected.

[0012] As a further technical solution of the present invention, the feature extraction module is also used to extract environmental features, extracting the operating environment features of the equipment, including the chemical environment (medium composition, pH) and the physical environment (temperature, pressure, flow rate). The feature extraction module selects the extracted feature data and selects key features for extraction, and uses the statistical method of correlation analysis and the knowledge in this field to select the features that have the greatest impact on equipment corrosion and maintenance decisions.

[0013] As a further technical solution of the present invention, the model building and training module can output specific maintenance methods and estimate the resources required for maintenance based on the newly collected equipment data.

[0014] As a further technical solution of the present invention, the model building and training module further includes:

[0015] a. Prediction and evaluation unit, which is used to input the newly collected equipment data into the trained and optimized decision tree model to evaluate the current corrosion status of the equipment and predict the best time for the next maintenance of the equipment;

[0016] b. A cross-validation unit, used to evaluate the performance of the trained decision tree model through a cross-validation method;

[0017] c. Optimization unit, which optimizes the decision tree model according to the evaluation results provided by the cross-validation unit.

[0018] As a further technical solution of the present invention, during the construction process of the decision tree model, the data set is divided by recursively selecting the optimal features until a stopping condition is met, and the stopping condition includes but is not limited to the number of samples in the node being less than a preset threshold and the purity of the divided subsets reaching a preset standard; the cross-validation unit adopts a K-fold cross-validation method to divide the data set into K non-overlapping subsets, and each subset is used as a test set in turn, and the remaining subsets are used as training sets, and K training and testing are performed to evaluate the performance of the decision tree model. The optimization unit performs a pruning operation on the decision tree model according to the evaluation results provided by the cross-validation unit, and the optimization unit is used to improve the generalization ability of the decision tree model.

[0019] As a further technical solution of the present invention, the cross-validation unit calculates the error and accuracy index of the model on the test set through a performance evaluation formula;

[0020] The cross validation error formula is expressed as:

[0021] CV=(1 / N)*Σ(yy^)2

[0022] in:

[0023] CV stands for cross validation error;

[0024] N represents the number of data samples;

[0025] y represents the actual observed value;

[0026] y^ represents the model prediction value.

[0027] The cross-validation error formula is used to calculate the average error between the actual observation value and the model prediction value, so as to evaluate the performance of the model. In K-fold cross-validation, this formula will be applied K times, each time using a different test set and training set, and finally the average of the K results is taken as the final performance evaluation result of the model.

[0028] As a further technical solution of the present invention, the prediction and evaluation unit further includes a maintenance suggestion generation subunit, and the maintenance suggestion generation subunit is used to generate corresponding maintenance suggestions according to the predicted optimal time for the next maintenance of the equipment.

[0029] As a further technical solution of the present invention, it also includes a user interface module, which is used to display the current corrosion status assessment result of the equipment, the predicted best time for the next maintenance of the equipment, and maintenance suggestions.

[0030] The intelligent automatic diagnostic probe detection system proposed by the present invention has the following beneficial effects:

[0031] The intelligent automatic diagnostic probe detection system of the present application can monitor the thickness change of equipment or pipelines online in real time. Once the thinning phenomenon is found, the alarm is immediately triggered and the relevant parameters are automatically captured for analysis. This instant response capability greatly shortens the time from fault discovery to processing and improves work efficiency.

[0032] The intelligent automatic diagnostic probe detection system of the present application integrates the corrosion mechanism model and advanced algorithms, and the system can accurately determine the corrosion mechanism existing in the thinning position. This precise diagnostic capability helps enterprises to understand the corrosion status of equipment or pipelines more accurately, so as to take targeted maintenance measures. The system not only provides the diagnosis of corrosion mechanism, but also automatically generates corresponding monitoring and detection plans based on the diagnosis results. The system can capture key parameters such as temperature, pressure, medium composition, flow rate, etc., and connect with the chemical plant platform. This data integration capability enables enterprises to have a more comprehensive understanding of various information in the production process, providing strong support for production management and optimization. The accuracy of the corrosion mechanism model is improved through continuous calculation and improvement, and the diagnostic capability of the system will continue to improve over time. The characteristics of continuous improvement enable the system to better adapt to the ever-changing production environment and corrosion conditions. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] Figure 1 This is a schematic diagram of the system architecture of an intelligent automatic diagnostic probe detection system proposed by the present invention. DETAILED DESCRIPTION

[0034] The technical solutions in the embodiments of the present invention will be described clearly and completely below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments.

