Intelligent detection data calibration system based on enhanced algorithm driving

Abstract

The application discloses an intelligent detection data calibration system based on an enhanced algorithm driver and specifically relates to the technical field of data calibration, a plurality of sensors are arranged at different positions in an environment, a plurality of same-type sensors are included in each position, redundancy or a machine learning algorithm is used to detect potential abnormal sensors, the range of the abnormal sensors is quickly narrowed, and efficiency is improved; a monitoring data fluctuation curve is analyzed, expected input cost description parameters and expected calibration similarity are obtained through analysis, uncalibratable sensors are screened out from the abnormal sensor set based on the analysis result, and calibration resources are prevented from being wasted on sensors that are not worth inputting; only calibratable sensors are ensured to be included in a calibration process, which helps improve the data accuracy and reliability of the entire monitoring system; detailed cost analysis about sensor maintenance or replacement is provided for decision makers, which helps make more economic and reasonable decisions, and solves the problem of resource waste in the intelligent detection data calibration process.

Description

Technical Field

[0001] The present invention relates to the technical field of data calibration, and more particularly to an intelligent detection data calibration system driven by an enhanced algorithm. Background Art

[0002] In the process of environmental monitoring through sensors, there is a lack of intelligent detection and data calibration of sensor data, which results in the collected sensors not accurately reflecting the actual situation; environmental monitoring sensors include at least temperature sensors, humidity sensors, air pressure sensors, and harmful gas concentration sensors; due to the diverse data sources and complex processing procedures, there may be inconsistencies between data at different time points. In order to ensure the accuracy and consistency of sensor data, data calibration is required.

[0003] However, it still has many shortcomings in actual use. For example, the existing intelligent detection data has the problem of wasting calibration resources, which is specifically reflected in the following aspects:

[0004] Many current systems rely on frequent manual intervention or resetting of calibration parameters, which is not only inefficient but also prone to waste of resources;

[0005] In a multi-sensor monitoring system, some sensors may have low calibration feasibility due to aging, damage, or design defects. Unnecessary calibration of such sensors is undoubtedly a huge waste of resources.

[0006] If calibration resources (such as computing resources and energy) cannot be allocated reasonably and evenly, critical measurement ranges may not receive adequate calibration support due to insufficient resources, thus affecting the accuracy and reliability of the entire monitoring system. Summary of the Invention

[0007] To overcome the above-mentioned shortcomings of the prior art, the present invention provides an intelligent detection data calibration system driven by an enhanced algorithm. By screening out non-calibrable sensors, calibration resources are avoided from being wasted on sensors that are not worth investing in. This ensures that only calibrable sensors are included in the calibration process, which helps to improve the data accuracy and reliability of the entire monitoring system. In addition, it provides decision makers with a detailed cost analysis of sensor repair or replacement to address the issues raised in the above-mentioned background technology.

[0008] To achieve the above objectives, the present invention provides the following technical solution: an intelligent detection data calibration system driven by an enhanced algorithm, comprising:

[0009] The abnormal sensor initial screening module sets up several sensors at different locations in the environment. Each location includes several sensors of the same type. If the monitoring data of sensors of the same type are different, it indicates that the sensor at that location is abnormal, and the abnormal sensor set is screened out.

[0010] The abnormal sensor testing module places a calibrated sensor and an abnormal sensor in the same environment for testing, and records the monitoring data fluctuation curve; obtains the standard monitoring data fluctuation curve A of the calibrated sensor and the abnormal monitoring data fluctuation curve B of the abnormal sensor;

[0011] The feasibility analysis module is used to analyze the monitoring data fluctuation curve, obtain the expected input cost description parameters and expected calibration similarity, and filter out non-calibrable sensors from the abnormal sensor set based on the analysis results, including:

[0012] Applying a correction instruction to the abnormal monitoring data fluctuation curve B to maximize the expected calibration similarity between the standard curve A and the abnormal monitoring data fluctuation curve B, wherein the expected calibration similarity is an indicator of the similarity between the standard monitoring data fluctuation curve A and the abnormal monitoring data fluctuation curve B of the abnormal sensor; the expected calibration similarity is recorded as Yxs;

[0013] Establish a mapping relationship between correction instructions and fault type and fault severity. Based on the correction instructions and mapping relationship, output the fault severity and type. Establish a fault repair cost map. Input the fault severity and type and output the time cost, economic cost, and computing resource cost of repair. Based on the time cost, economic cost, and computing resource cost, obtain the expected input cost description parameters.

