Intelligent detection method of air filter
The intelligent detection method uses a sensor group and neural network model to analyze air filter data, addressing inefficiencies in traditional methods by predicting air filter service life accurately and ensuring real-time air quality management.
Patent Information
- Application Number
- CN202411617709.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-11-13
- Publication Date
- 2024-12-27
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
Traditional air filter detection methods rely on manual judgment, which is subjective and inefficient, failing to guarantee the effective operation and real-time freshness of air quality.
An intelligent detection method using a sensor group, neural network model, and regression analysis to collect and analyze environmental and component data, predicting the remaining service life of air filters through a combined model trained with transfer learning.
Enables real-time detection of air filter status and accurate prediction of service life, ensuring effective air quality management by correlating airflow environments with air handling capacity.
Smart Images
Figure CN119203058A_ABST
Abstract
Description
Technical Field
[0001] The invention belongs to the technical field of filter detection, and in particular relates to an intelligent detection method for an air filter. Background Art
[0002] The intelligent detection method of air filters was born in the context of the continuous development of air quality control and purification technology. As people pay more and more attention to indoor air quality, air filters are important equipment for improving indoor air quality, and their performance detection and evaluation are particularly critical. Traditional air filter detection methods usually rely on manual judgment, and make relevant judgments based on the quality of the actual air inflow or the years of use. They are easily affected by human subjectivity, are inefficient, and cannot guarantee the effective operation of air filters and thus cannot guarantee the real-time freshness of air quality. Summary of the invention
[0003] In order to solve the above technical problems, the present invention proposes an intelligent detection method for an air filter to solve the above problems existing in the prior art.
[0004] To achieve the above object, the present invention provides an intelligent detection method for an air filter, comprising:
[0005] Collect environmental data and component data related to the air filter; perform feature recognition on the environmental data through a neural network model to obtain the environmental processing capacity data of the air filter; perform regression analysis on the component data through regression coefficients to generate component status data; generate the remaining service life of the air filter based on the component data and environmental processing capacity data.
[0006] Optionally, data is collected from the air filter through a sensor group, wherein the sensor group includes a first environmental sensor, a second environmental sensor and a component sensor, which are respectively arranged at an air inlet and an air outlet of the air filter, and the component sensor is arranged in the air filter, and the sensor group is connected to a gateway device.
[0007] Optionally, the environmental data includes air quality, temperature and humidity, and wind speed, and the component data includes pressure and displacement of the air filter.
[0008] Optionally, the neural network model adopts a combination model of a convolutional neural network model and a long short-term memory artificial neural network model, wherein the neural network model includes several convolutional pooling layers, an LSTM layer and two fully connected layers connected in sequence.
[0009] Optionally, before the feature recognition of the environmental data by the neural network model, the following steps are also included:
[0010] The existing first environmental data and second environmental parameters are used as independent variables through regression analysis, and the corresponding environmental processing capacity data are used as dependent variables to generate fitting coefficients. The initial data set is generated through the fitting coefficients, and the neural network model is initially trained through the initial data set. The existing first environmental data and second environmental parameters and the corresponding environmental processing capacity data are used as training data sets. The neural network model after initial training is trained through the training data set through the transfer learning method, and the environmental data is feature recognized through the trained neural network model.
[0011] Optionally, the process of performing regression analysis on component data using regression coefficients includes:
[0012] Obtain historical component data, perform regression analysis on the historical component data and corresponding component status data, and generate regression coefficients and offsets;
[0013] The component data is calculated through the regression coefficient and the offset to generate the corresponding component status data.
[0014] Optionally, the regression analysis of the component data by using the regression coefficients may also include:
[0015] A threshold value is judged on the component data, and an early warning is issued according to the judgment result.
[0016] Optionally, the component data and environmental processing capacity data are identified by a deep learning model to generate the remaining service life of the air filter.
