Noise sensor operation monitoring system and method based on multivariate data fusion processing

Abstract

The invention discloses a noise sensor operation monitoring system and method based on multivariate data fusion processing, and relates to the technical field of data processing. The information collection module is used for acquiring historical environment data and historical noise data; and the information processing module is used for exploring the influence relationship between the environment and the noise and the corresponding information through an exploration method to obtain a target relationship and reference information. According to the method, the set compensation model is matched with the dynamic analysis method, whether the environment and the noise are changed or not is analyzed according to the real-time environment data to obtain the change data, and compensation is performed through the compensation model according to the change data to obtain the corrected noise; according to the method, the monitored noise data with deviation of the noise sensor caused by environment change can be compensated and corrected, so that the accuracy of the monitored noise data is improved, the predicted noise is obtained through matching of the set prediction matching method according to the user demand and the storage library, the user can conveniently know the required future noise, and the user experience is improved. Therefore, corresponding measures can be taken to control the noise.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and in particular to a noise sensor operation monitoring system and method based on multivariate data fusion processing. Background Art

[0002] A noise sensor is a device used to measure the level of environmental noise. It is widely used in industries such as industry, transportation, and urban management to help monitor and control noise pollution. When the noise sensor is in operation, it is easily affected by the external environment, resulting in monitoring errors. In order to reduce monitoring errors, it is often necessary to monitor and calibrate the noise sensor during operation.

[0003] The patent publication number is CN116124481A, which is a locomotive noise intelligent monitoring system, including: a data acquisition module, which is used to set a sound sensor according to the locomotive track position to collect noise signals, form a noise difference vector data set, and form a nonlinear model; a data analysis module, which is used to perform noise sample confirmation operations according to the track noise characteristics, and mark the noise sample data after confirmation; a noise monitoring module, which is used to substitute the noise sample data into the nonlinear model, identify abnormal noise through Bayesian judgment of the data distribution of track noise, and discover the abnormal contact state between the track and the vehicle.

[0004] The monitoring systems based on the above and similar principles mostly rely on noise sensors. In a dynamic environment where the environment is prone to change, the errors generated by the noise sensor monitoring data will increase, which will lead to errors in the overall noise monitoring, making it difficult to feedback accurate real-time noise conditions. Noise pollution not only has a direct impact on personal health, but also affects multiple aspects such as psychological conditions, ecological environment and social economy. In order to reduce the harm caused by noise pollution, it is necessary to take effective noise control measures. In order to optimize and reduce noise pollution, it is often necessary to impose corresponding speed limits, traffic restrictions or traffic guidance on roads. In the prior art, when performing noise monitoring, it is not convenient to predict future noise conditions based on historical noise conditions, and thus it is not convenient for management departments to take measures to reduce noise pollution while ensuring traffic to the greatest extent. Therefore, the present invention is proposed. Summary of the invention

[0005] The object of the present invention is to provide a noise sensor operation monitoring system and method based on multivariate data fusion processing to solve the problems raised in the above background technology.

[0006] To achieve the above object, the present invention provides the following technical solution: a noise sensor operation monitoring system based on multivariate data fusion processing, the system comprising: Information collection module: obtain historical environmental data and historical noise data; Information processing module: Use the exploration method to explore the influence relationship and corresponding information between the environment and noise to obtain the target relationship and reference information, and establish a compensation model based on the target relationship and reference information; Monitoring and compensation module: acquires the monitoring data and environmental data of the noise sensor in real time to obtain real-time noise data and real-time environmental data, analyzes whether the environment and noise have changed based on the real-time environmental data through a dynamic analysis method to obtain change data, extracts the real-time noise data from the change data to obtain the noise to be corrected, and compensates the noise to be corrected based on the change data through a compensation model to obtain the corrected noise; Reference establishment module: replacing the noise to be corrected in the real-time noise data with the corrected noise to obtain the corrected real-time noise, and establishing a storage library for storing the corrected real-time noise, real-time environmental data, historical environmental data and historical noise data; Matching output module: obtains predicted demand and user address information, determines the target time point based on the predicted demand, obtains predicted noise through the prediction matching method based on the target time point and the repository, and feeds back the predicted noise to the user based on the user address information; The research method includes: presetting a standard environment and quantitative noise, obtaining a reference noise by monitoring the quantitative noise in the standard environment based on a noise sensor, presetting additional parameters, adjusting the standard environment based on the additional parameters to obtain an adjusted environment, obtaining a comparative noise by monitoring the quantitative noise in the adjusted environment based on the noise sensor, integrating the reference noise, comparative noise, the standard environment and the adjusted environment to obtain reference information, and obtaining the influence relationship between the environment and the noise through a relationship acquisition method based on the reference information to obtain a target relationship.

[0007] Furthermore, the process of presetting additional parameters and adjusting the standard environment based on the added parameters to obtain an adjusted environment is: splitting the standard environment to obtain several standard items, presetting fixed parameters based on the standard items to obtain sub-parameters, integrating the sub-parameters of all standard items to obtain added parameters, presetting adjustment objects, the adjustment objects include adjusting a single standard item and adjusting multiple standard items, determining the adjustment objects based on the standard items to obtain the target objects, and performing incremental and decremental adjustments on the standard items in the target objects based on the added parameters to obtain the adjusted environment.

