Temperature sensing data compensation method and system based on machine learning
By establishing a temperature sensing deviation correlation model and an adaptive learning model, a temperature compensation parameter set is generated, which solves the problem that traditional methods cannot adapt to complex environments, achieves high-precision compensation of temperature data, and improves the accuracy and stability of the system.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-12-26
- Publication Date
- 2026-04-03
AI Technical Summary
Traditional temperature sensor compensation methods cannot adapt to complex and ever-changing environments, resulting in the inability to effectively eliminate temperature sensing data deviations, which affects the accuracy and reliability of subsequent decision-making and control.
By acquiring raw data from temperature sensors and environmental impact characteristic data, a temperature sensing deviation correlation model is established. An adaptive learning model is used to generate a temperature compensation parameter set, and the model parameters are adjusted to optimize the compensation process, thereby achieving adaptive temperature data compensation.
It improves the accuracy and reliability of temperature data, is highly adaptable to complex and changing environments, significantly eliminates temperature sensing data bias, and enhances system performance and stability.
Smart Images
Figure CN121786770A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of artificial intelligence, and more specifically, to a method and system for compensating temperature sensing data based on machine learning. Background Technology
[0002] Temperature sensors are widely used in numerous fields such as industrial production, environmental monitoring, and smart homes to acquire temperature data. However, in practical applications, temperature sensors are susceptible to interference from various environmental factors, such as ambient humidity, air pressure, electromagnetic radiation, and airflow speed. These factors can cause deviations between the raw data output by the temperature sensor and the actual temperature. Traditional temperature sensor data compensation methods typically employ fixed compensation formulas or simple linear correction models. These methods are often based on preset assumptions and are difficult to adapt to complex and changing environmental conditions. Because the degree and manner in which various environmental factors affect temperature sensors differ under different environments, fixed-mode methods cannot accurately capture these dynamic changes, resulting in limited compensation effects and an inability to effectively eliminate deviations in temperature sensor data. Consequently, the accuracy and reliability of subsequent decision-making and control based on temperature data are affected. Therefore, a compensation method that can adapt to different environments and effectively eliminate temperature sensor data deviations is needed. Summary of the Invention
[0003] In view of the aforementioned problems, and in conjunction with the first aspect of the present invention, embodiments of the present invention provide a temperature sensing data compensation method based on machine learning, the method comprising: Acquire the raw temperature sensing data output by the temperature sensor and the environmental impact feature data corresponding to the sensing environment. The environmental impact feature data is a set of environmental features related to the temperature sensing deviation. A correlation model is constructed between the raw temperature sensing data and the environmental impact characteristic data to generate a temperature sensing deviation correlation model. The temperature sensing deviation correlation model contains the dynamic correlation between the raw temperature sensing data and the environmental impact characteristic data. An adaptive learning model is invoked to optimize the temperature sensing deviation correlation model, generating a temperature compensation parameter set adapted to the current environment. The dimension of the temperature compensation parameter set is consistent with the acquisition dimension of the original temperature sensing data. The temperature compensation parameter set is applied to the raw temperature sensing data to perform a deviation correction operation and generate preliminary compensated temperature data. The optimization parameters of the adaptive learning model are adjusted in reverse based on the initial compensated temperature data, the temperature compensation parameter set is updated, and the updated temperature compensation parameter set is applied again to the original temperature sensing data to output the final compensated temperature data.
[0004] In another aspect, embodiments of the present invention also provide a temperature sensing data compensation system based on machine learning, including a processor and a machine-readable storage medium connected to the processor. The machine-readable storage medium is used to store programs, instructions, or code, and the processor is used to execute the programs, instructions, or code in the machine-readable storage medium to implement the above-described method.
[0005] Based on the above, this embodiment of the invention provides a rich information foundation for accurate compensation by comprehensively acquiring the raw data output by the temperature sensor and the environmental impact characteristic data related to temperature sensing deviation. The temperature sensing deviation correlation model generated by correlation modeling can deeply explore the dynamic correlation between the raw temperature sensing data and the environmental impact characteristic data, accurately grasping the formation pattern of temperature sensing deviation under different environments. Optimizing the temperature sensing deviation correlation model by calling the adaptive learning model can generate a temperature compensation parameter set adapted to the current environment. This parameter set is consistent with the dimensions of the raw data acquisition, ensuring the targeting and effectiveness of the compensation. After applying the temperature compensation parameter set to the raw data to generate preliminary compensated temperature data, the optimized parameters of the adaptive learning model are adjusted based on the preliminary compensated data, the parameter set is updated, and applied again. This feedback mechanism allows the compensation process to continuously self-optimize and adjust, highly adapting to complex and changing environments, effectively eliminating deviations in the temperature sensing data, significantly improving the accuracy and reliability of the temperature data, providing solid data support for various subsequent applications based on temperature data, and improving the performance and stability of the entire system. Attached Figure Description
[0006] Figure 1 This is a schematic diagram of the execution flow of the temperature sensing data compensation method based on machine learning provided in an embodiment of the present invention.
[0007] Figure 2 This is a schematic diagram of exemplary hardware and software components of a temperature sensing data compensation system based on machine learning provided in an embodiment of the present invention. Detailed Implementation
[0008] The present invention will now be described in detail with reference to the accompanying drawings. Figure 1 This is a flowchart illustrating a temperature sensing data compensation method based on machine learning according to an embodiment of the present invention. The following is a detailed description of the temperature sensing data compensation method based on machine learning.
