Bed height detection method based on machine learning

Abstract

The invention provides a bed height detection method based on machine learning, and the method comprises the steps: S1, collecting the historical data of a chlorination furnace through a DCS system, and forming a corresponding sample set; s2, regression simulation is carried out by adopting multiple algorithms, and a model is established; s3, evaluating the model to obtain an optimal model as an output model; s4, deploying the optimal model as an application template; s5, packaging the application template, and opening the application template through an API (Application Program Interface) for external calling; s6, the API is connected with a chlorination furnace data monitoring system, so that the application template can call the monitoring data of the chlorination furnace, and an execution result is displayed on an interactive interface. By fusing multi-source sensor data, establishing an intelligent calculation model and combining a mechanism model and a machine learning algorithm, a calculation result always fits an actual production state, and finally accurate process parameter guidance is provided for a chlorination reaction process through high-precision bed height data, so that the product quality stability is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of chemical product production, and in particular to a bed height detection method based on machine learning. Background Art

[0002] In the chlorination process, chlorination bed height is a key process parameter, and its accuracy directly affects reaction efficiency, energy consumption, and product quality. Currently, the bed height calculated using the pressure differential formula exhibits significant discrepancies, leading to deviations between the measured value and actual operating conditions. These issues, primarily manifested in significant bed height differences and large measurement errors, severely interfere with precise control of the production process.

[0003] Patent publication number CN104019780A discloses a method for measuring the bed height of a boiling chlorination furnace using bed pressure drop. The method comprises the following steps: symmetrically placing six pressure pipes at the top of the chlorination furnace, at the bottom of the bed, and at a height of one meter from the bottom of the bed; installing differential pressure transmitters between the pressure pipes to measure the bed pressure drop; programming a DCS control program based on the bed pressure drop relationship ΔP = φρgh and the bed height relationship h1 = ΔP1 / ΔP3 and h2 = ΔP2 / ΔP4; and connecting the differential pressure transmitter to the DCS control system in the control room to collect data, calculate the bed heights h1 and h2, and display them on the DCS computer control interface. This patent can estimate the bed height based on the pressure difference. However, since the pressure difference formula assumes the bed to be an infinitely large flat plate, it ignores the friction on the equipment wall and the fluid retention at the corners. In addition, the chlorination reaction is a highly exothermic process, and the drastic change in temperature will change the porosity of the bed, resulting in a significant drop in the bed height calculated by the pressure difference formula, which seriously affects the precise control of the chlorination furnace production process. Summary of the Invention

[0004] In view of this, the present invention aims to propose a bed height detection method based on machine learning. Through efficient data processing and intelligent analysis, the accuracy and reliability of process parameters in the production process can be improved, the problem of large measurement errors in the traditional pressure difference method can be solved, a bed height calculation model that is more in line with actual production can be established, and the accuracy and reliability of production process parameter control can be improved, providing the industry with a popularizable intelligent solution.

[0005] The present invention discloses a bed height detection method based on machine learning, comprising:

[0006] Step S1: using the DCS system to collect historical data of the chlorination furnace to form a corresponding sample set;

[0007] Step S2: Using multiple algorithms to perform regression simulation on the data of the sample set and establish corresponding models;

[0008] Step S3: Evaluate the models obtained by using different algorithms in step S2, and use the optimal model as the output model;

[0009] Step S4: Deploy the optimal model as an application template, and automatically call all data when the application template is run;

[0010] Step S5: Encapsulate the deployed application template through the DCS platform, and open the encapsulated model through the API interface for external calls;

[0011] Step S6: Connect the API interface to the chlorination furnace data monitoring system so that the application template can call the monitoring data of the chlorination furnace and display the latest execution results on the interactive interface.

[0012] Furthermore, the historical data of the chlorination furnace in step S1 at least include: chlorination furnace temperature, chlorine gas flow rate, furnace top pressure, furnace bottom pressure and bed height value.

[0013] Furthermore, in step S1, the historical data is preprocessed, including at least defect processing and data standardization processing.

[0014] Furthermore, the multiple algorithms in step S2 include one or more of nonlinear SVM regression, linear regression, random forest regression, and BP neural network regression.

[0015] Furthermore, the output model in step S3 is a bed height prediction model obtained by regression simulation using a random forest regression algorithm.