[0014] If the expected input cost description parameter and the expected calibration similarity exceed the corresponding threshold, it indicates that the abnormal sensor is an uncalibrable sensor.

[0015] Preferably, the abnormal sensor initial screening module further includes identifying a set of abnormal sensors in a complex environment by using a deep learning model, including the following steps:

[0016] Step S11, collecting temperature data: continuously collecting temperature data from all temperature sensors and extracting key features from the temperature data;

[0017] Step S12: Divide the collected data into a training set, a validation set, and a test set, and mark them to distinguish between data with normal temperature and data with abnormal temperature;

[0018] Step S13: Selecting a long short-term memory network as the model. During the training phase, the LSTM model is trained using the training set data to enable it to learn the patterns of normal temperature data and attempt to distinguish these patterns from abnormal data. The performance of the model is evaluated using the validation set data, including the accuracy, recall rate, and F1 score indicators. Model parameters are adjusted based on the evaluation results to optimize model performance. The optimized model is used as a sensor anomaly prediction model for real-time detection.

[0019] Step S14: After the training is completed, the sensor anomaly prediction model is used to detect anomalies in sensor data in real time. When new data is input into the sensor anomaly prediction model, a prediction result is output, indicating the confidence level of whether the data is abnormal, and the abnormal sensor set is screened out.

[0020] Preferably, the key features include: time series features, statistical features and spatial features. Time series features refer to the rate, acceleration and periodic pattern of temperature change; statistical features refer to the mean, variance, maximum value and minimum value, which represent the information of data distribution; spatial features refer to: the layout characteristics of sensors in space, which represent the spatial layout information of sensors, that is, the temperature change characteristics caused by the spatial layout of sensors.

[0021] Preferably, the abnormal sensor initial screening module also includes a data integration unit and a noise filtering unit. The data integration unit is responsible for integrating the raw data from different sensors, including time synchronization and data format unification, to ensure data consistency and comparability; the noise filtering unit uses a filtering algorithm (such as Kalman filtering, median filtering, etc.) to pre-process the sensor data to remove noise data caused by environmental interference, equipment jitter and other factors, thereby improving data quality.

[0022] Preferably, the correction instruction includes at least forward and backward movement and temperature-based correction; the forward and backward movement refers to a translation operation on the abnormal monitoring data fluctuation curve B, which is used to adjust the time offset and baseline offset; temperature-based correction refers to adjusting the abnormal monitoring data fluctuation curve B according to the temperature change to eliminate the influence of temperature on the sensor reading.

[0023] Preferably, the calculation formula of the expected input cost description parameter YC is

[0024]

[0025] Among them, Trepair is the total time required for calibration, CT is the total cost required for calibration; EC is the economic cost, directly expressed in monetary units; TC_time is the value after the time cost is converted into monetary units, which is the time cost multiplied by one unit of time cost, and f1 is the conversion coefficient from time cost to economic cost; TC_com is the value after the computing power resource cost is converted into monetary units, and f2 is the conversion coefficient from computing power resource cost to economic cost.

[0026] Preferably, if the expected input cost description parameter and the expected calibration similarity exceed corresponding thresholds, the expected input cost description parameter YC and the expected calibration similarity Yxs are input into the calibration feasibility analysis model, and the calibration feasibility index JK is output;

[0027] The calibration feasibility analysis model satisfies the formula Where YC_norm represents the expected input cost description parameter after linear normalization, Yxs_norm represents the expected calibration similarity after linear normalization, and λ is a correction constant to avoid division by zero;

[0028] Determine whether the calibration feasibility index exceeds the threshold THa. If the calibration feasibility index exceeds the threshold THa, it indicates that the corresponding abnormal sensor is an uncalibrable sensor, and a second-level warning is issued, prompting the sensor to be replaced;

[0029] If the calibration feasibility index does not exceed the threshold THa, it indicates that the corresponding abnormal sensor is a calibrable sensor, and the calibrable sensor set is obtained.

[0030] Preferably, a mapping relationship between correction instructions and fault type and fault degree is established by analyzing the commonalities and differences of correction instructions under different fault types and degrees; the fault type and degree are inferred based on the correction instructions of the mapping relationship and the abnormal monitoring data fluctuation curve B.