[0009] Step S110: Obtain the raw temperature sensing data output by the temperature sensor and the environmental impact feature data corresponding to the sensing environment. The environmental impact feature data is a set of environmental features related to the temperature sensing deviation.
[0010] In this embodiment, the method is applied to a temperature monitoring scenario in an industrial production workshop. Temperature sensors are deployed in different equipment areas of the workshop to collect temperature data during equipment operation. The raw temperature sensing data is a sequence of temperature values output by the temperature sensor during continuous acquisition. This sequence contains multiple temperature data points, each corresponding to a collection timestamp. Environmental impact characteristic data refers to environmental factor data related to temperature sensing deviation within the workshop, including vibration frequency data of the equipment during operation, air velocity data within the workshop, electromagnetic radiation intensity data around the equipment, and air humidity data. These environmental factors can affect the measurement accuracy of the temperature sensor, thus causing deviations in the raw temperature sensing data. When collecting raw temperature sensing data and environmental impact characteristic data, data desensitization technology is used for privacy-sensitive data (such as specific parameters of equipment operation that may involve enterprise production privacy). Sensitive identification information in the data is replaced or deleted, and encrypted transmission is used to ensure that the data is not leaked during transmission.
[0011] Step S120: Perform correlation modeling on the original temperature sensing data and environmental impact characteristic data to generate a temperature sensing deviation correlation model. The temperature sensing deviation correlation model contains the dynamic correlation between the original temperature sensing data and the environmental impact characteristic data.
[0012] In this embodiment, in order to accurately generate the temperature sensing deviation correlation model, it is necessary to perform multi-dimensional processing and analysis on the original temperature sensing data and environmental impact characteristic data.
[0013] Step S121: Extract the temporal variation features from the raw temperature sensing data. The temporal variation features reflect the fluctuation pattern of the raw temperature sensing data over the acquisition time.
[0014] In this embodiment, the extraction of temporal variation features of the original temperature sensing data is achieved by analyzing the changes in temperature data points over consecutive timestamps. Specifically, for each temperature data point, the temperature difference between it and the previous timestamp is calculated to obtain a temperature difference sequence; simultaneously, the average and variance of the temperature data points within each time window are calculated to obtain a temperature average sequence and a temperature variance sequence. Combining the temperature difference sequence, the temperature average sequence, and the temperature variance sequence forms the temporal variation features of the original temperature sensing data, which can comprehensively reflect the fluctuation pattern of the original temperature sensing data over time.
[0015] Step S122: Extract environmental variable features from the environmental impact feature data. The environmental variable features reflect the characteristic distribution status of different environmental factors.
[0016] In this embodiment, the environmental impact characteristic data includes vibration frequency data, air velocity data, electromagnetic radiation intensity data, and air humidity data. For vibration frequency data, its frequency distribution characteristics are extracted, including the proportion of vibration intensity within different frequency ranges; for air velocity data, its velocity variation characteristics are extracted, including the maximum, minimum, and rate of change of velocity; for electromagnetic radiation intensity data, its intensity distribution characteristics are extracted, including the proportion of radiation duration within different intensity ranges; and for air humidity data, its humidity variation characteristics are extracted, including the average humidity and its fluctuation range. These extracted features are combined to form environmental variable characteristics, which can reflect the characteristic distribution state of different environmental factors.
[0017] Step S123: Associate and map the time-series variation features with environmental variable features to establish a corresponding relationship between the time-series variation features and environmental variable features.
[0018] In this embodiment, a series of processing steps are required to establish the corresponding relationship between the two.
[0019] Step S1231: Determine the feature dimensions of the time-series variation features and the feature dimensions of the environmental variable features, and adjust the dimensional representation method of the time-series variation features to maintain consistency with the dimensional representation method of the environmental variable features.
[0020] In this embodiment, the feature dimensions of the time-series variation features include the temperature difference dimension, the average temperature dimension, and the temperature variance dimension; the feature dimensions of the environmental variable features include the vibration frequency distribution dimension, the air velocity variation dimension, the electromagnetic radiation intensity distribution dimension, and the air humidity variation dimension. Through feature transformation methods, each dimension of the time-series variation features is converted into the same representation form as the corresponding dimension of the environmental variable features. For example, the numerical range of the temperature difference dimension is adjusted to be consistent with the numerical range of the air velocity variation dimension, ensuring that both maintain consistency in their dimensional representation methods.
[0021] Step S1232: Align the temporal change features and environmental variable features according to the collection time sequence to form feature sequence pairs with consistent time dimension.
[0022] In this embodiment, the acquisition timestamps of the original temperature sensing data and environmental impact feature data are synchronized. Therefore, the temporal variation features and environmental variable features can be directly aligned according to the acquisition time order. Each feature sequence pair contains a temporal variation feature and an environmental variable feature corresponding to a timestamp, forming a set of feature sequence pairs with consistent time dimensions.
[0023] Step S1233: Perform sliding window processing on the feature sequence pairs to divide them into multiple feature sub-sequence pairs. Each feature sub-sequence pair contains time-series change feature segments and environmental variable feature segments within the same time window.