[0017] Compared with the prior art, the present invention has the following advantages and technical effects:
[0018] Through the above technical solution of the present invention, by obtaining environmental parameters and component parameters, the airflow environment of the air filter input and output is determined through the environmental parameters, and at the same time, according to the correlation between the airflow environment and the air filter's air handling capacity, a combined model that can characterize the time-frequency characteristics and time series characteristics is constructed. The environmental handling capacity can be effectively predicted through the combined model. At the same time, in order to facilitate the use of the model, some data sets are constructed through relevant regression methods, and the combined model is initially trained through the data, and the combined model is trained again through the existing data through transfer learning to improve the recognition accuracy of the combined model. Subsequently, the component status is detected in real time, and the service life is predicted based on the environmental handling capacity and supplemented by the component data. Through the above technical solution, the present invention can effectively detect the environmental handling capacity of the air filter and effectively predict its service life. BRIEF DESCRIPTION OF THE DRAWINGS
[0019] The drawings constituting a part of the present application are used to provide a further understanding of the present application. The illustrative embodiments and descriptions of the present application are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:
[0020] Figure 1 The figure is a schematic diagram of a method flow of an embodiment of the present invention. DETAILED DESCRIPTION
[0021] It should be noted that, in the absence of conflict, the embodiments and features in the embodiments of the present application can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.
[0022] It should be noted that the steps shown in the flowcharts of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and that, although a logical order is shown in the flowcharts, in some cases, the steps shown or described can be executed in an order different from that shown here.
[0023] Embodiment 1
[0024] like Figure 1 As shown, this embodiment provides an intelligent detection method for an air filter, comprising:
[0025] Relevant data of the air filter is collected through the sensor group, wherein the relevant data includes environmental data and component data of the air filter. The relevant data of the air filter is transmitted to a local processing chip or a remote processing platform, and feature recognition and extraction are performed on the transmitted data. The processing power of environmental data extraction and the status of the air filter component data are jointly analyzed to detect the status of the air filter and its remaining service life.
[0026] The sensor group includes different sensors and gateway devices for sensor data transmission. Sensors are installed at various key positions of the air filter. Specifically, for example, a relevant environmental sensor is set at the air inlet of the air filter, which is defined as the first environmental sensor here to characterize the input airflow or environment of the air filter. A relevant environmental sensor is set around the air outlet of the air filter, which is defined as the second environmental sensor here to characterize the output airflow or environmental data of the air filter. The environmental sensors include air quality sensors, temperature and humidity sensors, wind speed sensors, etc., which can be added according to relevant needs. Relevant pressure sensors and displacement sensors can be set on the air filter to characterize the pressure or displacement or other physical quantity states of the air filter. It should also be noted that the above sensors can be adaptively set according to manual experience or the actual situation of the workshop, which will not be elaborated here.
[0027] For sensors in sensor groups set at different positions, a wireless transmission gateway device is used to wirelessly transmit the above sensors. The gateway device is used in the present invention mainly to solve the use of the above-mentioned large range and large number of sensors. When the relevant control terminal or local processor cannot provide enough communication interfaces for the sensors, all sensors are connected to the control terminal through the gateway device on the basis of the limited communication interface of the control terminal. The information of the above sensors is transmitted to the control terminal through the gateway device. A single gateway device can connect multiple sensors set for a single air filter or multiple air filters in one area. Through the above connection structure, the direct connection between the sensor and the control terminal is avoided, and the sensor group with multiple sensors can be transmitted data, avoiding the problem of insufficient control terminal interfaces. At the same time, the use of gateway devices for wireless connection improves the scalability of the number of sensor groups.
[0028] After the above-mentioned sensor group is deployed, the present invention performs intelligent detection on the air filter through local processing or remote terminal processing. During the intelligent detection process, the data collected by the sensor is transmitted to the local processing chip or remote terminal (computer or server) through the gateway device, and the environmental data and sensor data are subjected to relevant preprocessing such as data denoising and enhancement. The preprocessing related methods include median filtering, data interpolation and completion and other related methods.
[0029] After preprocessing the above environmental data and component data, firstly, feature recognition is performed separately for different environmental data and component data. During the recognition process, different environmental data are correlated and processed. The environmental data include environmental data collected by the first environmental sensor and the second environmental sensor. During the environmental data processing, the input airflow of the air filter is used as the input of the air filter, and the output airflow of the air filter is used as the output of the air filter. There is a correlation between the environmental processing capacity of the air filter and the input airflow and the output airflow. Based on this, the first environmental sensor and the second environmental sensor can be used as variables according to the above content, and the processing capacity of the air filter can be used as the independent variable to fit the relevant regression relationship. The numerical value of the environmental processing capacity is determined by the expert scoring method, wherein the environmental processing capacity is determined by the dust reduction capacity and the freshness capacity. Specifically, the dust reduction capacity represents the air filtration capacity. The device can reduce the dust density of the air. At the same time, for the same air, the change in the dust density of the air is quantified, such as a reduction of 0.1g / cm2 is scored as 1, a reduction of 0.2g / cm2 is scored as 2, and so on. The score is used as a quantitative value to determine the reduction capacity; the freshness capacity represents the degree of optimization of the freshness of the air before and after passing through the air filter, wherein the freshness of the air before and after the air filter is scored by relevant professionals and set to 1-10 points. The size of the freshness score represents different degrees of freshness from fresh to turbid. The scoring results before and after are used as quantitative values, and the difference between the freshness before and after is calculated. The quantitative values of the dust reduction capacity and the freshness capacity are weighted and calculated, wherein the weights of the weighted sum can be set to 0.5 and 0.5, or set according to the experience of relevant personnel to obtain the value of the environmental treatment capacity.