[0035] An intelligent automatic diagnostic probe detection system is based on an online thickness measuring probe. It detects thinning alarms of equipment or pipelines and automatically captures parameters for on-site and corrosion mechanism judgment, including temperature, pressure, medium composition, flow rate, and can be connected to a chemical plant platform for capture. After capture, the parameters are calculated through an algorithm based on a corrosion mechanism model (the accuracy of the corrosion mechanism model is improved through continuous calculation and improvement), and the corrosion mechanism existing at the thinning position is calculated, and a corresponding monitoring and detection plan is given based on the calculated corrosion mechanism.

[0036] Reference Figure 1 As shown, it includes: data collection module, data preprocessing module, feature extraction module and model building and training module;

[0037] The data collection module is used for data collection. It collects data through electrochemical sensors, environmental sensors and image capture devices installed on the diagnostic probe. The data collection module connects to the Party A platform to connect to DCS and LIMS. It collects temperature, pressure, flow, etc. in real time through DCS. LIMS requires Party A to upload test and analysis data at regular intervals and collects data on a periodic basis. For basic information of equipment / pipelines, enter it according to the pipeline / equipment ledger (material, thickness, corrosion allowance, service life, etc.). If there is a material change, replacement information must be provided.

[0038] The data preprocessing module is used to clean and normalize the data collected by the data collection module. By cleaning the collected data, outliers and erroneous data are removed, and data of different units and magnitudes are normalized to make the data on the same scale; the collected data is normalized, such as flow rate needs to be converted into flow velocity, and the unit is unified. For example, the corrosion current density data is converted from milliamperes per square centimeter to a standard dimensionless value so that it can be analyzed together with other parameters.

[0039] The feature extraction module is used for corrosion feature extraction. It extracts features directly related to corrosion from the collected data, including corrosion rate and corrosion type. The feature extraction module can calculate the change in wall thickness of the equipment to be tested within a time period to determine the corrosion rate, and judge the corrosion type based on the corrosion morphology image analysis of the equipment to be tested, and extract the basic information of the equipment to be tested; for example, the collected chemical environment (medium composition, pH, particle size, etc.) and physical environment (temperature, pressure, flow rate, etc.) of the pipeline / equipment are extracted.

[0040] When an alarm occurs, the data collected over a long period of time is extracted, including basic information of equipment / pipelines, real-time capture of DCS data (temperature, pressure, flow) when an alarm occurs, and capture of LIMS data for the current cycle and the previous two cycles: Compare the three-cycle test data to see if there is a large fluctuation or compare the test data of this cycle with the design threshold.

[0041] The model building and training module adopts a decision tree artificial intelligence model and uses the data in the feature extraction module to train the model and evaluate the corrosion status of the equipment to be tested. The model building and training module evaluates the model performance through cross-validation method, optimizes the model according to the evaluation results, and inputs the newly collected equipment data into the trained model. The model building and training module evaluates the current corrosion status of the equipment and predicts the best time for the next maintenance of the equipment.

[0042] Electrochemical sensors are used to obtain corrosion potential and corrosion current density, environmental sensors are used to obtain temperature, humidity, pH and salinity data, image capture devices are used to capture the morphology of the equipment to be tested, and the data collection module is also used to collect basic information of the equipment to be tested, including material composition, service life and past maintenance records; the data collection and preprocessing module is also used to collect basic information of the equipment to be tested, including material composition, service life and past maintenance records of the equipment to be tested. For example, in a chemical pipeline system, sensors will collect data on the medium composition, flow rate, temperature, pressure (DCS data) in the real-time transmission pipeline and basic information of the equipment, such as material composition, service life, past maintenance records (ledgers), etc.