[0024] In this embodiment, the size of the sliding window is determined based on the interval of the acquisition time and the frequency of data change, and the sliding step is one acquisition time interval. Through sliding window processing, the feature sequence pair set is divided into multiple feature sub-sequence pairs, and each feature sub-sequence pair contains time-series change feature segments and environmental variable feature segments corresponding to multiple consecutive timestamps.
[0025] Step S1234: Analyze the correlation of the changing trends of feature segments in each feature subsequence pair, and extract the combination of feature segments with consistent changing trends.
[0026] In this embodiment, for each feature subsequence pair, the changing trends of the time-series feature segment and the environmental variable feature segment are analyzed separately. The changing trend of the time-series feature segment is determined by calculating the direction and magnitude of the change in temperature difference within the segment, while the changing trend of the environmental variable feature segment is determined by calculating the direction and magnitude of the change in each environmental factor data within the segment. The changing trends of the time-series feature segment and the environmental variable feature segment are compared, and feature segment combinations with consistent changing trends are extracted, i.e., combinations where the changing direction and magnitude of the time-series feature segment are the same as those of the environmental variable feature segment.
[0027] Step S1235: Establish a correspondence between time-series change features and environmental variable features based on the co-occurrence frequency of feature fragment combinations. The correspondence includes the time synchronization information of feature fragment combinations.
[0028] In this embodiment, the co-occurrence frequency of all feature fragment combinations in the feature subsequence pair set is calculated. The co-occurrence frequency is the ratio of the number of times a feature fragment combination appears to the total number of feature subsequence pairs. Feature fragment combinations with high co-occurrence frequencies are identified as feature combinations with corresponding relationships. At the same time, the temporal synchronization information of the feature fragment combinations is recorded, that is, the timestamp correspondence between the time-series change feature fragments and the environmental variable feature fragments in the feature fragment combination.
[0029] Step S124: Mining deviation patterns in the original temperature sensing data based on the corresponding correlation. The deviation pattern is the form of deviation between the original temperature sensing data and the changes in environmental variable characteristics.
[0030] In this embodiment, based on the established correlation between time-series variation characteristics and environmental variable characteristics, the deviation of the original temperature sensing data under different environmental variable characteristics is analyzed. For example, when the vibration frequency distribution in the environmental variable characteristics changes, the changes in the temperature value sequence of the original temperature sensing data are observed to determine the deviation manifestation of the original temperature sensing data, such as whether the temperature value is too high or too low, and the trend of the deviation, thereby uncovering the deviation pattern of the original temperature sensing data.
[0031] Step S125: Quantify the correlation between the deviation pattern and the characteristics of environmental variables. Based on the quantification results, construct a temperature sensing deviation correlation model. The temperature sensing deviation correlation model can output the corresponding deviation pattern according to the changes in the characteristics of environmental variables.
[0032] Step S1251: Determine the characteristic indicators of the deviation model and the characteristic indicators of the environmental variable characteristics. The characteristic indicators of the deviation model and the characteristic indicators of the environmental variable characteristics are used to characterize their respective characteristic attributes.
[0033] In this embodiment, the characteristic indicators of the deviation mode include the direction indicator of the deviation (such as positive or negative deviation) and the magnitude indicator of the deviation change; the characteristic indicators of the environmental variable include the peak value indicator of vibration frequency, the average value indicator of air velocity, the maximum value indicator of electromagnetic radiation intensity, and the fluctuation range indicator of air humidity.
[0034] Step S1252: Calculate the correlation strength between each deviation pattern characteristic index and each environmental variable characteristic index. The correlation strength reflects the degree of mutual influence between the deviation pattern characteristic index and the environmental variable characteristic index.
[0035] In this embodiment, correlation analysis is used to calculate the correlation strength between deviation pattern characteristic indicators and environmental variable characteristic indicators. For example, the correlation strength between the deviation variation amplitude indicator and the peak value of vibration frequency is calculated, and this correlation strength represents the degree of influence of the peak value change of vibration frequency on the deviation variation amplitude.
[0036] Step S1253: Construct an association matrix based on the association strength. The elements of the association matrix are the association strength values between the corresponding deviation pattern characteristic indicators and environmental variable characteristic indicators.
[0037] In this embodiment, the rows of the correlation matrix represent the characteristic indicators of the deviation pattern, the columns represent the characteristic indicators of the environmental variable, and each element in the matrix is the correlation strength value between the corresponding deviation pattern characteristic indicator and the environmental variable characteristic indicator.
[0038] Step S1254: Perform feature dimensionality reduction on the association matrix to retain core association information and simplify the dimensions of the association matrix.
[0039] In this embodiment, a feature dimensionality reduction method is used to process the association matrix, removing elements with low association strength and retaining core association information with high association strength, thereby simplifying the dimension of the association matrix and reducing the complexity of subsequent model construction.
[0040] Step S1255: Construct a temperature sensing deviation correlation model based on the dimensionality-reduced correlation matrix. The temperature sensing deviation correlation model can output the corresponding deviation pattern feature index according to the input environmental variable feature index.
[0041] In this embodiment, the temperature sensing deviation correlation model is a mapping model based on the correlation matrix. It takes environmental variable feature indicators as input and outputs the corresponding deviation pattern feature indicators through the mapping relationship of the correlation matrix, thereby realizing the output of the corresponding deviation pattern according to the changes in environmental variable features.