[0008] Furthermore, the relationship acquisition method includes: classifying the reference information into environment and noise to obtain a noise information set and an environment information set, splitting the noise information set and the environment information set to obtain a plurality of noise information and a plurality of environment information, establishing the correlation between the noise information and the environment information, extracting the environment in which a single standard item is adjusted from the environment information set to obtain a single-factor environment, extracting the environment in which multiple standard items are adjusted from the environment information set to obtain a multi-factor environment, extracting the noise information of the single-factor environment and the multi-factor environment based on the correlation to obtain single-factor noise and multi-factor noise, sorting the single-factor environment and the multi-factor environment from small to large based on the parameters of the same standard item to obtain single-factor sorting information and multi-factor sorting information, and combining the single-factor sorting information and the multi-factor sorting information based on the correlation. The factor sorting information corresponds to sorting single-factor noise and multi-factor noise to obtain single-factor sorting noise and multi-factor sorting noise. Based on the difference of adjacent noises in the single-factor sorting noise and multi-factor sorting noise, the single-factor difference set and the multi-factor difference set are obtained. The single-factor noise range and the multi-factor noise range with the same difference in the single-factor difference set and the multi-factor difference set are extracted to obtain the single-factor range and the multi-factor range. Based on the single-factor range and the multi-factor range, the corresponding single-factor environment and the multi-factor environment are extracted in conjunction with the correlation to obtain the linear relationship range. The single-factor range and the multi-factor range in the single-factor noise and the multi-factor noise are eliminated and the corresponding single-factor environment and the multi-factor environment are extracted in conjunction with the correlation to obtain the non-linear relationship range. The linear relationship range and the non-linear relationship range are integrated to obtain the target relationship.

[0009] Furthermore, the method of establishing the compensation model includes: Data acquisition and processing: Based on the reference information and target relationship, the training data is derived through the data derivation method to obtain the derived data, the reference information and the derived data are integrated to obtain the preliminary data, the noise difference generated by the adjacent environmental data in the preliminary data is obtained to obtain the target difference, and the target difference is marked in the preliminary data to obtain the training data; Model selection and training: Select a deep learning model as the model matrix, import the training data into the model matrix for training to obtain the initial model; Model adjustment and output: preset verification data and verification results, import the verification data into the initial model to obtain result information, compare the verification results and result information to obtain comparison information, and adjust and optimize the initial model based on the comparison information to obtain a compensation model.

[0010] Furthermore, the data derivation method includes: the target relationship is a linear relationship range and a nonlinear relationship range, an added value is preset, the added value is increased or decreased for the environment in the linear relationship range to obtain a linear derived environment, based on the relationship between the added value and the added parameter, the noise information of the environment is correspondingly reduced or increased to obtain a linear derived noise, the added value of the environment in the nonlinear relationship range is increased or decreased to obtain a nonlinear derived environment, nonlinear derived noise is obtained based on the nonlinear derived environment and quantitative noise in conjunction with a noise sensor, and the linear derived environment, linear derived noise, nonlinear derived environment and nonlinear derived noise are integrated to obtain derived data.

[0011] Furthermore, the dynamic analysis method includes: presetting an interval time, obtaining environmental data of the interval time before the real-time time point to obtain past environmental data, obtaining noise information of the past environmental data and the real-time environmental data to obtain a first noise and a second noise, and when the real-time environmental data changes and the second noise does not change and when the real-time environmental data changes and the second noise changes, integrating the past environmental data, the real-time environmental data, the first noise and the second noise to obtain change data.

[0012] Furthermore, the prediction and matching method includes: extracting time features and holiday features of historical noise data and real-time noise data in the repository to obtain a reference feature set, the reference feature set includes a number of reference features corresponding to the historical noise data and the real-time noise data, obtaining the time features and holidays of the target time point to obtain the target feature, presetting a similarity value, traversing the reference feature set based on the target feature to select the reference feature whose similarity with the target feature exceeds the similarity value to obtain the selected feature, and selecting the noise data corresponding to the target feature in the repository based on the selected feature to obtain the predicted noise.

[0013] The noise sensor operation monitoring method based on multivariate data fusion processing uses the above-mentioned noise sensor operation monitoring system based on multivariate data fusion processing.

[0014] Compared with the prior art, the present invention has the following beneficial effects: The noise sensor operation monitoring system and method based on multivariate data fusion processing uses a set compensation model in conjunction with a dynamic analysis method to analyze whether the environment and noise have changed based on real-time environmental data to obtain change data, and uses a compensation model to compensate for the noise to be corrected based on the change data to obtain corrected noise. This can compensate and correct the monitoring noise data of the noise sensor that has deviations due to changes in the environment to improve the accuracy of the monitoring noise data. The set prediction matching method is used to match and output according to user needs in conjunction with a storage library to obtain predicted noise, which can facilitate users to understand the required future noise so as to make corresponding measures to control the noise.