[0030] In the present invention, a deep learning model is used as a model for fitting regression relationships, wherein the deep learning model can select a combination model of a long short-term memory artificial neural network LSTM model and a convolutional neural network CNN model, wherein the CNN network model is used as a pre-processing of the corresponding feature data of the combined model, and LSTM is used as the temporal feature of the combined model. Specifically, the convolutional neural network model includes a number of convolutional layers and pooling layers connected in sequence, and the model parameters, such as the convolution kernel size, sliding step size, boundary filling value and pooling parameters, are initialized and set during the construction of the convolutional neural network. The CNN neural network performs relevant convolution and pooling operations on the environmental parameters to extract the frequency domain features and time domain features of the environment, and then inputs the environmental features into the LSTM layer. The LSTM network includes hyperparameters and weight parameters. The hyperparameters include the number of hidden layers, loss function, number of neurons in each layer, activation function, number of iterations, and learning rate of the LSTM network. The weight parameters are the weight coefficients inside the LSTM network. The above parameters are initialized. In the LSTM network model, according to the relevant features output by the CNN network, the LSTM fits the time series features with long-term dependencies, and inputs the time series features into two fully connected layers. The time series features are fitted through the two fully connected layers to generate the environmental processing capacity corresponding to the air filter.
[0031] After the model parameters are initialized, the above-mentioned combined model is trained. Since the data required for the above-mentioned model training is a long-term process, in order to quickly use the model, first, according to the existing environmental data corresponding to the input airflow and the output airflow, that is, the first environmental data and the second environmental data are used as independent variables, and the relevant data of the air filter processing capacity are used as dependent variables, regression fitting is performed by multivariate regression fitting. After the regression fitting is completed, the regression correlation coefficient item is included. By adjusting the environmental data values of different input airflows and output airflows, different environmental processing capacities are generated. The above data are integrated as the initial data set, and the combined model is initially trained by the data set. After the initial training is completed, the combined model after the initial training is trained by the existing above-mentioned data. After training, the combined model is used. During the use process, some environmental data corresponding to the input airflow and the output airflow are collected, and the processing capacity of the air filter is determined. After the determination, the parameters of the combined model are also trained and updated in real time in the later stage. This content can be realized through a remote control terminal. If a local recognition method is adopted, the model parameters are periodically transmitted in real time. After the transmission is completed, the local model is updated according to the transmitted model parameters.
[0032] The component data is judged by a relevant regression model to determine the corresponding component state. In the component data, the pressure data and the displacement data are detected, wherein the pressure data and the displacement data are counted in real time, wherein the pressure data and the displacement data are measured by corresponding pressure sensors and displacement sensors, wherein the corresponding pressure sensors are arranged at the connection of different components to indicate the pressure state of a certain component, and the displacement sensors are arranged on different components to indicate the displacement state of the corresponding components, and the measurements are started when the air filter is working normally, and within a certain period of time, the average values of the pressure data and the displacement data are obtained, after obtaining the above-mentioned component information, the mean value of the component data is used as the independent variable through correlation analysis, and the component state is used as the dependent variable for correlation analysis, so as to fit the correlation between the component data and the component state, wherein the correlation analysis method can select the linear regression method or the nonlinear regression method, after fitting, the regression coefficient and the offset are extracted, and the component state is determined according to the scoring of the components of the air filter by relevant professionals, and after determination, the regression coefficient and the offset corresponding to the regression analysis process are extracted, and the currently collected component information is calculated by the regression coefficient and the offset to generate the corresponding component state value. At the same time, during the statistical process of pressure and displacement data, if the pressure data drops sharply by a certain threshold value or the displacement data exceeds a certain threshold value, it is considered that the air filter may be damaged, and no relevant identification is performed. The air filter or its fixing device is damaged and the air filter cannot work normally.