[0016] Furthermore, in step S3, the average relative error is calculated by the maximum relative error and the minimum relative error of each model. The closer the average relative error value is to 0, the healthier the model is. The model with the average relative error closest to 0 is determined as the optimal model.

[0017] Furthermore, in step S5, the application template is encapsulated by the service designer built into the DCS platform.

[0018] Furthermore, the detection method further comprises:

[0019] Step S7: Set a timer task to call the model at intervals of a first preset time and update the result value.

[0020] Furthermore, the first preset time is 8-12 minutes.

[0021] Furthermore, the first preset time is 10 minutes.

[0022] Compared with the existing technology, the bed height detection method based on machine learning described in the present invention has the following advantages:

[0023] (1) The present invention eliminates the calculation errors of the differential pressure method caused by factors such as gas flow rate fluctuations and uneven particle distribution by fusing multi-source sensor data and establishing an intelligent calculation model, achieving a measurement accuracy of bed height within ±2%. By combining the mechanism model with the machine learning algorithm, a dynamic calculation model is developed that can adapt to changes in raw materials and fluctuations in working conditions in real time, so that the calculation results always fit the actual production status. Ultimately, through high-precision bed height data, accurate process parameter guidance is provided for the chlorination reaction process, the reaction efficiency is optimized, and the product quality stability is improved;

[0024] (2) Using a large amount of historical data parameters and data science to perform computational analysis and optimization to find the optimal solution and guide on-site production;

[0025] (3) Build an environment based on the DCS industrial Internet platform to achieve rapid deployment of bed height calculation models, optimize production process operating parameters, reduce energy consumption, and improve economic benefits;

[0026] (4) Promote enterprise informatization construction, effectively integrate production and information resources, and enhance the guidance of informatization on process production. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. 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 paying any creative work.

[0028] Figure 1 This is a flow chart of the bed height detection method based on machine learning according to an embodiment of the present invention. DETAILED DESCRIPTION

[0029] In order to make the technical means, objectives and effects of the present invention easier to understand, the embodiments of the present invention are described in detail below.

[0030] It should be noted that all directional and positional terms in the present invention, such as "up," "down," "left," "right," "front," "back," "vertical," "horizontal," "inside," "outside," "top," "low," "lateral," "longitudinal," and "center," are used only to explain the relative positional relationships and connections between components in a specific state. They are intended solely to facilitate the description of the present invention and do not require that the present invention be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on the present invention. Furthermore, terms such as "first" and "second" in the present invention are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly indicating the number of the technical features indicated.

[0031] In the description of the present invention, unless otherwise expressly specified or limited, the terms "mounted," "connected," and "connected" should be understood broadly. For example, they may refer to fixed, detachable, or integral connections; mechanical connections; direct connections or indirect connections through an intermediary; and internal communication between two components. Those skilled in the art will understand the specific meanings of these terms in the present invention based on the specific circumstances.

[0032] Throughout this specification, reference to terms such as "one embodiment," "some embodiments," "illustrative embodiments," "examples," "specific examples," or "some examples" means that a specific feature, structure, material, or characteristic described in conjunction with that embodiment or example is included in at least one embodiment or example of the present invention. In this specification, illustrative uses of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples.

[0033] Example 1

[0034] The present invention discloses a bed height detection method based on machine learning, comprising:

[0035] Step S1: using the DCS system to collect historical data of the chlorination furnace to form a corresponding sample set;

[0036] Step S2: Using multiple algorithms to perform regression simulation on the data of the sample set and establish corresponding models;

[0037] Step S3: Evaluate the models obtained by using different algorithms in step S2, and use the optimal model as the output model;

[0038] Step S4: Deploy the optimal model as an application template, and automatically call all data when the application template is run;

[0039] Step S5: Encapsulate the deployed application template through the DCS platform, and open the encapsulated model through the API interface for external calls;

[0040] Step S6: Connect the API interface to the chlorination furnace data monitoring system so that the application template can call the monitoring data of the chlorination furnace, predict the bed height according to the monitoring data, and display the latest bed height prediction result on the interactive interface.

[0041] In the prior art, the bed height in the chlorination furnace is usually predicted and calculated by the pressure difference method, and the calculation formula is shown in formula (1):

[0042] (1)

[0043] Where H is the bed height, ΔP is the pressure difference, ρ is the average density of chlorine, titanium tetrachloride, nitrogen, carbon monoxide, and carbon dioxide gases in the bed, and g is the acceleration due to gravity.