[0015] At the same time, the reference information is derived from the target relationship through the set data derivation method to enrich the number of data used to train the model and enrich the diversity of the data in order to improve the accuracy of the model. The multiple relationship between the added value and the added parameter is recorded. The linear derived environment can be obtained by increasing or decreasing the added value of the environment in the linear relationship range. The linear derived noise corresponding to the linear derived environment can be calculated through the multiple relationship. The nonlinear derived environment can be obtained by increasing and decreasing the added value of the environment in the nonlinear relationship range. A nonlinear derived environment is created and quantitative noise is added to cooperate with a noise sensor to obtain nonlinear derived noise. Through the prediction matching method, the features in the historical noise data can be extracted and combined with user needs to obtain the features of the future time points for similarity matching to obtain the predicted noise of the future time points, so that users can know the possible predicted noise at future time points in advance, so as to facilitate application in different fields.

[0016] At the same time, through the set dynamic analysis method, it is determined whether the real-time monitored noise needs to be compensated to improve the smoothness of the overall system operation and reduce the system load, which is convenient for use. When the real-time environmental data changes and the second noise does not change, and when the real-time environmental data changes and the second noise changes, it proves that the real-time noise data monitored by the noise sensor needs to be compensated. At the same time, a threshold for the change of real-time environmental data can be set, that is, when the threshold is exceeded, the data monitored by the noise sensor needs to be compensated, and when it does not exceed the threshold, no compensation is required. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 It is a schematic diagram of the main structure of the present invention; Figure 2 A schematic diagram of a module structure is established for reference of the present invention; Figure 3 It is a schematic diagram of the structure of the matching output module of the present invention; Figure 4 A schematic diagram of the structure of the compensation model establishment process of the present invention; Figure 5 It is a structural schematic diagram of the relationship acquisition method of the present invention.

[0018] Description of the drawings: Figure 5 (a) is a schematic diagram of the linear relationship range and nonlinear relationship range structure of single-factor sorting noise. Figure 5 (b) is a schematic diagram of the linear relationship range and nonlinear relationship range structure of multi-factor sorting noise. DETAILED DESCRIPTION

[0019] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0020] Multivariate data fusion processing refers to the integration, analysis and processing of data from different sources, different formats and different characteristics to extract more comprehensive and accurate information. Multivariate data fusion processing refers to the integration and processing of data from different sources (such as noise data and environmental data) to improve the accuracy, reliability and intelligence level of the monitoring system. By correcting the data collected by the noise sensor through multivariate data fusion processing, environmental interference can be eliminated and data accuracy can be improved. The corrected noise data is more accurate and reliable, and can support the intelligent operation and decision-making of the noise monitoring system.

[0021] like Figure 1-Figure 5 As shown, the present invention provides a technical solution: a noise sensor operation monitoring system based on multivariate data fusion processing, the system comprising: Information collection module: obtain historical environmental data and historical noise data; Example

[0022] In the process of connecting the noise sensor to the operation monitoring system, it is first necessary to determine the noise sensor model and obtain the historical noise data of the noise sensor. For example, when it is necessary to connect the noise sensor of model Microphone (1 / 2" Prepolarized, 40kHz)-Brüel&Kjær4135 to the operation monitoring system, the historical monitoring noise data of the noise sensor of this model and the environmental data of the noise sensor are obtained. The specific environmental data can be obtained through the corresponding sensor. When the noise sensor of this model is adjusted, the historical noise data of the adjusted noise sensor needs to be obtained accordingly.

[0023] Information processing module: Use the exploration method to explore the influence relationship and corresponding information between the environment and noise to obtain the target relationship and reference information, and establish a compensation model based on the target relationship and reference information; Real-time Example 2 The influence relationship between environment and noise is determined based on the collected historical environmental data and historical noise data. Specifically, the time relationship between the historical environmental data and the historical noise data needs to be established first to ensure that the environmental data corresponds to the noise data. When the collected historical environmental data is the first environmental data: temperature: 20°C, humidity: 50%, air pressure: 1010hPa, wind speed: 5km / h; the second environmental data: temperature: 25°C, humidity: 55%, air pressure: 1030hPa, wind speed: 7km / h; the third environmental data: temperature: 22°C, humidity: 50%, air pressure: 101 0hPa, wind speed: 8km / h; the collected historical noise data are: the first noise data is 70dB; the second noise data is 60dB; the third noise data is 65dB; that is, it is necessary to pair the noise and the environment according to the time relationship, and match the generation time of the first environmental data, the second environmental data and the third environmental data with the generation time of the first noise data, the second noise data and the third noise data to obtain the corresponding environmental data and noise data, and then extract multiple corresponding environmental data and noise data through the exploration method to obtain the impact relationship and corresponding information between noise and environment.

[0024] Monitoring and compensation module: acquires the monitoring data and environmental data of the noise sensor in real time to obtain real-time noise data and real-time environmental data, analyzes whether the environment and noise have changed based on the real-time environmental data through a dynamic analysis method to obtain change data, extracts the real-time noise data from the change data to obtain the noise to be corrected, and compensates the noise to be corrected based on the change data through a compensation model to obtain the corrected noise; Real-time Example 3 When the noise sensor is running in real time, by installing temperature and humidity sensors, air pressure sensors and wind speed sensors around the working environment of the noise sensor, the temperature, humidity, air pressure and wind speed data can be obtained, that is, the real-time environmental data is obtained, and then the real-time noise data can be obtained through the noise value monitored by the noise sensor. In the process of information generation and output, ensure that the generation and output of real-time environmental data and real-time noise data are consistent, that is, when the noise sensor generates noise at 11:50, the corresponding monitoring environment sensor generates environmental data at 11:50, that is, temperature, humidity, air pressure and wind speed data. According to the collected Real-time environmental data and real-time noise data are analyzed through dynamic analysis methods to determine whether the environment and noise have changed to obtain change data. Specifically, it is determined whether the noise changes when the environment changes, and whether the noise changes when the environment does not change. When the environment changes, the occurrence or non-change of noise may cause deviations in the monitored noise values. The environmental change threshold can be set according to the applicable environment of the noise sensor. When the temperature is between -10°C and 35°C and the humidity is between 10% and 65%, it is determined that the noise sensor will not have deviations, that is, when the environmental change is within the environmental change threshold, there is no need to compensate for the noise.