[0031] Preferably, the system further includes an abnormal sensor calibration module for calibrating the calibratable sensor set, including:

[0032] The area division unit is used to divide the sensor data into several areas according to the range and number them. n represents the number of areas and i represents the area number.

[0033] The calibration requirement analysis unit obtains the amplitude and frequency of data deviation in each area based on the standard monitoring data fluctuation curve A and the abnormal monitoring data fluctuation curve B of the abnormal sensor; constructs a calibration requirement index based on the amplitude and frequency of the data deviation, and outputs the calibration requirement index for each area;

[0034] The calibration resource allocation unit assigns calibration priority to each area based on the calibration demand index; and reasonably allocates calibration resources according to the calibration demand index of each area.

[0035] The model selection unit uses a polynomial function to correct the abnormal monitoring data fluctuation curve B of each area and uses statistical methods to find the polynomial function corresponding to each area;

[0036] An iterative calibration unit ensures that the intelligent detection data meets preset accuracy requirements through iterative calibration. The iterative calibration refers to performing a preliminary calibration on the calibrable sensor, then collecting monitoring data in actual applications to evaluate the calibration accuracy, and then fine-tuning the calibration parameters based on the evaluated accuracy. This process is repeated until satisfactory calibration accuracy is achieved.

[0037] Preferably, the calibration requirement index Jx satisfies the formula

[0038]

[0039] Among them, pcf_avg and pcp_avg represent the average data deviation amplitude and data deviation frequency of the region, respectively; pcf_max and pcp_max represent the maximum deviation amplitude and the highest deviation frequency of the region, respectively; α and β are weight coefficients used to adjust the relative importance of deviation amplitude and frequency in the index; the maximum deviation amplitude and the highest deviation frequency are the maximum values ​​of all regions and are used for normalization processing.

[0040] Preferably, it also includes a calibration quality management module, which records the calibration process data and analyzes it to obtain the calibration stability parameter and calibration accuracy parameter, wherein the calibration process data includes the actual calibration time, the interval time between two adjacent calibration behaviors and the actual error after calibration; the calibration stability parameter and the calibration accuracy parameter are input into the calibration quality assessment model, and the calibration quality assessment is output. If the calibration quality assessment index is lower than the threshold THb, an early warning is issued to prompt attention to the calibration behavior and calibration instructions; the calibration quality assessment model satisfies the formula JP=Form(Jw*Jz), wherein Jw represents the calibration stability parameter, Jz represents the calibration accuracy parameter, and Form(·) represents the linear normalization function.

[0041] Preferably, the calibration stability parameter satisfies the formula

[0042]

[0043] Among them, t 1i represents the actual calibration time of the i-th region, t 2i represents the interval between two adjacent calibration actions of the i-th region, t 10 Indicates the preset calibration time, t 20 Indicates the preset calibration interval; the weight coefficient K of the i-th area is expressed by the ratio of the range of the i-th area to the total range. i; The calibration accuracy parameter satisfies the formula

[0044]

[0045] Among them, d i represents the actual error after calibration of the i-th region, and d0 represents the preset error.

[0046] Technical effects and advantages of the present invention:

[0047] (1) The intelligent detection data calibration system provided by the present invention uses redundancy or machine learning algorithms to detect potential abnormal sensors, which can quickly narrow the scope of abnormal sensors and improve efficiency. By screening out sensors that cannot be calibrated, calibration resources can be avoided from being wasted on sensors that are not worth investing in. It ensures that only calibrable sensors are included in the calibration process, which helps to improve the data accuracy and reliability of the entire monitoring system. It provides decision makers with a detailed cost analysis of sensor repair or replacement, which helps them make more economical and reasonable decisions and solves the problem of resource waste in the intelligent detection data calibration process.

[0048] (2) The intelligent detection data calibration system provided by the present invention analyzes the calibration stability parameters and calibration accuracy parameters through the provided calibration quality management module, which helps the system administrator or operator to better understand the calibration status, discover problems in time and take corresponding measures, thereby ensuring the accuracy and reliability of the monitoring data; by issuing early warning prompts to pay attention to calibration behavior and calibration instructions, the system can also promote the continuous optimization and improvement of calibration work. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] Figure 1 This is a flowchart of the intelligent detection data calibration system based on feasibility analysis of the present invention.