[0044] During actual operation, the bed height is also manually measured when the chlorination furnace is shut down. It is found that there is a difference between the bed height and the bed height predicted by the pressure difference method, with the difference value ranging from 1.2 to 1.3 cm. This leads to a deviation between the measured value and the actual operating conditions, seriously affecting the precise control of the production process. In this example, a mechanism model is built through the DCS data science platform to achieve accurate calculation of the bed height, making the measurement results closer to the actual production conditions. By building an efficient data processing system and intelligent analysis model, the measurement accuracy and control reliability of the process parameters are significantly improved, making the calculation results closer to the actual conditions, enhancing the reliability and operability of production guidance, and providing a scalable intelligent solution for the industry. After a certain period of operation, it is found that the bed height predicted by the incentive model formed in this example and the bed height detected by the shutdown have a difference of about 0.6 cm, which significantly improves the prediction accuracy of the bed height.

[0045] Specifically, the historical data of the chlorination furnace includes at least: chlorination furnace temperature, chlorine gas flow rate, furnace top pressure, furnace bottom pressure and bed height value. In this example, the chlorination furnace temperature, chlorine gas flow rate, furnace top pressure and furnace bottom pressure are input as independent variables, and the bed height value is input as the dependent variable. The bed height detection value is the bed height value predicted by the pressure difference formula. It should be noted that due to the small number of shutdowns in a year, in some cases, the shutdown does not exceed 6 times a year, resulting in insufficient bed height data measured by manual shutdown, which is not enough to support large sample size training. Model training can only be performed using the predicted value of the pressure difference method. However, in actual operation, it was found that the bed height value predicted by the method provided in this example is closer to the manual detection value at the time of shutdown than the bed height value calculated by the pressure difference method. In other words, the bed height predicted by the model formed by this example is closer to the actual value, which is conducive to improving the precise control of the production process. As an optional and preferred example, the bed height detection value includes only the bed height measured during manual shutdown, and does not include the bed height value predicted by the pressure differential formula. It should be noted that this requires a large amount of manual measurement data to achieve this. This setting can further improve the accuracy of the dependent variable input value, thereby facilitating a more accurate bed height prediction. Furthermore, the chlorination furnace temperature, chlorine flow rate, furnace top pressure, and furnace bottom pressure can be detected by multiple sensors in the prior art. The detection frequency can be referenced in the prior art and will not be further elaborated here.

[0046] In step S1, historical data must be preprocessed, including at least defect handling and data standardization. Specifically, data preprocessing includes various methods, such as defect handling, data design, and data replication. These settings can improve data quality. Preprocessing abnormal data can reduce its impact on prediction results, making the data more accurate, reliable, and complete, facilitating subsequent analysis and processing.

[0047] Data preprocessing can adopt the processing methods used in the prior art such as filtering method or normalization method, which are not limited here. Specifically, defect processing includes deleting blank lines, deleting abnormal data, etc. In the field of chlorination furnace operation, common defects currently include: data missing defects during shutdown and maintenance, blank line defects caused by data signal loss, abnormal data caused by damage to instruments and their accessories, etc. The general missing value processing for the above defects includes replacing the variables with missing values ​​with maximum value, minimum value, average value, mode or a custom value. The blank line is to directly delete the entire row of data with missing values. For obviously abnormal data, it can be deleted or replaced with maximum value, minimum value, average value, mode or a custom value. It should be noted that a corresponding threshold range or graphic range can be set for the identification of abnormal values. The values ​​exceeding the corresponding area are identified as abnormal values ​​and can be deleted or replaced. Data design refers to the structured design and feature engineering of the data set before model construction to make the data suitable for model training. Data replication refers to the data augmentation operation performed during model training to enhance the robustness of the model, thereby improving the generalization ability of the model and providing a high-quality and diverse input data foundation for subsequent model training.

[0048] In this example, the multiple algorithms in step S2 include one or more of nonlinear SVM regression, linear regression, random forest regression, and BP neural network regression. In these algorithms, key input parameters include several independent variable parameters, such as the temperature in the chlorination furnace, chlorine flow rate, furnace top pressure, and furnace bottom pressure, as well as bed height as a dependent variable parameter. These algorithms perform error assessments based on prefabricated data sets and self-prediction. For data sets with the same independent and dependent variables, the performance of a regression algorithm under one or more parameter combinations, or the performance gap between multiple regression algorithms, is compared to test the reliability of the regression models. Ultimately, the optimal regression model is obtained based on evaluation metrics (such as relative error, mean absolute error, root mean square error, etc.) or graphical presentation.