[0025] Reference establishment module: replacing the noise to be corrected in the real-time noise data with the corrected noise to obtain the corrected real-time noise, and establishing a storage library for storing the corrected real-time noise, real-time environmental data, historical environmental data and historical noise data; Matching output module: obtains predicted demand and user address information, determines the target time point based on the predicted demand, obtains predicted noise through the prediction matching method based on the target time point and the repository, and feeds back the predicted noise to the user based on the user address information; Example

[0026] First, the user's predicted needs and user address information are obtained from the user himself. Specifically, the predicted needs can be obtained through the user's own input or sending, and the user address information, that is, the phone number or email address, is obtained by recording the user's input or sending method. When the user has a predicted need, the predicted noise that meets the user's predicted needs is predicted according to the predicted need through the prediction matching method, that is, the historical noise data corresponding to the predicted date is matched according to the previous noise data and date characteristics to obtain the predicted noise. When there are multiple predicted noises, the noise with the highest similarity during matching is selected to obtain the predicted noise, and the predicted noise is sent to the user according to the user's address information.

[0027] It should be noted that in the information collection module, historical environmental data and historical noise data can be obtained from the working log of the noise sensor or a noise sensor of the same model. The noise sensor refers to the noise sensor that needs to apply this monitoring system. The historical environmental data is the environmental data of the working environment of the noise sensor, including temperature, humidity, air pressure, etc. The historical noise data is a collection of noise values ​​monitored by the noise sensor in the past. In the information processing module, the influence relationship between the environment and noise and the corresponding information are explored through the set exploration method to obtain the target relationship and reference information. The compensation model is established through the target relationship and the reference information. The compensation model is used to compensate for the noise data monitored by the noise sensor to improve the accuracy of noise monitoring by the noise sensor. In the monitoring compensation module, real-time noise data can be directly obtained through the noise sensor. Real-time environmental data can be obtained by installing corresponding environmental sensors around the noise sensor, such as temperature and humidity sensors and air pressure sensors, etc., through the set dynamic The analysis method analyzes whether the environment and noise have changed according to the real-time environmental data to obtain the change data, and compensates the noise to be corrected through the compensation model according to the change data to obtain the corrected noise. The corrected noise is a combination of the compensation value obtained from the compensation model and the real-time noise data. Referring to the establishment module, the noise to be corrected in the real-time noise data is replaced by the corrected noise to obtain the corrected real-time noise, and a repository is established to store the real-time and historical environmental data and noise data, so as to facilitate the subsequent prediction of future noise. In the matching output module, the prediction demand is obtained through the user, specifically the user's predicted time demand, the user address information is the address information for sending information to the user, which may specifically include an IP address and an email address, etc. The target time point is the user's predicted demand. The predicted noise is obtained according to the target time point in conjunction with the repository through the set prediction matching method, and then the predicted noise is fed back to the user according to the user address information, so that the user can understand the required future noise, so as to make corresponding measures to control the noise.

[0028] like Figure 1 As shown, the research method includes: presetting a standard environment and quantitative noise, obtaining a reference noise by monitoring the quantitative noise in the standard environment based on a noise sensor, presetting additional parameters, adjusting the standard environment based on the additional parameters to obtain an adjusted environment, obtaining a comparative noise by monitoring the quantitative noise in the adjusted environment based on the noise sensor, integrating the reference noise, comparative noise, standard environment and adjusted environment to obtain reference information, and obtaining the influence relationship between the environment and noise through a relationship acquisition method based on the reference information to obtain a target relationship.

[0029] It should be noted that the standard environment and quantitative noise are formulated according to actual usage conditions. The reference noise is obtained by monitoring the quantitative noise in the set standard environment through the noise sensor. The added parameters are preset according to actual usage conditions. The specific added parameters are consistent with the individual items in the standard environment. The standard environment is adjusted by adding parameters to obtain the adjusted environment. The comparative noise is obtained by monitoring the quantitative noise in the set adjusted environment through the noise sensor. The influence relationship between the environment and noise is obtained according to the reference information through the set relationship acquisition method to obtain the target relationship.