[0042] Step S130: Call the adaptive learning model to optimize the temperature sensing deviation correlation model and generate a temperature compensation parameter set that adapts to the current environment. The dimension of the temperature compensation parameter set is consistent with the acquisition dimension of the original temperature sensing data.
[0043] In this embodiment, the adaptive learning model is a neural network model, which is used to optimize the temperature sensing deviation correlation model based on the environmental impact feature data of the current environment, and generate a temperature compensation parameter set that can adapt to the current environment.
[0044] Step S131: Extract the real-time environmental features from the environmental impact feature data corresponding to the current environment. The real-time environmental features are the environmental variable features at the current collection time.
[0045] In this embodiment, the current environment is the current equipment operating environment in the industrial production workshop, and the real-time environmental characteristics are the vibration frequency data, air velocity data, electromagnetic radiation intensity data, and air humidity data at the current acquisition time. These data are acquired in real time through corresponding sensors.
[0046] Step S132: Input the real-time environmental features into the temperature sensing deviation association model and output the current deviation pattern corresponding to the real-time environmental features.
[0047] In this embodiment, the peak value of vibration frequency, the average value of air velocity, the maximum value of electromagnetic radiation intensity, and the fluctuation range of air humidity in the real-time environmental characteristics are input into the temperature sensing deviation correlation model. The model outputs the corresponding feature indicators of the current deviation mode through the internal correlation matrix mapping relationship, including the direction indicator of the current deviation and the magnitude of the deviation change, thereby obtaining the current deviation mode.
[0048] Step S133: Input the current deviation mode and real-time environmental features into the feature processing layer of the adaptive learning model, perform feature fusion, and generate a fused feature vector.
[0049] In this embodiment, the feature processing layer of the adaptive learning model includes a feature concatenation module and a feature transformation module. First, the feature indicators of the current deviation mode and the feature indicators of the real-time environment are input into the feature concatenation module for feature concatenation, forming a concatenated feature vector. Then, the concatenated feature vector is input into the feature transformation module, and through feature transformation operations, the concatenated feature vector is converted into a fixed-dimensional fused feature vector, which contains comprehensive information of the current deviation mode and the real-time environment.
[0050] Step S134: The fused feature vector is nonlinearly transformed through the hidden layer of the adaptive learning model to strengthen the nonlinear correlation between features in the fused feature vector.
[0051] Step S1341: Input the fused feature vector into the first hidden layer of the adaptive learning model, and perform a preliminary nonlinear transformation on the fused feature vector through neuron activation operations to generate the first feature vector.
[0052] In this embodiment, the first hidden layer contains multiple neurons, each of which processes the corresponding dimension of the fused feature vector. The neuron activation operation is a non-linear activation function operation, which performs a non-linear transformation on each dimension of the fused feature vector to generate a first feature vector, which contains preliminary non-linear feature information.
[0053] Step S1342: Input the first feature vector into the second hidden layer, and perform feature interaction through a fully connected network to enhance the correlation strength between different feature dimensions in the first feature vector.
[0054] In this embodiment, the second hidden layer is a fully connected network layer, where each neuron is connected to all dimensions of the first feature vector. Through the computation of the fully connected network, information interaction between different feature dimensions in the first feature vector is realized, enhancing the correlation strength between different feature dimensions and generating the second feature vector.
[0055] Step S1343: Reorganize the feature vectors output by the second hidden layer, adjust the order of feature dimensions, and highlight the position of key feature dimensions.
[0056] In this embodiment, the key feature dimensions are those that are highly correlated with the generation of temperature compensation parameters. Through feature recombination operations, these key feature dimensions are adjusted to the beginning of the feature vector so that these key features can be given priority in subsequent processing.
[0057] Step S1344: Input the recombined feature vector into the third hidden layer, extract local feature associations through convolution operation, and generate a second feature vector containing local association information.
[0058] In this embodiment, the third hidden layer is a convolutional layer containing multiple convolutional kernels. The convolutional kernels slide across the recombined feature vector, performing convolution calculations on local regions of the feature vector to extract local feature association information and generate a second feature vector containing local association information.
[0059] Step S1345: Perform global feature integration on the second feature vector, combining local correlation information with global feature distribution to generate a nonlinearly transformed feature vector, thereby strengthening the overall nonlinear correlation between features.
[0060] In this embodiment, the second feature vector is processed by global average pooling to obtain global feature distribution information. Then, the local correlation information is fused with the global feature distribution information to generate a nonlinearly transformed feature vector, which strengthens the overall nonlinear correlation between features.
[0061] Step S135: Generate parameters for the transformed features through the output layer of the adaptive learning model, and generate a temperature compensation parameter set that is adapted to the current environment. The temperature compensation parameter set matches the degree of deviation of the current deviation mode.
[0062] In this embodiment, the output layer of the adaptive learning model is a fully connected layer, and the number of neurons in this layer is consistent with the dimension of the temperature compensation parameter set. Through the calculation of the output layer, the feature vector after nonlinear transformation is converted into a temperature compensation parameter set. Each parameter in this parameter set corresponds to one acquisition dimension of the original temperature sensing data, which can compensate for the deviation of the original temperature sensing data. Moreover, the value of the parameter set matches the degree of deviation of the current deviation mode. The greater the degree of deviation, the larger the corresponding temperature compensation parameter value.