[0049] In step S3, the average relative error is calculated by the maximum relative error and the minimum relative error of each model. The closer the average relative error value is to 0, the higher the model health. The model with the average relative error closest to 0 is determined as the optimal model. Through the above settings, the optimal regression model can be screened out from multiple algorithms, thereby effectively reducing the measurement error in the work, providing accurate process parameter guidance for the chlorination reaction process, optimizing reaction efficiency, and improving product quality stability.

[0050] Specifically, in step S4, automatically calling all data using the template means that it can automatically call all input chlorination furnace temperature, chlorine gas flow rate, furnace top pressure, furnace bottom pressure, and bed height detection value data. The bed height detection value data at this time includes the bed height value in the previously collected data set, and also includes the bed height value predicted by the parameters such as chlorination furnace temperature, chlorine gas flow rate, furnace top pressure, and furnace bottom pressure obtained when calling the application module through the interface.

[0051] In step S5, the application template is encapsulated by the service designer built into the DCS platform.

[0052] The business designer can perform related operations such as visual configuration management and dynamic rule orchestration. Among them, visual configuration management can provide a graphical interface, so that rules can be quickly built by dragging and dropping, instead of manually writing configuration files, lowering the configuration threshold and reducing human errors. Dynamic rule orchestration supports visual orchestration of business logic (such as traffic distribution, circuit breaker rules, service dependencies, etc.), and after successful orchestration, it takes effect in real time without the need for restart, significantly improving ease of use.

[0053] Specifically, after the interface call in step S6 is successful, the value called by the mechanism model (ie, application template) can be written back to the corresponding attribute value of the object instance, thereby obtaining the latest execution result value and displaying it on the interactive interface.

[0054] As one of the preferred embodiments, the output model is a bed height prediction model obtained through regression simulation using a random forest regression algorithm. Specifically, during testing, historical DCS data (e.g., relevant data from the last three months, totaling 9095 sets) is first exported for preprocessing to set the quantity, minimum sample size, and maximum sample size. Blank rows and abnormal data are deleted, retaining normal values. The independent variables are set as furnace temperature, chlorine flow rate, furnace top pressure, and furnace bottom pressure. The data range needs to be set based on the actual production process on site. The dependent variable is bed height. A random forest regression algorithm is then used to simulate and calculate the maximum and minimum relative errors to evaluate and calculate the average relative error. The closer the error value is to 0, the higher the healthiness. The average relative error of the random forest regression algorithm, as evaluated by the aforementioned algorithms, is 0.11678369774912493, with an assessed model healthiness of 83.56%. Its prediction accuracy is relatively better than other models. Therefore, the bed height prediction model obtained using the forest regression algorithm is used as the preferred model. Table 1 shows the comparison between the bed height predicted by the bed height prediction model obtained by regression simulation using the random forest regression algorithm and the actual detection data. Due to the large number of data sets, only 10 sets of data are extracted here for display.

[0055] Table 1 Comparison of detection values ​​and predicted data

[0056] Specifically, the random forest regression algorithm is processed in the following way:

[0057] The impurity of the parent node is calculated by formula (2):

[0058] (2)

[0059] Among them, MSE P is the mean square error of the parent node, which is used to measure the impurity of the parent node, n is the number of samples of the parent node, y i is the value of the dependent variable at the i-th node, y n is the average value of n dependent variables at the parent node.

[0060] The weighted sum of the sub-node MSE is calculated using formula (3):

[0061] (3)

[0062] Among them, MSE C is the weighted sum of the mean squared error of the child nodes, MSE L is the mean square error of the left child node, MSE R is the mean square error of the right child node, n L is the number of samples of the left child node, n R is the number of samples of the right child node. L 、MSE R The calculation principle and MSE P The same is no longer limited here.

[0063] Among them, the splitting rule of random forest regression is as follows:

[0064] Traverse all features and all possible thresholds, find the splitting method that maximizes the value of formula (4), and complete the node splitting:

[0065] (4)

[0066] Among them, ΔMSE is the mean square error MSE of the parent node P and the weighted sum MSE of the child nodes MSE C The difference.