[0030] In the specific implementation process, it is first necessary to set the standard environment, quantitative noise and add parameters. For example, the standard environment is temperature: 20°C, humidity: 50%, air pressure: 1010hPa, and wind speed: 5km / h; the quantitative noise is 65dB, and the added parameters are temperature 1°C, humidity 1%, air pressure 10hPa, and wind speed 1km / h; the regulated environment can be obtained by adjusting the standard environment by adding parameters. There are multiple regulated environments. For example, the first regulated environment is temperature: 21°C, humidity: 51%, air pressure: 1020hPa, and wind speed: 6km / h; the second regulated environment is temperature: 22°C, humidity: 52%, air pressure: 1030hPa, and wind speed: 7km / h; the nth regulated environment is temperature: 20+(n*1)°C, humidity: 50+(n*1)%, air pressure: 1010+( n*10) hPa, wind speed: 5+ (n*1) km / h; the above-mentioned adjustment environment is in an increasing state, and it can also be in a decreasing state. For example, the first adjustment environment is temperature: 19°C, humidity: 49%, air pressure: 1000hPa, wind speed: 4km / h; the second adjustment environment is temperature: 18°C, humidity: 48%, air pressure: 990hPa, wind speed: 3km / h; the nth adjustment environment is temperature: 20- (n*1)°C, humidity: 50- (n*1)%, air pressure: 1010- (n*10) hPa, wind speed: 5- (n*1) km / h; at the same time, the number of specific adjustment items can be single or multiple. The above-mentioned is to adjust multiple items, and the adjustment of a single item is, for example, temperature: 21°C, humidity: 50%, air pressure: 1010hPa, wind speed: 5km / h.

[0031] The process of presetting additional parameters and adjusting the standard environment based on the added parameters to obtain an adjusted environment is as follows: splitting the standard environment to obtain several standard projects, presetting fixed parameters based on the standard projects to obtain sub-parameters, integrating the sub-parameters of all standard projects to obtain added parameters, presetting adjustment objects, the adjustment objects include adjusting a single standard project and adjusting multiple standard projects, determining the adjustment objects based on the standard projects to obtain the target objects, and performing incremental and decremental adjustments on the standard projects in the target objects based on the added parameters to obtain the adjusted environment.

[0032] It should be noted that the added parameters are formulated in accordance with the units of the standard items, that is, sub-parameters are obtained according to the preset fixed parameters of the standard items, and all preset sub-parameters are integrated to obtain the added parameters. The added parameters include the adjustment parameters of all standard items in the standard environment. The adjustment object is selected according to the actual usage. The selection method can be determined by the user, including single selection or multiple selections. The upper limit of multiple selections is the number of standard items in the standard environment. Based on the added parameters, the standard items in the target object are adjusted in sequence by increments and decrements to obtain the adjustment environment.

[0033] During the specific implementation process, when the standard items in the standard environment include temperature, humidity, air pressure and wind speed, there are sub-parameters corresponding to the temperature, humidity, air pressure and wind speed in the added parameters. At this time, the adjustment object can be single or multiple. When single, it can be one of the temperature, humidity, air pressure and wind speed. When multiple, it can be two, three or all of the humidity, humidity, air pressure and wind speed. The standard items in the adjustment object are adjusted incrementally and decrementally by adjusting the object in conjunction with the added parameters. For example, the first adjustment is to increase the added parameter by 1 unit, the second adjustment is to continue to increase the added parameter by 1 unit on the basis of the previous adjustment, and so on. The decrement is to decrease the added parameter by 1 unit in sequence.

[0034] like Figure 5As shown, the relationship acquisition method includes: classifying the reference information into environment and noise to obtain a noise information set and an environment information set, splitting the noise information set and the environment information set to obtain a plurality of noise information and a plurality of environment information, establishing the correlation between the noise information and the environment information, extracting the environment in which a single standard item is adjusted in the environment information set to obtain a single-factor environment, extracting the environment in which multiple standard items are adjusted in the environment information set to obtain a multi-factor environment, extracting the noise information of the single-factor environment and the multi-factor environment based on the correlation to obtain single-factor noise and multi-factor noise, sorting the single-factor environment and the multi-factor environment from small to large based on the parameters of the same standard item to obtain single-factor sorting information and multi-factor sorting information, and matching the single-factor sorting information and the multi-factor sorting information based on the correlation. The sorting information corresponds to sorting single-factor noise and multi-factor noise to obtain single-factor sorting noise and multi-factor sorting noise; based on the difference of adjacent noises in the single-factor sorting noise and multi-factor sorting noise, the single-factor difference set and the multi-factor difference set are obtained; the single-factor noise range and the multi-factor noise range with the same difference in the single-factor difference set and the multi-factor difference set are extracted to obtain the single-factor range and the multi-factor range; based on the single-factor range and the multi-factor range, the corresponding single-factor environment and the multi-factor environment are extracted in conjunction with the correlation to obtain the linear relationship range; the single-factor range and the multi-factor range in the single-factor noise and the multi-factor noise are eliminated and the corresponding single-factor environment and the multi-factor environment are extracted in conjunction with the correlation to obtain the non-linear relationship range; the linear relationship range and the non-linear relationship range are integrated to obtain the target relationship.

[0035] It should be noted that the process of classifying the reference information into environment and noise to obtain the noise information set and the environment information set is to extract the noise in the reference information to obtain the noise information set, and to extract the environment in the reference information to obtain the environment information set. By establishing the correlation between the noise information and the environment information, it is convenient to find other information according to one of them in the future. The single-factor environment is an environment in which a single standard item is adjusted in the environment information set, and the multi-factor environment is an environment in which multiple standard items are adjusted in the environment information. Based on the correlation, the corresponding noise can be extracted according to the single-factor environment and the multi-factor environment to obtain single-factor noise and multi-factor noise, and then the difference between adjacent single-factor noise and multi-factor noise is obtained to determine whether it changes linearly to obtain the linear relationship range and the nonlinear relationship range. By obtaining the linear relationship range and the nonlinear relationship range, it is convenient to deeply obtain the relationship between the environment and the noise, so as to generate more data for model training according to the linear relationship range and the nonlinear relationship range in the future, so as to improve the richness of the data and facilitate use, such as Figure 5 As shown in (a) and (b), difference 1, difference 2 and difference 3 represent different differences, that is, difference 1 is only the same as difference 1, and difference 1 is different from difference 2 and difference 3.