[0063] Step S140: Apply the temperature compensation parameter set to the original temperature sensing data, perform deviation correction operation, and generate preliminary compensated temperature data.
[0064] In this embodiment, by associating the temperature compensation parameter set with the original temperature sensing data, the deviation in the original temperature sensing data is corrected, and preliminary compensated temperature data is obtained.
[0065] Step S141: Establish a mapping between the temperature compensation parameter set and the original temperature sensing data, so that each original temperature sensing data point matches the corresponding temperature compensation parameter.
[0066] In this embodiment, the dimension of the temperature compensation parameter set is consistent with the acquisition dimension of the original temperature sensing data, and each temperature compensation parameter corresponds to the acquisition timestamp of an original temperature sensing data point. A corresponding mapping relationship is established by matching the parameters of the temperature compensation parameter set with the original temperature sensing data points in the order of acquisition time.
[0067] Step S142: Based on the corresponding mapping, the temperature compensation parameters are associated and fused with the original temperature sensing data points to generate intermediate correction data.
[0068] In this embodiment, each original temperature sensing data point is associated with its corresponding temperature compensation parameter based on the established mapping relationship. Specifically, for each original temperature sensing data point, the application method of the temperature compensation parameter is determined according to the direction index of the current deviation mode. For example, when the deviation is positive, the value of the original temperature sensing data point is subtracted from the corresponding temperature compensation parameter; when the deviation is negative, the value of the original temperature sensing data point is added to the corresponding temperature compensation parameter, thereby generating intermediate correction data.
[0069] Step S143: Perform time-series continuity processing on the intermediate correction data, and perform transition processing on the changing trends of adjacent data points. Achieve a smooth transition of the changing trends by adjusting the data point values.
[0070] Step S1431: Arrange the intermediate correction data in the order of collection time to form an ordered data sequence.
[0071] In this embodiment, the intermediate correction data is arranged in chronological order of the collection timestamps to form an ordered data sequence, with each data point corresponding to a collection timestamp.
[0072] Step S1432: Extract the numerical changes of adjacent data points in the ordered data sequence and analyze the distribution characteristics of the changes.
[0073] In this embodiment, for each pair of adjacent data points in an ordered data sequence, the difference between the value of the latter data point and the value of the former data point is calculated to obtain the change in value. By statistically analyzing the changes in value of all adjacent data point pairs, the distribution characteristics of the changes, such as the central tendency and dispersion of the changes, are analyzed.
[0074] Step S1433: Determine the reasonable range of variation for adjacent data points based on the distribution characteristics. The reasonable range of variation reflects the normal fluctuation range of the data in the time series.
[0075] In this embodiment, a reasonable range of variation is determined based on the distribution characteristics of the numerical change. This range includes the numerical changes of most adjacent data point pairs, reflecting the normal fluctuation range of the data in time series.
[0076] Step S1434: Perform transitional adjustments on adjacent data point pairs that exceed the reasonable range of change, adjusting the data point values so that the change falls within the reasonable range of change.
[0077] In this embodiment, for adjacent data point pairs whose numerical changes exceed the reasonable range, the value of the next data point is adjusted according to the values of the adjacent data points and the reasonable range of changes, so that the numerical changes of the adjacent data point pairs fall within the reasonable range of changes.
[0078] Step S1435: Rearrange the adjusted intermediate correction data to form a time-series continuous correction data sequence, and achieve a smooth transition of the change trend of adjacent data points through numerical adjustment.
[0079] In this embodiment, the adjusted intermediate correction data are rearranged according to the acquisition time order to form a time-continuous correction data sequence, and the change trend of adjacent data points in this sequence achieves a smooth transition.
[0080] Step S144: Extract the valid data portion from the intermediate corrected data after time series processing and remove invalid data interference.
[0081] In this embodiment, invalid data refers to data points in the intermediate correction data that do not conform to the normal temperature change pattern, such as data points whose values exceed the normal operating temperature range of the equipment. By setting data filtering conditions, such as the normal operating temperature range of the equipment, data points in the intermediate correction data that exceed this range are removed, and the valid data portion is extracted.
[0082] Step S145: Integrate the effective data into preliminary compensated temperature data, and ensure that the acquisition time series of the preliminary compensated temperature data is consistent with the acquisition time series of the original temperature sensing data.
[0083] In this embodiment, the extracted effective data is integrated according to the acquisition time sequence to form preliminary compensated temperature data. The acquisition time sequence of this data is the same as that of the original temperature sensing data.
[0084] Step S150: Based on the preliminary compensated temperature data, adjust the optimization parameters of the adaptive learning model in reverse, update the temperature compensation parameter set, and apply the updated temperature compensation parameter set back to the original temperature sensing data to output the final compensated temperature data.
[0085] In this embodiment, by analyzing the characteristics of the preliminary compensated temperature data, the optimization parameters of the adaptive learning model are adjusted in reverse to optimize the temperature compensation parameter set, and the original temperature sensing data is corrected again to obtain the final compensated temperature data.
[0086] Step S151: Extract the temporal characteristics of the preliminary compensated temperature data. The temporal characteristics reflect the fluctuation pattern of the data after preliminary compensation.
[0087] In this embodiment, the preliminary compensation temperature data is a temperature value sequence within a continuous acquisition period. The temporal characteristics of this sequence are extracted, including the average value, variance, and trend of the data. These characteristics reflect the fluctuation pattern of the data after preliminary compensation.