[0050] Figure 2 This is a structural block diagram of the intelligent detection data calibration system driven by the enhanced algorithm of the present invention. DETAILED DESCRIPTION

[0051] Exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although exemplary embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments set forth herein. Instead, these embodiments are provided to enable a more thorough understanding of the present disclosure and to fully convey the scope of the present disclosure to those skilled in the art.

[0052] At the same time, it should be understood that for the convenience of description, the sizes of the various parts shown in the drawings are not drawn according to the actual proportional relationship.

[0053] The following description of at least one exemplary embodiment is merely illustrative in nature and is in no way intended to limit the present disclosure, its application, or uses.

[0054] Technologies, methods, and equipment known to ordinary technicians in the relevant art may not be discussed in detail, but where appropriate, the technologies, methods, and equipment should be considered part of the specification.

[0055] Example 1, see Figure 1 The present invention provides a flowchart of an intelligent detection data calibration system based on feasibility analysis. Figure 1 The intelligent detection data calibration system shown is driven by an enhanced algorithm and includes:

[0056] The abnormal sensor initial screening module sets up several sensors at different locations in the environment. Each location includes several sensors of the same type. If the monitoring data of sensors of the same type are different, it indicates that the sensor at that location is abnormal, and the abnormal sensor set is screened out.

[0057] It should be further explained in the embodiment of the present invention that the abnormal sensor initial screening module also includes a data integration unit and a noise filtering unit. The data integration unit is responsible for integrating the raw data from different sensors, including time synchronization and data format unification to ensure data consistency and comparability; the noise filtering unit uses a filtering algorithm (such as Kalman filtering, median filtering, etc.) to pre-process the sensor data to remove noise data caused by environmental interference, equipment jitter, etc., thereby improving data quality;

[0058] Explanation: Several sensors are set up at different locations in the environment, with each location containing several sensors of the same type. This is done to leverage redundancy to detect potential anomalies. At each location, the monitoring data from sensors of the same type is compared. If these data differ, this is usually a strong signal that one or more sensors at that location may be anomaly. Based on these comparison results, a preliminarily selected set of anomalous sensors can be identified. These anomalous sensors will require further testing or calibration, quickly narrowing down the number of anomalous sensors and improving efficiency.

[0059] The abnormal sensor testing module places a calibrated sensor and an abnormal sensor in the same environment for testing, and records the monitoring data fluctuation curve; obtains the standard monitoring data fluctuation curve A of the calibrated sensor and the abnormal monitoring data fluctuation curve B of the abnormal sensor;

[0060] Explanation: The monitoring data fluctuation curve of the calibrated sensor is recorded as the standard monitoring data fluctuation curve A; the monitoring data fluctuation curve of the abnormal sensor is recorded as the abnormal monitoring data fluctuation curve B;

[0061] The feasibility analysis module is used to analyze the monitoring data fluctuation curve, obtain the expected input cost description parameters and expected calibration similarity, and filter out non-calibrable sensors from the abnormal sensor set based on the analysis results, including:

[0062] Applying a correction instruction to the abnormal monitoring data fluctuation curve B to maximize the expected calibration similarity between the standard curve A and the abnormal monitoring data fluctuation curve B, wherein the expected calibration similarity is an indicator of the similarity between the standard monitoring data fluctuation curve A and the abnormal monitoring data fluctuation curve B of the abnormal sensor; the expected calibration similarity is recorded as Yxs;

[0063] Establish a mapping relationship between correction instructions and fault type and fault severity. Based on the correction instructions and mapping relationship, output the fault severity and type. Establish a fault repair cost map. Input the fault severity and type and output the time cost, economic cost, and computing resource cost of repair. Based on the time cost, economic cost, and computing resource cost, obtain the expected input cost description parameters.

[0064] Explanation: Collect historical maintenance records, including maintenance time, economic expenditure (such as spare parts cost, labor cost) and computing resource usage under different fault types and degrees; use the collected data to establish a fault repair cost map. The fault repair cost map is a multi-dimensional database or model that contains the correspondence between fault type, fault degree and maintenance time cost, economic cost and computing resource cost; input the fault type and degree, and output the maintenance time cost, economic cost and computing resource cost.