[0043] The feature extraction module is also used to extract environmental features, extracting the operating environment features of the equipment, including the chemical environment (medium composition, pH) and physical environment (temperature, pressure, flow rate). The feature extraction module selects the extracted feature data and selects key features for extraction, and uses the statistical method of correlation analysis and knowledge in this field to select the features that have the greatest impact on equipment corrosion and maintenance decisions. For example, the concentration of ammonia ions, pressure, temperature in the hydrochloric acid pipeline and the metal material of the equipment are key features because they are closely related to the corrosion process.

[0044] The model building and training module can output specific maintenance methods and estimate the resources required for maintenance based on the newly collected equipment data. According to the judgment method of the corrosion mechanism, capture: temperature, pressure, flow rate, sensitive materials, etc. For example, a set of data including the temperature, pressure, medium corrosivity and corresponding maintenance time and maintenance method of the chemical reactor is input into the model for training. If it is found that the model has a low accuracy rate in predicting a certain type of corrosion, the parameters of the model can be adjusted or more relevant data can be added for training. For metal equipment with slight surface corrosion, it is recommended to use the method of re-applying anti-corrosion paint after surface grinding; for equipment with pitting, it is recommended to use local repair methods (such as repair welding, filling repair materials). Resource demand estimation, including manpower, materials (anti-corrosion coatings, repair materials, etc.), tools (welding equipment, grinding tools, etc.).

[0045] The model building and training module further includes:

[0046] a. Prediction and evaluation unit, which is used to input the newly collected equipment data into the trained and optimized decision tree model to evaluate the current corrosion status of the equipment and predict the best time for the next maintenance of the equipment;

[0047] b. A cross-validation unit, used to evaluate the performance of the trained decision tree model through a cross-validation method;

[0048] c. Optimization unit, which optimizes the decision tree model according to the evaluation results provided by the cross-validation unit.

[0049] During the construction of the decision tree model, the data set is divided by recursively selecting the optimal features until the stopping condition is met. The stopping condition includes but is not limited to the number of samples in the node being less than the preset threshold and the purity of the divided subsets reaching the preset standard; the cross-validation unit uses the K-fold cross-validation method to divide the data set into K non-overlapping subsets, and each subset is used as a test set in turn, and the remaining subsets are used as training sets, and K training and testing are performed to evaluate the performance of the decision tree model. The optimization unit prunes the decision tree model according to the evaluation results provided by the cross-validation unit. The optimization unit is used to improve the generalization ability of the decision tree model.

[0050] The cross-validation unit calculates the error and accuracy indicators of the model on the test set through the performance evaluation formula; the cross-validation error formula is expressed as:

[0051] CV=(1 / N)*Σ(yy^)2

[0052] in:

[0053] CV stands for cross validation error;

[0054] N represents the number of data samples;

[0055] y represents the actual observed value;

[0056] y^ represents the model prediction value.

[0057] The cross-validation error formula is used to calculate the average error between the actual observation value and the model prediction value, so as to evaluate the performance of the model. In K-fold cross-validation, this formula will be applied K times, each time using a different test set and training set, and finally the average of the K results is taken as the final performance evaluation result of the model.

[0058] Suppose we have 5 data samples (N=5), and the actual observed value (y) and model predicted value (y^) of each sample are as follows:

[0059] Sample 1: y1=3, y^1=2.5

[0060] Sample 2:y 2 =5,y^2 =4.8

[0061] Sample 3:y 3 =7,y^ 3 =6.2

[0062] Sample 4:y 4 =9,y^ 4 =9.5

[0063] Sample 5:y 5 =11,y^ 5 =10.1

[0064] Now, we calculate the error using the formula for cross validation error (CV):

[0065] CV=(1 / N)*Σ(yy^)2; calculate the square of the error of each sample and sum them up:

[0066]

[0067] Sum the squares of these errors:

[0068]

[0069] Finally, we calculate the CV by dividing the sum by the number of samples, N (5):

[0070]

[0071] Therefore, in this example, the cross validation error (CV) is 0.4.