[0036] In the specific implementation process: when it is necessary to obtain the specific relationship between the environment and noise, and when the noise monitoring value is linearly related to a single environmental parameter, a model is established: ,in is the noise monitoring value, is the slope (influence coefficient), is the change of environmental parameters, As the benchmark noise value, in the specific use process, the least squares method is used to fit the data and solve the model parameters. The noise monitoring value is linearly related to multiple environmental parameters, and the model is established: ,in is the variation of different environmental parameters, is the influence coefficient of each parameter, The noise level is the benchmark value. In actual use, the least squares method is used to fit the data. When it is necessary to obtain the specific relationship between the environment and the noise, and the noise monitoring value is nonlinearly related to the environmental parameters, a suitable nonlinear regression model (such as polynomial regression, logarithmic regression, exponential regression, etc.) is selected for data fitting.

[0037] like Figure 4 As shown, the method of establishing a compensation model includes: Data acquisition and processing: Based on the reference information and target relationship, the training data is derived through the data derivation method to obtain the derived data, the reference information and the derived data are integrated to obtain the preliminary data, the noise difference generated by the adjacent environmental data in the preliminary data is obtained to obtain the target difference, and the target difference is marked in the preliminary data to obtain the training data; Model selection and training: Select a deep learning model as the model matrix, import the training data into the model matrix for training to obtain the initial model; Model adjustment and output: preset verification data and verification results, import the verification data into the initial model to obtain result information, compare the verification results and result information to obtain comparison information, and adjust and optimize the initial model based on the comparison information to obtain a compensation model.

[0038] It should be noted that in the data acquisition and processing stage, the reference information is derived from the target relationship through the set data derivation method to enrich the number of data used to train the model and enrich the diversity of the data, so as to improve the accuracy of the model. By marking the target difference, the model can judge whether it is a linear or nonlinear relationship. At the same time, the linear relationship range and the nonlinear relationship range in the target relationship can be imported into the model for the model to learn. In the model selection and training stage, the deep learning model can select convolutional neural network as the model matrix, import the training data and the target relationship into the model matrix for training to obtain the initial model. In the model adjustment and output stage, verification data and verification results are preset. The verification data is the preset environmental information, and the verification result is the compensation value of the noise sensor under the environmental information. The environmental information is imported into the initial model to obtain the result information, and the result information is compared with the verification result to obtain the comparison information. The initial model is adjusted and optimized according to the comparison information to obtain the compensation model.

[0039] In the specific implementation process, a deep learning model is selected as the model matrix. For example, a feedforward neural network can be selected as the model matrix to create a compensation model. First, the acquired training data is processed to remove incomplete, repeated or noisy data to ensure the quality of the data. The data is converted to a unified range (such as normalization and standardization) to ensure that different features will not affect the model training due to different dimensions. The processed training data is imported into the model matrix model for training to obtain the initial model. The training data is input into the model in batches for forward propagation. By calculating the gradient (partial derivative of the loss with respect to the model parameters), the back propagation algorithm is executed to update the model parameters. In each training round, the model will traverse all the training data and update the parameters. After the training is completed, the verification data is used to evaluate the generalization ability of the model, the final accuracy is calculated, and the model is adjusted and optimized according to the accuracy or other indicators to obtain a compensation model.

[0040] The data derivation method includes: the target relationship is a linear relationship range and a nonlinear relationship range, an added value is preset, the added value is increased or decreased for the environment in the linear relationship range to obtain a linear derived environment, based on the relationship between the added value and the added parameter, the noise information of the environment is correspondingly reduced or increased to obtain a linear derived noise, the added value is increased or decreased for the environment in the nonlinear relationship range to obtain a nonlinear derived environment, nonlinear derived noise is obtained based on the nonlinear derived environment and quantitative noise in combination with a noise sensor, and the linear derived environment, linear derived noise, nonlinear derived environment and nonlinear derived noise are integrated to obtain derived data.

[0041] It should be noted that the added value is formulated according to actual usage. The definition of the added value is consistent with that of the added parameter, but the added value is smaller than the added parameter. The multiple relationship between the added value and the added parameter is recorded. The linear derivative environment can be obtained by increasing or decreasing the added value for the environment in the linear relationship range. The linear derivative noise corresponding to the linear derivative environment can be calculated through the multiple relationship. The nonlinear derivative environment can be obtained by increasing and decreasing the added value for the environment in the nonlinear relationship range. A nonlinear derivative environment is created and quantitative noise is added to cooperate with a noise sensor to obtain nonlinear derivative noise. At the same time, the nonlinear derivative environment can also be further segmented and analyzed to extract the segmented linear derivative environment in the nonlinear derivative environment. Whether to continue the analysis is determined based on actual usage.