[0065] If the expected input cost description parameter and the expected calibration similarity exceed the corresponding threshold, it indicates that the abnormal sensor is an uncalibrable sensor;

[0066] Explanation: The higher the calibration cost of an abnormal sensor, the less necessary it is to calibrate it. To save costs, non-calibrable sensors are screened out to save costs. By screening out non-calibrable sensors, the system can avoid wasting calibration resources on sensors that are not worth investing in. Ensuring that only calibrable sensors are included in the calibration process helps improve the data accuracy and reliability of the entire monitoring system. Providing decision makers with a detailed cost analysis of sensor repair or replacement helps them make more economical and reasonable decisions.

[0067] It is to be explained that the embodiment of the present invention does not specifically limit the calculation method of the expected calibration similarity. The expected calibration similarity may be a mean square error, a correlation coefficient, or other indicators for measuring the similarity between two curves.

[0068] What needs to be further explained in the embodiments of the present invention is that the correction instructions include at least forward and backward movement and temperature-based correction; the forward and backward movement refers to the translation operation performed on the abnormal monitoring data fluctuation curve B, which is used to adjust the time offset and baseline offset, such as zero drift; temperature-based correction refers to adjusting the abnormal monitoring data fluctuation curve B according to the temperature change to eliminate the influence of temperature on the sensor reading, such as temperature drift.

[0069] In the embodiment of the present invention, it is necessary to further explain that the calculation formula of the expected input cost description parameter YC is:

[0070]

[0071] Among them, Trepair is the total time required for calibration, CT is the total cost required for calibration; EC is the economic cost, directly expressed in monetary units; TC_time is the value after the time cost is converted into monetary units, which is the time cost multiplied by one unit of time cost, and f1 is the conversion coefficient from time cost to economic cost; TC_com is the value after the computing power resource cost is converted into monetary units, and f2 is the conversion coefficient from computing power resource cost to economic cost.

[0072] It is necessary to further explain in the embodiment of the present invention that if the expected input cost description parameter and the expected calibration similarity exceed the corresponding threshold, the expected input cost description parameter YC and the expected calibration similarity Yxs are input into the calibration feasibility analysis model, and the calibration feasibility index JK is output;

[0073] The calibration feasibility analysis model satisfies the formula Where YC_norm represents the expected input cost description parameter after linear normalization, Yxs_norm represents the expected calibration similarity after linear normalization, and λ is a correction constant to avoid division by zero;

[0074] Determine whether the calibration feasibility index exceeds the threshold THa. If the calibration feasibility index exceeds the threshold THa, it indicates that the corresponding abnormal sensor is an uncalibrable sensor, and a secondary warning is issued, prompting you to replace the sensor. If the calibration feasibility index does not exceed the threshold THa, it indicates that the corresponding abnormal sensor is a calibrable sensor, and a set of calibrable sensors is obtained.

[0075] Explanation: The larger the calibration feasibility index, the higher the feasibility of sensor calibration. The lower the calibration feasibility index, the more difficult the calibration of the sensor device is and the lower the calibration benefit is. The sensor device is marked as unusable.

[0076] What needs to be further explained in the embodiments of the present invention is that a mapping relationship between correction instructions and fault type and fault degree is established by analyzing the commonalities and differences of correction instructions under different fault types and degrees; the fault type and degree are inferred based on the correction instructions of the mapping relationship and the abnormal monitoring data fluctuation curve B.

[0077] BACKGROUND: Assume that multiple temperature sensors are deployed in an industrial automation environment. The temperature sensors are distributed on different machines and pipelines and are used to monitor the temperature of the equipment to prevent failures caused by overheating.

[0078] In a possible embodiment, the abnormal sensor initial screening module further includes identifying a set of abnormal sensors in a complex environment by using a deep learning model, including the following steps:

[0079] Step S11, collecting temperature data: continuously collecting temperature data from all temperature sensors and extracting key features from the temperature data;

[0080] Step S12: Divide the collected data into a training set, a validation set, and a test set, and mark them to distinguish between data with normal temperature and data with abnormal temperature;

[0081] Step S13: Selecting a long short-term memory network as the model. During the training phase, the LSTM model is trained using the training set data to enable it to learn the patterns of normal temperature data and attempt to distinguish these patterns from abnormal data. The performance of the model is evaluated using the validation set data, including the accuracy, recall rate, and F1 score indicators. Model parameters are adjusted based on the evaluation results to optimize model performance. The optimized model is used as a sensor anomaly prediction model for real-time detection.

[0082] Step S14: After the training is completed, the sensor anomaly prediction model is used to detect anomalies in sensor data in real time. When new data is input into the sensor anomaly prediction model, a prediction result is output, indicating the confidence level of whether the data is abnormal, and the abnormal sensor set is screened out.