[0072] The prediction and evaluation unit also includes a maintenance suggestion generation subunit, which is used to generate corresponding maintenance suggestions based on the predicted optimal time for the next maintenance of the equipment.

[0073] It also includes a user interface module, which is used to display the current corrosion status assessment results of the equipment, the predicted best time for the next maintenance of the equipment, and maintenance suggestions

[0074] Working principle: Based on the online thickness measuring probe, the thinning alarm of equipment / pipeline is detected, and the parameters for on-site and corrosion mechanism judgment are automatically captured (such as temperature, pressure, medium composition, flow rate, etc., which can be connected to the chemical plant platform for capture). After capturing, the parameters are calculated based on the corrosion mechanism model (the accuracy of the corrosion mechanism model is improved through continuous calculation and improvement) through the algorithm, and the corrosion mechanism existing in the thinning position is calculated, and the corresponding monitoring and detection plan is given according to the calculated corrosion mechanism.

[0075] The system can monitor the thickness changes of equipment or pipelines online in real time. Once thinning is found, an alarm will be triggered immediately and relevant parameters will be automatically captured for analysis. This instant response capability greatly shortens the time from fault discovery to processing and improves work efficiency.

[0076] By integrating the corrosion mechanism model and advanced algorithms, the system can accurately determine the corrosion mechanism existing in the thinning position. This precise diagnostic capability helps enterprises to understand the corrosion status of equipment or pipelines more accurately, so as to take targeted maintenance measures. The system not only provides the diagnosis of corrosion mechanism, but also automatically generates corresponding monitoring and inspection plans based on the diagnosis results. The system can capture key parameters such as temperature, pressure, medium composition, flow rate, etc., and connect with the chemical plant platform. This data integration capability enables enterprises to have a more comprehensive understanding of various information in the production process, providing strong support for production management and optimization. The accuracy of the corrosion mechanism model is improved through continuous calculation and improvement. The diagnostic capability of the system will continue to improve over time. The continuous improvement feature enables the system to better adapt to the ever-changing production environment and corrosion conditions.

[0077] Through accurate diagnosis and intelligent decision support, the system can help enterprises to detect and deal with potential corrosion problems in a timely manner, thus avoiding serious damage and downtime of equipment or pipelines. This helps to reduce the maintenance costs and production interruption risks of enterprises.

[0078] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can make equivalent replacements or changes according to the technical scheme and inventive concept of the present invention within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. An intelligent automatic diagnostic probe detection system, characterized in that: Based on the online thickness measuring probe, the thinning alarm of equipment or pipeline is detected, and the parameters for on-site and corrosion mechanism judgment are automatically captured, including temperature, pressure, medium composition, flow rate, and can be connected to the chemical plant platform for capture. After capture, the parameters are calculated based on the corrosion mechanism model (the accuracy of the corrosion mechanism model is improved through continuous calculation and improvement) through the algorithm, and the corrosion mechanism existing in the thinning position is calculated, and the corresponding monitoring and detection plan is given according to the calculated corrosion mechanism.

2. The intelligent automatic diagnostic probe detection system according to claim 1, characterized in that: include: Data collection module, data preprocessing module, feature extraction module and model building and training module; The data collection module is used for data collection, and collects data through an electrochemical sensor, an environmental sensor and an image capture device installed on the diagnostic probe; The data preprocessing module is used to clean and normalize the data collected by the data collection module, by cleaning the collected data, removing outliers and erroneous data, and normalizing data of different units and magnitudes to make the data on the same scale; The feature extraction module is used for corrosion feature extraction, and extracts features directly related to corrosion from the collected data, including corrosion rate and corrosion type. The feature extraction module can calculate the change in wall thickness of the device to be detected within a time period to determine the corrosion rate, and determine the corrosion type based on the corrosion morphology image analysis of the device to be detected, and extract basic information of the device to be detected at the same time; The model building and training module adopts a decision tree artificial intelligence model, and uses the data in the feature extraction module to train the model to evaluate the corrosion status of the equipment to be detected. The model building and training module evaluates the model performance through a cross-validation method, optimizes the model according to the evaluation results, and inputs the newly collected equipment data into the trained model. The model building and training module evaluates the current corrosion status of the equipment and predicts the best time for the next maintenance of the equipment.