[0088] Step S152: Input the temporal features into the feedback adjustment module of the adaptive learning model to generate the adjustment amount of the optimization parameters of the adaptive learning model.
[0089] For example, in step S1521: the time series features are converted into a feature vector form that the feedback adjustment module can recognize, and the dimension of the feature vector is adjusted to meet the input requirements of the feedback adjustment module.
[0090] In this embodiment, the time-series features include multiple feature indicators. These feature indicators are converted into feature vectors that can be recognized by the feedback adjustment module through a feature transformation method, and the dimension of the feature vectors is adjusted to be consistent with the input dimension of the feedback adjustment module.
[0091] Step S1522: Input the transformed feature vector into the feature analysis unit of the feedback adjustment module to extract the key components in the feature vector that are related to the optimization of the adaptive learning model.
[0092] In this embodiment, the feature analysis unit of the feedback adjustment module analyzes the transformed feature vector and identifies the feature components related to the optimization parameters of the adaptive learning model. These key components can reflect the influence of the fluctuation pattern of the preliminary compensation temperature data on the optimization of the adaptive learning model.
[0093] Step S1523: Analyze the key components through the adjustment amount calculation unit of the feedback adjustment module to determine the adjustment direction of each optimization parameter.
[0094] In this embodiment, the adjustment calculation unit determines the adjustment direction of each adaptive learning model optimization parameter based on the changes in the values of the key components. For example, when the value of a key component is greater than a preset threshold, it is determined that the corresponding optimization parameter needs to be increased; when the value of a key component is less than the preset threshold, it is determined that the corresponding optimization parameter needs to be decreased.
[0095] Step S1524: Determine the adjustment range of each optimization parameter based on the values of the key components. The correlation between the values of the key components and the adjustment range is determined by a preset mapping rule.
[0096] In this embodiment, the preset mapping rule is a correspondence table between the values of key components and their adjustment ranges. The adjustment range of each optimization parameter is determined by looking up the correspondence table based on the values of the key components.
[0097] Step S1525: Integrate the adjustment direction and adjustment magnitude to generate the adjustment amount of the adaptive learning model optimization parameters. The adjustment amount includes the specific adjustment value of each parameter.
[0098] In this embodiment, the adjustment direction and adjustment range of each optimization parameter are integrated to generate an adjustment amount. Each element of the adjustment amount corresponds to a specific adjustment value of an adaptive learning model optimization parameter.
[0099] Step S153: Adjust the optimization parameters of the adaptive learning model based on the adjustment amount, and update the feature processing weights and transformation coefficients of the adaptive learning model.
[0100] In this embodiment, the optimization parameters of the adaptive learning model are adjusted according to the generated adjustment values. Specifically, for each optimization parameter, the feature processing weights and transformation coefficients of the model are adjusted according to the specific adjustment values in the adjustment values, such as increasing or decreasing the connection weights of neurons in the feature processing layer, adjusting the transformation coefficients of the hidden layer, etc., thereby updating the adaptive learning model.
[0101] Step S154: Call the updated adaptive learning model to re-optimize the temperature sensing deviation correlation model and generate an updated temperature compensation parameter set.
[0102] In this embodiment, real-time environmental features are input into the updated adaptive learning model. The model, through internal feature processing and parameter generation, re-optimizes the temperature sensing deviation correlation model, generating an updated temperature compensation parameter set. This parameter set better adapts to changes in the current environment, improving the accuracy of temperature compensation.
[0103] Step S155: Apply the updated temperature compensation parameter set to the original temperature sensing data, perform deviation correction operation, and generate and output the final compensated temperature data.
[0104] In this embodiment, following the methods of steps S141 to S145, the updated temperature compensation parameter set is applied to the original temperature sensing data, a deviation correction operation is performed, the final compensated temperature data is generated, and the data is output to the temperature monitoring system of the industrial production workshop for monitoring and analysis of equipment operating status.
[0105] In this embodiment, the above steps achieve deviation compensation for the raw temperature sensing data output by the temperature sensor in the industrial production workshop, improving the measurement accuracy of the temperature data and providing accurate temperature data support for monitoring and analyzing equipment operating status. Simultaneously, privacy protection technology is employed during data acquisition and processing to ensure data security.
[0106] Figure 2 The illustration shows exemplary hardware and software components of a machine learning-based temperature sensing data compensation system 100 that can implement the ideas of this application, according to some embodiments of this application. For example, a processor 120 can be used in the machine learning-based temperature sensing data compensation system 100 and to perform the functions in this application.
[0107] The machine learning-based temperature sensing data compensation system 100 can be a general-purpose server or a special-purpose server; both can be used to implement the machine learning-based temperature sensing data compensation method of this application. Although only one server is shown in this application, for convenience, the functions described in this application can be implemented in a distributed manner on multiple similar platforms to balance the processing load.
[0108] For example, the machine learning-based temperature sensing data compensation system 100 may include a network port 110 connected to a network, one or more processors 120 for executing program instructions, a communication bus 130, and various forms of storage media 140, such as a disk, ROM, or RAM, or any combination thereof. Exemplarily, the machine learning-based temperature sensing data compensation system 100 may also include program instructions stored in ROM, RAM, or other types of non-transitory storage media, or any combination thereof. The methods of this application can be implemented according to these program instructions. The machine learning-based temperature sensing data compensation system 100 also includes an I / O interface 150 between the computer and other input / output devices.