[0083] Furthermore, the key features include: time series features, statistical features and spatial features. Time series features refer to the rate, acceleration and periodic pattern of temperature change; statistical features refer to the mean, variance, maximum value and minimum value, which represent the information of data distribution; spatial features refer to: the layout characteristics of sensors in space, which represent the spatial layout information of sensors, that is, the temperature change characteristics caused by the spatial layout of sensors.

[0084] For example, suppose that at a certain point in time, the temperature reading of a temperature sensor suddenly rises sharply, exceeding the normal operating range. After data processing and feature extraction, this abnormal temperature change pattern is input into the LSTM model. Based on the previously learned normal temperature pattern, the model finds that this new data point does not match the normal pattern, so it outputs a high-confidence abnormal signal. The abnormal signal is captured by the system and triggers the corresponding alarm or fault handling mechanism, thereby achieving accurate detection of abnormal situations and realizing the use of deep learning models to identify and filter out abnormal sensor sets in complex environments.

[0085] Example 2: The difference between this embodiment of the present invention and Example 1 is that the intelligent detection data calibration system further includes an abnormal sensor calibration module.

[0086] The abnormal sensor calibration module is used to calibrate the calibratable sensor set, including:

[0087] The area division unit is used to divide the sensor data into several areas according to the range and number them. n represents the number of areas and i represents the area number.

[0088] Explanation: Because sensors may exhibit different error characteristics within different measuring ranges, dividing the measuring range into multiple regions can more accurately identify and address the calibration requirements of each region. Dynamically adjusting the region boundaries based on real-time changes in sensor data ensures that the data characteristics within each region are relatively stable, thereby improving calibration accuracy.

[0089] The calibration requirement analysis unit obtains the amplitude and frequency of data deviation in each area based on the standard monitoring data fluctuation curve A and the abnormal monitoring data fluctuation curve B of the abnormal sensor; constructs a calibration requirement index based on the amplitude and frequency of the data deviation, and outputs the calibration requirement index for each area;

[0090] The calibration resource allocation unit assigns calibration priorities to each region based on the calibration demand index. Based on the calibration demand index of each region, calibration resources, such as time, manpower, and equipment, are allocated reasonably to ensure that regions with higher calibration demand indexes are prioritized.

[0091] The model selection unit uses a polynomial function to correct the abnormal monitoring data fluctuation curve B of each area and uses statistical methods to find the polynomial function corresponding to each area;

[0092] Explanation: Polynomial functions are simple and computationally efficient, making them suitable for areas where the deviation relationship is relatively smooth and continuously changing. If the polynomial function cannot be calibrated, a deviation prediction model can be built by training a neural network. This input uses historical data and outputs the corresponding deviation prediction value. The sensor output value is then adjusted based on the deviation prediction value to improve calibration accuracy.

[0093] An iterative calibration unit ensures that the intelligent detection data meets preset accuracy requirements through iterative calibration. The iterative calibration refers to performing a preliminary calibration on the calibrable sensor, then collecting monitoring data in actual applications to evaluate the calibration accuracy, and then fine-tuning the calibration parameters based on the evaluated accuracy. This process is repeated until satisfactory calibration accuracy is achieved.

[0094] It is necessary to further explain in the embodiment of the present invention that the calibration requirement index Jx satisfies the formula

[0095]

[0096] Among them, pcf_avg and pcp_avg represent the average data deviation amplitude and data deviation frequency of the region, respectively; pcf_max and pcp_max represent the maximum deviation amplitude and the highest deviation frequency of the region, respectively; α and β are weight coefficients used to adjust the relative importance of deviation amplitude and frequency in the index; the maximum deviation amplitude and the highest deviation frequency are the maximum values ​​of all regions and are used for normalization processing.

[0097] Example 3, see Figure 2 The structure block diagram of the intelligent detection data calibration system based on the enhanced algorithm drive is shown in FIG. The difference between the embodiment of the present invention and embodiment 2 is that the intelligent detection data calibration system further includes a calibration quality management module.