3. The intelligent automatic diagnostic probe detection system according to claim 2, characterized in that: The electrochemical sensor is used to obtain corrosion potential and corrosion current density, the environmental sensor is used to obtain temperature, humidity, pH and salinity data, the image capture device is used to capture the morphology of the equipment to be detected, and the data collection module is also used to collect basic information of the equipment to be detected, including material composition, service life and past maintenance records; the data collection and preprocessing module is also used to collect basic information of the equipment to be detected, including material composition, service life and past maintenance records of the equipment to be detected.

4. The intelligent automatic diagnostic probe detection system according to claim 1, characterized in that: The feature extraction module is also used to extract environmental features, extracting the operating environment features of the equipment, including the chemical environment (medium composition, pH) and the physical environment (temperature, pressure, flow rate). The feature extraction module selects the extracted feature data and selects key features for extraction, and uses the statistical method of correlation analysis and knowledge in this field to select the features that have the greatest impact on equipment corrosion and maintenance decisions.

5. The intelligent automatic diagnostic probe detection system according to claim 1, characterized in that: The model building and training module can output specific maintenance methods and estimate the resources required for maintenance based on the newly collected equipment data.

6. The intelligent automatic diagnostic probe detection system according to claim 5, characterized in that: The model building and training module further includes: a. Prediction and evaluation unit, which is used to input the newly collected equipment data into the trained and optimized decision tree model to evaluate the current corrosion status of the equipment and predict the best time for the next maintenance of the equipment; b. A cross-validation unit, used to evaluate the performance of the trained decision tree model through a cross-validation method; c. Optimization unit, which optimizes the decision tree model according to the evaluation results provided by the cross-validation unit.

7. The intelligent automatic diagnostic probe detection system according to claim 6, characterized in that: During the construction process of the decision tree model, the data set is divided by recursively selecting the optimal features until a stopping condition is met, and the stopping condition includes but is not limited to the number of samples in the node being less than a preset threshold and the purity of the divided subsets reaching a preset standard; the cross-validation unit adopts a K-fold cross-validation method to divide the data set into K non-overlapping subsets, and each subset is used as a test set in turn, and the remaining subsets are used as training sets, and K training and testing are performed to evaluate the performance of the decision tree model; the optimization unit performs a pruning operation on the decision tree model according to the evaluation result provided by the cross-validation unit, and the optimization unit is used to improve the generalization ability of the decision tree model.

8. The intelligent automatic diagnostic probe detection system according to claim 7, characterized in that: The cross-validation unit calculates the error and accuracy index of the model on the test set through a performance evaluation formula; The cross validation error formula is expressed as: CV=(1 / N)*Σ(yy^)2 in: CV stands for cross validation error; N represents the number of data samples; y represents the actual observed value; y^ represents the model prediction value. The cross-validation error formula is used to calculate the average error between the actual observation value and the model prediction value, so as to evaluate the performance of the model. In K-fold cross-validation, this formula will be applied K times, each time using a different test set and training set, and finally the average of the K results is taken as the final performance evaluation result of the model.

9. The intelligent automatic diagnostic probe detection system according to claim 8, characterized in that: The prediction and evaluation unit further includes a maintenance suggestion generating subunit, which is used to generate corresponding maintenance suggestions according to the predicted optimal time for the next maintenance of the equipment.

10. The intelligent automatic diagnostic probe detection system according to claim 9, characterized in that: It also includes a user interface module, which is used to display the current corrosion status assessment results of the equipment, the predicted best time for the next maintenance of the equipment, and maintenance suggestions.

Images