[0067] It should be noted that when a child node meets any of the following conditions, it stops splitting and becomes a leaf node:

[0068] 1. The number of checkpoint samples is less than the preset node sample threshold. The node sample threshold is a preset value. Optionally, the node sample threshold can be 10 or other values.

[0069] 2. When the split depth reaches the preset depth, the pre-trial depth is the preset value. Optionally, the preset depth is 5 or other values;

[0070] 3. When the purity reaches the preset purity threshold, the preset purity threshold is a preset value. Optionally, the preset depth is 0.5 or other values. The purity is the mean square error corresponding to the node, which can be calculated according to the principle of formula (2);

[0071] 4. When ΔMSE reaches a preset threshold, optionally, the preset threshold of ΔMSE is 0.1 or another value. In this case, it indicates that the improvement of the split is insufficient and the improvement of subsequent splits is not significant.

[0072] In addition, the random forest algorithm also requires the preset number of decision trees m. Each decision tree uses the autonomous sampling method to extract samples from the collected data set and is independently constructed. The output value of the final model is calculated according to formula (5):

[0073] (5)

[0074] Among them, y is the predicted value output by the random forest regression model, m is the preset number of decision trees, and y j The predicted value output by the jth decision tree. m is preset before the model is built. Optionally, m is 10 or another preset value.

[0075] Specifically, the detection method further includes:

[0076] Step S7: Set a timer task to call the model at intervals of a first preset time and update the result value.

[0077] The scheduled task executes the corresponding model call at a preset interval or time point. By feeding the newly detected chlorination furnace parameters into the object instance, replacing the corresponding attribute values ​​in the model (chlorination furnace temperature, chlorine flow rate, and pressure difference), the latest bed height measurement results are calculated. This setting allows for regular monitoring of the latest bed height values, providing accurate process parameter guidance for the chlorination reaction process, optimizing reaction efficiency, and improving product quality stability.

[0078] Specifically, the first preset time is 8-12 minutes. Preferably, the first preset time is 10 minutes. Through the above settings, the detection system transmits the detected chlorination furnace temperature, chlorine flow rate, and pressure difference to the called model every 8-12 minutes, thereby obtaining the latest detection values, making it easier for operators to grasp bed height information in real time and provide accurate guidance for subsequent operations.

[0079] In practice, users can use the DCS system to build models based on the platform's data science module, quickly build and publish them by dragging and dropping, and then encapsulate services through the business designer. The DCS platform is used to build built-in components for the model and establish process optimization flows. At the same time, the platform creates applications including: data input and output, data preprocessing, signal processing, feature engineering, machine learning algorithms, customization, model evaluation, etc. During the design phase of model building, in a componentized form, through the visual components provided by the platform, users can drag and drop components to any location on the page. In the editing panel of the corresponding component, functions such as sample input and algorithm optimization can be performed.

[0080] After building the model, use the business designer to encapsulate the service. The business designer allows users to define the service's input and output parameters, as well as the logical flow of service calls. Through the business designer, users can turn the model into a service, enabling it to be called by other systems or applications.

[0081] Once the model is built, the service can be published. After the platform has packaged the service, users can publish it. This allows other users or systems to access and use the service over the network. The published service can be called through an API, enabling real-time data transmission and processing. This setup allows the model to be integrated into a domain-wide shared model, providing bed height monitoring services for multiple users or monitoring systems.

[0082] The following are examples of model service call instructions and requests:

[0083] post: / models / {modelCode} / versions / {modelVersion} / invocations.

[0084] Example request:

[0085] {

[0086] "outerId" : "uuiduuiduuid",

[0087] "inputData" : [ [ 1.5 ] ]

[0088] Among them, post is used to submit data to the server (API interface) to execute the corresponding model. The request includes the data to be processed (parameters input into the model), where modelCode is the unique identifier of the model, which is used to locate the specific model, and modelVersion is the version number of the model, which is used to distinguish different iterative versions of the same model; outerId is the client-defined request unique identification code, which is used to track request records, troubleshoot problems or associate business data to ensure the uniqueness of each request. InputData is the input data provided to the model in the request. Each request can include multiple data samples. Specifically in the example, the real-time collected process parameters (such as furnace top pressure, furnace bottom pressure, furnace temperature, chlorine flow rate, etc.) are passed into the model through inputData, and the model can return the calculated bed height. The specific operation details will not be repeated here.