[0098] Used to record calibration process data and analyze it to obtain calibration stability parameters and calibration accuracy parameters. The calibration process data includes the actual calibration time, the interval between two adjacent calibration actions, and the actual error after calibration;

[0099] Inputting a calibration stability parameter and a calibration accuracy parameter into a calibration quality assessment model and outputting a calibration quality assessment; the calibration quality assessment model satisfies the formula JP=Form(Jw*Jz), wherein Jw represents the calibration stability parameter, Jz represents the calibration accuracy parameter, and Form(·) represents a linear normalization function;

[0100] If the calibration quality assessment index is lower than the threshold THb, an early warning is issued, prompting attention to the calibration behavior and calibration instructions;

[0101] It is necessary to further explain in the embodiment of the present invention that the calibration stability parameter satisfies the formula

[0102]

[0103] Among them, t 1i represents the actual calibration time of the i-th region, t 2iinterval time of adjacent two calibration behaviors of the i-th region, t 10 preset calibration time, t 20 preset calibration behavior interval time; weight coefficient K of the i-th region is represented by ratio of range of the i-th region to total range i ; the calibration accuracy parameter satisfies formula

[0104]

[0105] wherein, d i actual error of the i-th region after calibration, d0 represents preset error.

[0106] It is explained that based on difference between actual consumption time and preset consumption time of calibration, calibration stability parameter is evaluated, divided by converted calibration demand index, influence of calibration demand on calibration stability is eliminated, in this way, calibration stability parameter can more purely reflect stability of calibration process itself, without interference of data deviation amplitude and frequency; based on actual error and preset error after calibration, calibration accuracy parameter is evaluated, divided by converted calibration demand index, influence of calibration demand on calibration accuracy is eliminated.

[0107] Finally: the above only for preferred embodiments of the present application, and is not used to limit the present application, any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application, should be included in the protection scope of the present application.

Claims

1. Intelligent detection data calibration system driven by enhanced algorithm, characterized by: include: The abnormal sensor initial screening module sets up several sensors at different locations in the environment. Each location includes several sensors of the same type. If the monitoring data of sensors of the same type are different, it indicates that the sensor at that location is abnormal, and the abnormal sensor set is screened out. The abnormal sensor testing module places a calibrated sensor and an abnormal sensor in the same environment for testing, and records the monitoring data fluctuation curve; obtains the standard monitoring data fluctuation curve A of the calibrated sensor and the abnormal monitoring data fluctuation curve B of the abnormal sensor; The feasibility analysis module is used to analyze the monitoring data fluctuation curve, obtain the expected input cost description parameters and expected calibration similarity, and filter out non-calibrable sensors from the abnormal sensor set based on the analysis results, including: Applying a correction instruction to the abnormal monitoring data fluctuation curve B to maximize the expected calibration similarity between the standard curve A and the abnormal monitoring data fluctuation curve B, wherein the expected calibration similarity is an indicator of the similarity between the standard monitoring data fluctuation curve A and the abnormal monitoring data fluctuation curve B of the abnormal sensor; the expected calibration similarity is recorded as Yxs; Establish a mapping relationship between correction instructions and fault type and fault severity. Based on the correction instructions and mapping relationship, output the fault severity and type. Establish a fault repair cost map. Input the fault severity and type and output the time cost, economic cost, and computing resource cost of repair. Based on the time cost, economic cost, and computing resource cost, obtain the expected input cost description parameters. The calculation formula for the expected input cost description parameter YC is: , where Trepair is the total time required for calibration, CT is the total cost required for calibration; EC is the economic cost, which is directly expressed in monetary units; TC_time is the value after the time cost is converted into monetary units, which is the time cost multiplied by one unit time cost. It is the conversion coefficient from time cost to economic cost; TC_com is the value after computing power resource cost is converted into monetary unit. It is the conversion coefficient from computing resource cost to economic cost; If the expected input cost description parameter and the expected calibration similarity exceed the corresponding threshold, it indicates that the abnormal sensor is an uncalibrable sensor.

2. The intelligent detection data calibration system based on enhanced algorithm driving according to claim 1 is characterized in that: The correction instruction includes at least forward and backward movement and temperature-based correction; the forward and backward movement refers to a translation operation performed on the abnormal monitoring data fluctuation curve B, which is used to adjust the time offset and baseline offset; Temperature-based correction refers to adjusting the abnormal monitoring data fluctuation curve B according to temperature changes to eliminate the impact of temperature on sensor readings.