[0109] For ease of explanation, only one processor is described in the machine learning-based temperature sensing data compensation system 100. However, it should be noted that the machine learning-based temperature sensing data compensation system 100 of this application may also include multiple processors, and therefore the steps performed by one processor described in this application may also be performed jointly by multiple processors or individually. For example, if the processor of the machine learning-based temperature sensing data compensation system 100 performs steps A and B, it should be understood that steps A and B may also be performed jointly by two different processors or individually by one processor. For example, the first processor performs step A, the second processor performs step B, or the first processor and the second processor jointly perform steps A and B.
[0110] Furthermore, this embodiment of the invention also provides a readable storage medium, wherein computer-executable instructions are preset in the readable storage medium, and when the processor executes the computer-executable instructions, the above-mentioned temperature sensing data compensation method based on machine learning is implemented.
[0111] It should be noted that, in order to simplify the description of the present invention and thus help to understand one or more embodiments of the invention, multiple features may sometimes be grouped into one embodiment, drawing or description thereof in the foregoing description of the embodiments of the present invention.
Claims
1. A temperature sensing data compensation method based on machine learning, characterized in that, The method includes: Acquire the raw temperature sensing data output by the temperature sensor and the environmental impact feature data corresponding to the sensing environment. The environmental impact feature data is a set of environmental features related to the temperature sensing deviation. A correlation model is constructed between the raw temperature sensing data and the environmental impact characteristic data to generate a temperature sensing deviation correlation model. The temperature sensing deviation correlation model contains the dynamic correlation between the raw temperature sensing data and the environmental impact characteristic data. An adaptive learning model is invoked to optimize the temperature sensing deviation correlation model, generating a temperature compensation parameter set adapted to the current environment. The dimension of the temperature compensation parameter set is consistent with the acquisition dimension of the original temperature sensing data. The temperature compensation parameter set is applied to the raw temperature sensing data to perform a deviation correction operation and generate preliminary compensated temperature data. The optimization parameters of the adaptive learning model are adjusted in reverse based on the initial compensated temperature data, the temperature compensation parameter set is updated, and the updated temperature compensation parameter set is applied again to the original temperature sensing data to output the final compensated temperature data.
2. The temperature sensing data compensation method based on machine learning according to claim 1, characterized in that, The process of correlating and modeling the original temperature sensing data and environmental impact characteristic data to generate a temperature sensing deviation correlation model includes: Extract the temporal variation features from the raw temperature sensing data. The temporal variation features reflect the fluctuation pattern of the raw temperature sensing data over the acquisition time. Extract environmental variable features from environmental impact characteristic data; environmental variable features reflect the characteristic distribution of different environmental factors. The temporal variation characteristics are correlated and mapped with the environmental variable characteristics to establish the corresponding correlation between the temporal variation characteristics and the environmental variable characteristics; Based on the corresponding correlation, the deviation patterns of the original temperature sensing data are mined. The deviation patterns are the deviation manifestations of the original temperature sensing data and the changes in the characteristics of environmental variables. The correlation between the quantification deviation pattern and the characteristics of environmental variables is determined. Based on the quantification results, a temperature sensing deviation correlation model is constructed. The temperature sensing deviation correlation model can output the corresponding deviation pattern according to the changes in the characteristics of environmental variables.
3. The temperature sensing data compensation method based on machine learning according to claim 2, characterized in that, The step of associating and mapping time-series variation features with environmental variable features to establish a corresponding relationship between them includes: Determine the feature dimensions of time-series variation characteristics and the feature dimensions of environmental variables, and adjust the dimensional representation method of time-series variation characteristics to maintain consistency with the dimensional representation method of environmental variables; Align temporal variation features and environmental variable features according to the collection time sequence to form feature sequence pairs with consistent time dimension; The feature sequence pairs are processed by a sliding window to divide them into multiple feature sub-sequence pairs. Each feature sub-sequence pair contains time-series change feature segments and environmental variable feature segments within the same time window. Analyze the correlation of the changing trends of feature segments in each feature subsequence pair, and extract combinations of feature segments with consistent changing trends; The correlation between temporal variation features and environmental variable features is established based on the co-occurrence frequency of feature fragment combinations. The correlation includes the temporal synchronization information of feature fragment combinations.
4. The temperature sensing data compensation method based on machine learning according to claim 1, characterized in that, The process of calling the adaptive learning model to optimize the temperature sensing deviation correlation model and generating a temperature compensation parameter set adapted to the current environment includes: Extract real-time environmental features from the environmental impact feature data corresponding to the current environment. The real-time environmental features are the environmental variable features at the current time of data collection. Input real-time environmental features into the temperature sensing deviation correlation model, and output the current deviation pattern corresponding to the real-time environmental features; The current deviation pattern and real-time environmental features are input into the feature processing layer of the adaptive learning model to perform feature fusion and generate a fused feature vector. The hidden layer of the adaptive learning model performs a non-linear transformation on the fused feature vector, thereby strengthening the non-linear correlation between features in the fused feature vector; The output layer of the adaptive learning model generates parameters for the transformed features, producing a temperature compensation parameter set that is adapted to the current environment. The temperature compensation parameter set matches the degree of deviation of the current deviation mode.