[0089] Specifically, during the operation of the model checking service, the DCS platform monitors the service's performance, including key indicators such as response time and number of calls. If service performance degrades or anomalies are detected, users can optimize and adjust the service using tools provided by the platform to ensure service stability and reliability.

[0090] During model development and service operation, users accumulate a wealth of experience, knowledge, and best practices. The DCS platform can store this knowledge in the form of documents, templates, or components, forming an internal enterprise knowledge base. This allows users to quickly reuse existing knowledge and components when encountering similar problems or requirements, further improving development efficiency and quality.

[0091] Through the above steps, this application provides a method for improving the accuracy of bed height calculation based on machine learning, enabling the rapid construction, release, and service packaging of the model. This method not only improves the efficiency of model development, but also makes it easier for the model to be integrated and used by other systems or applications.

[0092] With the development of business and the accumulation of data, users may need to continuously optimize and iterate the model. The DCS platform provides flexible model management functions, and users can easily update model algorithms, adjust parameter configurations, or introduce new data samples for retraining. These operations can be completed through the platform's visual interface without going deep into the underlying code, which greatly reduces the threshold for model optimization. To this end, the present invention provides a method for improving bed height accuracy detection based on machine learning, which realizes the rapid construction, release, service packaging, monitoring optimization, and knowledge precipitation and reuse of the model, providing strong support for the digital transformation and intelligent manufacturing of enterprises. Through the above specific implementation methods, the present application realizes the rapid construction, deployment and calling of the bed height calculation model, effectively improves production efficiency, reduces energy consumption, and has significant economic and social benefits.

[0093] As a preferred example, this example also provides a detection APP. Specifically, after the application template is generated and packaged, a corresponding detection APP is produced and generated. The detection APP is communicated with the API interface of the packaged model, and is also communicated with the parameter detection systems of several chlorination furnaces. The detection APP can obtain the detection data of the chlorination furnace in real time and can call the application template. The latest detection value of the bed height of each chlorination furnace is obtained through the application template, and the latest detection value is displayed on the human-computer interaction interface of the detection APP, or on the display interface connected to the detection APP.

[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 principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A bed height detection method based on machine learning, characterized in that: include: Step S1: using the DCS system to collect historical data of the chlorination furnace to form a corresponding sample set; Step S2: Using multiple algorithms to perform regression simulation on the data of the sample set and establish corresponding models; Step S3: Evaluate the models obtained by using different algorithms in step S2, and use the optimal model as the output model; Step S4: Deploy the optimal model as an application template, and automatically call all data when the application template is run; Step S5: Encapsulate the deployed application template through the DCS platform, and open the encapsulated model through the API interface for external calls; Step S6: Connect the API interface to the chlorination furnace data monitoring system so that the application template can call the monitoring data of the chlorination furnace and display the latest execution results on the interactive interface.

2. The bed height detection method based on machine learning according to claim 1, characterized in that: The historical data of the chlorination furnace in step S1 at least include: chlorination furnace temperature, chlorine gas flow rate, furnace top pressure, furnace bottom pressure and bed height value.

3. The bed height detection method based on machine learning according to claim 1, characterized in that: In step S1, the historical data is preprocessed, including at least defect processing and data standardization processing.

4. The bed height detection method based on machine learning according to claim 1, characterized in that: The multiple algorithms in step S2 include one or more of nonlinear SVM regression, linear regression, random forest regression, and BP neural network regression.

5. The bed height detection method based on machine learning according to claim 4, characterized in that: The output model in step S3 is a bed height prediction model obtained by regression simulation using a random forest regression algorithm.

6. The bed height detection method based on machine learning according to claim 1, characterized in that: In step S3, the average relative error is calculated by the maximum relative error and the minimum relative error of each model. The closer the average relative error value is to 0, the healthier the model is. The model with the average relative error closest to 0 is determined as the optimal model.

7. The bed height detection method based on machine learning according to claim 1, characterized in that: In step S5, the application template is encapsulated by the service designer built into the DCS platform.

8. The bed height detection method based on machine learning according to claim 1, characterized in that: The detection method further comprises: Step S7: Set a timer task to call the model at intervals of a first preset time and update the result value.

9. The bed height detection method based on machine learning according to claim 8, characterized in that: The first preset time is 8-12 minutes.

10. The bed height detection method based on machine learning according to claim 8, characterized in that: The first preset time is 10 minutes.

Images