[0033] The remaining service life of the air filter is obtained based on the above-mentioned component status data and environmental processing capacity, wherein a simple deep learning network model can be used to complete the task, wherein the component status data only participates in the fitting operation of the deep learning network model as an influencing item, and in the process of parameter setting of the deep learning network model, the weight parameters corresponding to the component data can be adjusted slightly. The deep learning network model can be selected as a CNN network, a fully connected neural network, which can be generated through training using an existing deep network model, and the remaining service life is predicted by inputting the environmental processing capacity and component data at the current time. When the air filter or its fixing device is damaged, the remaining service life is no longer predicted, and an alarm is promptly issued through the relevant alarm device.
[0034] In the remote terminal or local processing module, environmental data, component data, environmental processing capacity, service life, etc. are stored and displayed to guide relevant personnel in the maintenance and replacement of air filters.
[0035] Through the above technical solution of the present invention, by obtaining environmental parameters and component parameters, the airflow environment of the air filter input and output is determined through the environmental parameters, and at the same time, according to the correlation between the airflow environment and the air filter's air handling capacity, a combined model that can characterize the time-frequency characteristics and time series characteristics is constructed. The environmental handling capacity can be effectively predicted through the combined model. At the same time, in order to facilitate the use of the model, some data sets are constructed through relevant regression methods, and the combined model is initially trained through the data, and the combined model is trained again through the existing data through transfer learning to improve the recognition accuracy of the combined model. Subsequently, the component status is detected in real time, and the service life is predicted based on the environmental handling capacity and supplemented by the component data. Through the above technical solution, the present invention can effectively detect the environmental handling capacity of the air filter and effectively predict its service life.
[0036] The above are only preferred specific implementations of the present application, but the protection scope of the present application is not limited thereto. Any changes or substitutions that can be easily thought of by a person skilled in the art within the technical scope disclosed in the present application should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.
Claims
1. An intelligent detection method for an air filter, characterized in that: include: Collect environmental data and component data related to the air filter; perform feature recognition on the environmental data through a neural network model to obtain environmental processing capacity data of the air filter; The component data is subjected to regression analysis through regression coefficients to generate component status data; the remaining service life of the air filter is generated based on the component data and environmental processing capacity data.
2. The method according to claim 1, characterized in that Data is collected on the air filter through a sensor group, wherein the sensor group includes a first environmental sensor, a second environmental sensor and a component sensor, which are respectively arranged at the air inlet and the air outlet of the air filter. The component sensor is arranged in the air filter, and the sensor group is connected to a gateway device.
3. The method according to claim 1, characterized in that The environmental data includes air quality, temperature and humidity, and wind speed, and the component data includes pressure and displacement of the air filter.
4. The method according to claim 1, characterized in that: The neural network model adopts a combination model of a convolutional neural network model and a long short-term memory artificial neural network model, wherein the neural network model includes several convolutional pooling layers, LSTM layers and two fully connected layers connected in sequence.
5. The method according to claim 1, characterized in that Before the feature recognition of environmental data through the neural network model, it also includes: The existing first environmental data and second environmental parameters are used as independent variables through regression analysis, and the corresponding environmental processing capacity data are used as dependent variables to generate fitting coefficients. The initial data set is generated through the fitting coefficients, and the neural network model is initially trained through the initial data set. The existing first environmental data and second environmental parameters and the corresponding environmental processing capacity data are used as training data sets. The neural network model after initial training is trained through the training data set through the transfer learning method, and the environmental data is feature recognized through the trained neural network model.
6. The method according to claim 1, characterized in that The process of performing regression analysis on component data through regression coefficients includes: Obtain historical component data, perform regression analysis on the historical component data and corresponding component status data, and generate regression coefficients and offsets; The component data is calculated through the regression coefficient and the offset to generate the corresponding component status data.
7. The method according to claim 1, characterized in that The regression analysis of component data by regression coefficients also includes: A threshold value is judged on the component data, and an early warning is issued according to the judgment result.
8. The method according to claim 1, characterized in that: The component data and environmental processing capacity data are identified through a deep learning model to generate the remaining service life of the air filter.