5. The temperature sensing data compensation method based on machine learning according to claim 4, characterized in that, The method of performing nonlinear transformation on the fused feature vector through the hidden layer of the adaptive learning model to strengthen the nonlinear correlation between features in the fused feature vector includes: The fused feature vector is input into the first hidden layer of the adaptive learning model, and the fused feature vector is subjected to a preliminary nonlinear transformation through neuron activation operations to generate the first feature vector. The first feature vector is input into the second hidden layer, and feature interaction is performed through a fully connected network to enhance the correlation strength between different feature dimensions in the first feature vector. The feature vector output by the second hidden layer is reorganized, and the order of the feature dimensions is adjusted to highlight the position of the key feature dimensions. The recombined feature vector is input into the third hidden layer, and local feature associations are extracted through convolution operations to generate a second feature vector containing local association information. The second feature vector is subjected to global feature integration, which combines local correlation information with global feature distribution to generate a nonlinearly transformed feature vector, thereby strengthening the overall nonlinear correlation between features.
6. The temperature sensing data compensation method based on machine learning according to claim 1, characterized in that, The step of applying the temperature compensation parameter set to the original temperature sensing data, performing a deviation correction operation, and generating preliminary compensated temperature data includes: Establish a mapping between the temperature compensation parameter set and the original temperature sensing data, so that each original temperature sensing data point matches the corresponding temperature compensation parameter. Based on the corresponding mapping, the temperature compensation parameters are associated and fused with the original temperature sensing data points to generate intermediate correction data; The intermediate correction data is processed to ensure time continuity, and the changing trends of adjacent data points are processed to achieve a smooth transition of the changing trends by adjusting the data point values. Extract the valid data portion from the intermediate corrected data after time series processing and remove invalid data interference; The effective data is integrated into preliminary compensated temperature data, and the acquisition time series of the preliminary compensated temperature data is consistent with the acquisition time series of the original temperature sensing data.
7. The temperature sensing data compensation method based on machine learning according to claim 6, characterized in that, The process of performing time-series continuity processing on intermediate correction data and transition processing on the changing trends of adjacent data points, achieving a smooth transition of changing trends by adjusting data point values, includes: The intermediate correction data are arranged in chronological order of collection time to form an ordered data sequence; Extract the numerical changes of adjacent data points in an ordered data sequence and analyze the distribution characteristics of these changes. Based on the distribution characteristics, the reasonable range of variation of adjacent data points is determined. The reasonable range of variation reflects the normal fluctuation range of data in time series. For adjacent data point pairs that exceed the reasonable range of change, make transitional adjustments to adjust the data point values so that the change falls within the reasonable range of change. The intermediate corrected data are rearranged and adjusted to form a time-series corrected data sequence, and the changing trends of adjacent data points are smoothly transitioned through numerical adjustments.
8. The temperature sensing data compensation method based on machine learning according to claim 1, characterized in that, The process involves adjusting the optimization parameters of the adaptive learning model based on the initial compensated temperature data, updating the temperature compensation parameter set, and then applying the updated temperature compensation parameter set back to the original temperature sensing data to output the final compensated temperature data. This includes: Extract the temporal characteristics of the preliminary compensated temperature data; the temporal characteristics reflect the fluctuation pattern of the data after preliminary compensation. The temporal features are input into the feedback adjustment module of the adaptive learning model to generate the adjustment amount of the optimization parameters of the adaptive learning model; The optimization parameters of the adaptive learning model are adjusted based on the adjustment amount, and the feature processing weights and transformation coefficients of the adaptive learning model are updated. The updated adaptive learning model is invoked to re-optimize the temperature sensing deviation correlation model, generating an updated set of temperature compensation parameters. The updated temperature compensation parameter set is applied to the original temperature sensing data, a deviation correction operation is performed, and the final compensated temperature data is generated and output.
9. The temperature sensing data compensation method based on machine learning according to claim 2, characterized in that, The correlation between the quantified deviation pattern and environmental variable characteristics is used to construct a temperature sensing deviation correlation model based on the quantification results, including: Determine the characteristic indicators of the deviation pattern and the characteristic indicators of the environmental variable characteristics. The characteristic indicators of the deviation pattern and the characteristic indicators of the environmental variable characteristics are used to characterize their respective characteristic attributes. Calculate the correlation strength between each deviation pattern characteristic index and each environmental variable characteristic index. The correlation strength reflects the degree of mutual influence between the deviation pattern characteristic index and the environmental variable characteristic index. A correlation matrix is constructed based on the correlation strength, and the elements of the correlation matrix are the correlation strength values between the corresponding deviation pattern characteristic indicators and environmental variable characteristic indicators. Feature dimensionality reduction is performed on the association matrix to retain core association information and simplify the dimensions of the association matrix; A temperature sensing deviation correlation model is constructed based on the dimensionality-reduced correlation matrix. The temperature sensing deviation correlation model can output the corresponding deviation pattern feature index according to the input environmental variable feature index.
10. A temperature sensing data compensation system based on machine learning, characterized in that, The machine learning-based temperature sensing data compensation system includes a processor and a memory, the memory and the processor being connected. The memory is used to store programs, instructions or code, and the processor is used to execute the programs, instructions or code in the memory to implement the machine learning-based temperature sensing data compensation method according to any one of claims 1-9.