[0042] During the specific implementation process, an added value is preset. For example, when the added parameters are temperature 1°C, humidity 1%, air pressure 10hPa, and wind speed 1km / h, the added value can be 0.5 times the added parameter, that is, the added value is temperature 0.5°C, humidity 0.5%, air pressure 5hPa, and wind speed 0.5km / h; at the same time, the added value can also be temperature 0.5°C, humidity 1%, air pressure 10hPa, and wind speed 1km / h, which corresponds to a single adjustment or multiple adjustments in a linear derivative environment.

[0043] like Figure 1 As shown, the dynamic analysis method includes: presetting an interval time, obtaining environmental data of the interval time before the real-time time point to obtain past environmental data, obtaining noise information of the past environmental data and the real-time environmental data to obtain a first noise and a second noise, when the real-time environmental data changes and the second noise does not change and when the real-time environmental data changes and the second noise changes, integrating the past environmental data, the real-time environmental data, the first noise and the second noise to obtain change data.

[0044] It should be noted that the interval time is determined according to actual usage requirements. Through the set dynamic analysis method, it is determined whether the real-time monitored noise needs to be compensated to improve the smoothness of the overall system operation and reduce the system load, which is convenient for use. When the real-time environmental data changes and the second noise does not change, and when the real-time environmental data changes and the second noise changes, it proves that the real-time noise data monitored by the noise sensor needs to be compensated. At the same time, the threshold for the change of real-time environmental data can be set, that is, when the threshold is exceeded, the data monitored by the noise sensor needs to be compensated, and when it does not exceed the threshold, no compensation is required.

[0045] During the specific implementation process, the threshold for environmental data changes is set. For example, the threshold for environmental data changes is: temperature 15°C-25°C, humidity: 30%-70%, air pressure 950hPa-1050hPa, wind speed: 2km / h-10km / h; the preset interval time is 10 minutes, and every 10 minutes, the environmental data 10 minutes ago and the real-time environmental data will be extracted for judgment. When the real-time environmental data is in this range, no compensation will be performed.

[0046] like Figure 3 As shown, the prediction and matching method includes: extracting the time features and holiday features of historical noise data and real-time noise data in the repository to obtain a reference feature set, the reference feature set includes a number of reference features corresponding to the historical noise data and the real-time noise data, obtaining the time features and holidays of the target time point to obtain the target feature, presetting a similarity value, traversing the reference feature set based on the target feature to select the reference feature whose similarity with the target feature exceeds the similarity value to obtain the selected feature, and selecting the noise data corresponding to the target feature in the repository based on the selected feature to obtain the predicted noise.

[0047] It should be noted that the reference feature set includes the time features and holiday features when a single historical noise data and real-time noise data occur. For example, the time feature is 8 o'clock in the morning, and the holiday feature is a weekday. The similarity value is determined according to the actual usage. The larger the similarity value, the more accurate the selected prediction noise is. Through the prediction matching method, the features in the historical noise data can be extracted, and the features of the future time points can be obtained according to the user's needs. Similarity matching is performed to obtain the predicted noise of the future time points, so that users can know in advance the possible predicted noise at future time points, so as to facilitate application in different fields.

[0048] In the specific implementation process, the similarity value can be preset to 80%, and the prediction noise is selected on a preferential basis. The time characteristics and holiday characteristics can be more precise. For example, the time characteristic is 8:10 in the morning, and the holiday characteristic is Wednesday, which is a working day, etc., to improve the prediction noise accuracy of the prediction matching method.

[0049] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is limited by the attached embodiments and their equivalents.

Claims

1. A noise sensor operation monitoring system based on multivariate data fusion processing, characterized in that: The system comprises: Information collection module: obtain historical environmental data and historical noise data; Information processing module: Use the exploration method to explore the influence relationship and corresponding information between the environment and noise to obtain the target relationship and reference information, and establish a compensation model based on the target relationship and reference information; Monitoring and compensation module: acquires the monitoring data and environmental data of the noise sensor in real time to obtain real-time noise data and real-time environmental data, analyzes whether the environment and noise have changed based on the real-time environmental data through a dynamic analysis method to obtain change data, extracts the real-time noise data from the change data to obtain the noise to be corrected, and compensates the noise to be corrected based on the change data through a compensation model to obtain the corrected noise; Reference establishment module: replacing the noise to be corrected in the real-time noise data with the corrected noise to obtain the corrected real-time noise, and establishing a repository for storing the corrected real-time noise, real-time environmental data, historical environmental data and historical noise data; Matching output module: obtains predicted demand and user address information, determines the target time point based on the predicted demand, obtains predicted noise through the prediction matching method based on the target time point and the repository, and feeds back the predicted noise to the user based on the user address information; The research method includes: presetting a standard environment and quantitative noise, obtaining a reference noise by monitoring the quantitative noise in the standard environment based on a noise sensor, presetting additional parameters, adjusting the standard environment based on the additional parameters to obtain an adjusted environment, obtaining a comparative noise by monitoring the quantitative noise in the adjusted environment based on the noise sensor, integrating the reference noise, comparative noise, the standard environment and the adjusted environment to obtain reference information, and obtaining the influence relationship between the environment and the noise through a relationship acquisition method based on the reference information to obtain a target relationship.