3. The intelligent detection data calibration system based on enhanced algorithm driving according to claim 2 is characterized in that: If the expected input cost description parameter and the expected calibration similarity exceed the corresponding threshold, the expected input cost description parameter YC and the expected calibration similarity Yxs are input into the calibration feasibility analysis model, and the calibration feasibility index JK is output; The calibration feasibility analysis model satisfies the formula , where YC_norm represents the expected input cost description parameter after linear normalization, Yxs_norm represents the expected calibration similarity after linear normalization, and λ is a correction constant to avoid division by zero; Determine whether the calibration feasibility index exceeds the threshold THa. If the calibration feasibility index exceeds the threshold THa, it indicates that the corresponding abnormal sensor is an uncalibrable sensor, and a second-level warning is issued, prompting the sensor to be replaced; If the calibration feasibility index does not exceed the threshold THa, it indicates that the corresponding abnormal sensor is a calibrable sensor, and the calibrable sensor set is obtained.

4. The intelligent detection data calibration system based on enhanced algorithm driving according to claim 1 is characterized in that: By analyzing the commonalities and differences of correction instructions under different fault types and degrees, a mapping relationship between correction instructions and fault type and fault degree is established; the fault type and degree are inferred based on the correction instructions of the mapping relationship and the abnormal monitoring data fluctuation curve B.

5. The intelligent detection data calibration system based on enhanced algorithm driving according to claim 1 is characterized in that: Also included is an abnormal sensor calibration module for calibrating the calibratable sensor set, including: The area division unit is used to divide the sensor data into several areas according to the range and number them. n represents the number of areas and i represents the area number. The calibration requirement analysis unit obtains the amplitude and frequency of data deviation in each area based on the standard monitoring data fluctuation curve A and the abnormal monitoring data fluctuation curve B of the abnormal sensor; constructs a calibration requirement index based on the amplitude and frequency of the data deviation, and outputs the calibration requirement index for each area; The calibration resource allocation unit assigns calibration priorities to each region based on the calibration demand index; and reasonably allocates calibration resources based on the calibration demand index of each region; The model selection unit uses a polynomial function to correct the abnormal monitoring data fluctuation curve B of each area and uses statistical methods to find the polynomial function corresponding to each area; An iterative calibration unit ensures that the intelligent detection data meets preset accuracy requirements through iterative calibration. The iterative calibration refers to performing a preliminary calibration on the calibrable sensor, then collecting monitoring data in actual applications to evaluate the calibration accuracy, and then fine-tuning the calibration parameters based on the evaluated accuracy. This process is repeated until satisfactory calibration accuracy is achieved.

6. The intelligent detection data calibration system based on enhanced algorithm driving according to claim 5 is characterized in that: The calibration requirement index Jx satisfies the formula , where pcf_avg and pcp_avg represent the average data deviation amplitude and data deviation frequency of the region, respectively; pcf_max and pcp_max represent the maximum deviation amplitude and the highest deviation frequency of the region, respectively; α and β are weight coefficients used to adjust the relative importance of the deviation amplitude and frequency in the index; the maximum deviation amplitude and the highest deviation frequency are the maximum values ​​of all regions and are used for normalization processing.

7. The intelligent detection data calibration system based on enhanced algorithm driving according to claim 6 is characterized in that: The module also includes a calibration quality management module, which records the calibration process data and analyzes it to obtain calibration stability parameters and calibration accuracy parameters. The calibration process data includes the actual calibration time, the interval between two adjacent calibration actions, and the actual error after calibration. The calibration stability parameters and calibration accuracy parameters are input into the calibration quality assessment model, and the calibration quality assessment is output. If the calibration quality assessment index is lower than the threshold THb, an early warning is issued to prompt attention to the calibration behavior and calibration instructions. The calibration quality assessment model satisfies the formula , where Jw represents the calibration stability parameter, Jz represents the calibration accuracy parameter, and Form(·) represents the linear normalization function.

8. The intelligent detection data calibration system based on enhanced algorithm driving according to claim 7 is characterized in that: The calibration stability parameter satisfies the formula , where t 1i represents the actual calibration time of the i-th region, t 2i represents the interval between two adjacent calibration actions of the i-th region, t 10 Indicates the preset calibration time, t 20 Indicates the preset calibration interval; the weight coefficient K of the i-th area is expressed by the ratio of the range of the i-th area to the total range. i ; The calibration accuracy parameter satisfies the formula: , where d i represents the actual error after calibration of the i-th region, and d0 represents the preset error.

Images