2. The noise sensor operation monitoring system based on multivariate data fusion processing according to claim 1 is characterized in that: The process of presetting additional parameters and adjusting the standard environment based on the added parameters to obtain an adjusted environment is as follows: splitting the standard environment to obtain several standard projects, presetting fixed parameters based on the standard projects to obtain sub-parameters, integrating the sub-parameters of all standard projects to obtain added parameters, presetting adjustment objects, the adjustment objects include adjusting a single standard project and adjusting multiple standard projects, determining the adjustment objects based on the standard projects to obtain the target objects, and performing incremental and decremental adjustments on the standard projects in the target objects based on the added parameters to obtain the adjusted environment.

3. The noise sensor operation monitoring system based on multivariate data fusion processing according to claim 1 is characterized in that: The relationship acquisition method comprises: classifying the reference information into environment and noise to obtain a noise information set and an environment information set, splitting the noise information set and the environment information set to obtain a plurality of noise information and a plurality of environment information, establishing the correlation between the noise information and the environment information, extracting the environment in which a single standard item is adjusted from the environment information set to obtain a single-factor environment, extracting the environment in which multiple standard items are adjusted from the environment information set to obtain a multi-factor environment, extracting the noise information of the single-factor environment and the multi-factor environment based on the correlation to obtain single-factor noise and multi-factor noise, sorting the single-factor environment and the multi-factor environment from small to large based on the parameters of the same standard item to obtain single-factor sorting information and multi-factor sorting information, and matching the single-factor sorting information and the multi-factor sorting information based on the correlation. The single-factor noise and multi-factor noise are sorted corresponding to the sequence information to obtain the single-factor sorting noise and multi-factor sorting noise. The single-factor difference set and the multi-factor difference set are obtained based on the difference of adjacent noises in the single-factor sorting noise and the multi-factor sorting noise. The single-factor noise range and the multi-factor noise range with the same difference in the single-factor difference set and the multi-factor difference set are extracted to obtain the single-factor range and the multi-factor range. Based on the single-factor range and the multi-factor range, the corresponding single-factor environment and the multi-factor environment are extracted in conjunction with the correlation to obtain the linear relationship range. The single-factor range and the multi-factor range in the single-factor noise and the multi-factor noise are eliminated and the corresponding single-factor environment and the multi-factor environment are extracted in conjunction with the correlation to obtain the non-linear relationship range. The linear relationship range and the non-linear relationship range are integrated to obtain the target relationship.

4. The noise sensor operation monitoring system based on multivariate data fusion processing according to claim 1 is characterized in that: Methods for building compensation models include: Data acquisition and processing: Based on the reference information and target relationship, the training data is derived through the data derivation method to obtain the derived data, the reference information and the derived data are integrated to obtain the preliminary data, the noise difference generated by the adjacent environmental data in the preliminary data is obtained to obtain the target difference, and the target difference is marked in the preliminary data to obtain the training data; Model selection and training: Select a deep learning model as the model matrix, import the training data into the model matrix for training to obtain the initial model; Model adjustment and output: preset verification data and verification results, import the verification data into the initial model to obtain result information, compare the verification results and result information to obtain comparison information, and adjust and optimize the initial model based on the comparison information to obtain a compensation model.

5. The noise sensor operation monitoring system based on multivariate data fusion processing according to claim 4 is characterized in that: The data derivation method includes: the target relationship is a linear relationship range and a nonlinear relationship range, an added value is preset, the added value is increased or decreased for the environment in the linear relationship range to obtain a linear derived environment, based on the relationship between the added value and the added parameter, the noise information of the environment is correspondingly reduced or increased to obtain a linear derived noise, the added value is increased or decreased for the environment in the nonlinear relationship range to obtain a nonlinear derived environment, nonlinear derived noise is obtained based on the nonlinear derived environment and quantitative noise in combination with a noise sensor, and the linear derived environment, linear derived noise, nonlinear derived environment and nonlinear derived noise are integrated to obtain derived data.

6. The noise sensor operation monitoring system based on multivariate data fusion processing according to claim 1 is characterized in that: The dynamic analysis method includes: presetting an interval time, obtaining environmental data of the interval time before the real-time time point to obtain past environmental data, obtaining noise information of the past environmental data and the real-time environmental data to obtain a first noise and a second noise, and when the real-time environmental data changes and the second noise does not change and when the real-time environmental data changes and the second noise changes, integrating the past environmental data, the real-time environmental data, the first noise and the second noise to obtain change data.

7. The noise sensor operation monitoring system based on multivariate data fusion processing according to claim 1 is characterized in that: The prediction and matching method comprises: extracting time features and holiday features of historical noise data and real-time noise data in a storage repository to obtain a reference feature set, the reference feature set comprising a plurality of reference features corresponding to the historical noise data and the real-time noise data, obtaining time features and holidays at a target time point to obtain a target feature, presetting a similarity value, traversing the reference feature set based on the target feature to select a reference feature whose similarity with the target feature exceeds the similarity value to obtain a selected feature, and selecting noise data corresponding to the target feature in the storage repository based on the selected feature to obtain predicted noise.

8. A noise sensor operation monitoring method based on multivariate data fusion processing, characterized in that: A noise sensor operation monitoring system based on multivariate data fusion processing as described in any one of claims 1 to 7 is used.

Images