Artificial intelligence toxic gas detection alarm joint calibration method based on multi-sensor data fusion

Abstract

The invention relates to the technical field of toxic gas detection and calibration, in particular to an artificial intelligence toxic gas detection alarm joint calibration method based on multi-sensor data fusion, which comprises the steps of data acquisition, data preprocessing, data fusion, calibration model establishment, real-time calibration and calibration result evaluation. According to the artificial intelligence toxic gas detection alarm joint calibration method based on multi-sensor data fusion, data of a plurality of toxic gas sensors are integrated through a weighted fusion algorithm, environmental parameters such as temperature, humidity and air pressure are introduced for correlation fusion, and errors and environmental interference of a single sensor are effectively eliminated. For example, sensor weights are dynamically distributed based on historical detection precision and stability, so that data errors after fusion are reduced, and the problem that traditional single sensor calibration cannot cope with complex environments is solved.

Description

Technical Field

[0001] The present invention relates to the technical field of toxic gas detection and calibration, and specifically to a joint calibration method for an artificial intelligence toxic gas detection alarm based on multi-sensor data fusion. Background Art

[0002] With the rapid development of industrial production, safety incidents caused by toxic gas leaks are increasing. Therefore, accurate and timely detection of toxic gases is crucial. As a vital device for ensuring industrial production and personnel safety, the detection accuracy of toxic gas detectors directly impacts the effectiveness of safety protection. However, existing toxic gas detectors are often affected by environmental factors (such as temperature, humidity, and air pressure) and sensor aging during use, leading to deviations in detection data and reduced detection accuracy.

[0003] Currently, the most common calibration method focuses on periodic calibration of individual sensors. This approach is not only inefficient but also fails to fully account for the interactions between multiple sensors and the impact of environmental factors on the overall detection results. Furthermore, traditional calibration methods lack effective utilization and intelligent analysis of historical data, making them difficult to adapt to complex and changing detection environments and unable to accurately calibrate toxic gas detection alarms. Therefore, a new calibration method is urgently needed to improve the detection accuracy and reliability of toxic gas detection alarms. Summary of the Invention

[0004] The purpose of the present invention is to provide an artificial intelligence toxic gas detection alarm joint calibration method based on multi-sensor data fusion. By fusing the data of multiple sensors and using artificial intelligence algorithms for analysis and processing, accurate joint calibration of toxic gas detection alarms can be achieved, thereby improving detection accuracy and reliability and reducing the risk of safety accidents caused by inaccurate toxic gas detection.

[0005] To achieve the above objectives, the present invention provides the following technical solution: a joint calibration method for an artificial intelligence toxic gas detection alarm based on multi-sensor data fusion, comprising the following steps:

[0006] S1: Data acquisition: The data acquisition module collects in real time the toxic gas concentration data detected by multiple toxic gas sensors, as well as environmental parameter data such as temperature, humidity, and air pressure detected by environmental parameter sensors. Different types of toxic gas sensors can detect a variety of toxic gases to obtain more comprehensive detection information.

[0007] S2: Data preprocessing: The collected data is transferred to the data processing module for preprocessing, including removing abnormal data, filling missing data, and normalizing the data. By analyzing the statistical characteristics of the data, such as mean and variance, abnormal data that deviates significantly from the normal range is identified and removed.

[0008] S3: Data fusion: The pre-processed data is transmitted to the artificial intelligence calibration module, and the detection data of multiple toxic gas sensors and environmental parameter data are fused using a weighted fusion algorithm. A corresponding weight is assigned to each sensor based on factors such as its historical detection accuracy and stability, and the weighted average is calculated to obtain the fused toxic gas concentration data.

[0009] S4: Establishing a calibration model: The AI ​​calibration module is based on deep learning algorithms, such as neural network algorithms, and uses a large amount of historical calibration data and fused data to establish a joint calibration model. During the model training process, the error between the model output and the actual calibration results is minimized by adjusting the parameters of the neural network, such as weights and thresholds, to improve the calibration accuracy and generalization ability of the model.

[0010] S5: Real-time calibration: The fused data is input into the established joint calibration model. The model outputs calibration parameters based on the input data, and the toxic gas detection alarm is calibrated in real time. New data is continuously collected to update and optimize the joint calibration model to adapt to the ever-changing detection environment and sensor performance changes.

[0011] S6: Calibration result evaluation: Regularly evaluate the detection accuracy of the calibrated toxic gas detection alarm, calculate the detection error by comparing it with the standard gas concentration, and if the detection error exceeds the preset threshold, recalibrate and adjust and optimize the joint calibration model to ensure that the detection accuracy of the toxic gas detection alarm always meets the requirements.

[0012] Preferably, the outlier elimination in S2 is performed by analyzing statistical characteristics of the data such as the mean and variance to identify and remove data that obviously deviates from the normal range.

[0013] Preferably, in S2, missing data are filled by linear interpolation, and normalization is used to map the data to a specific interval for subsequent data processing and analysis.

[0014] Preferably, the influence of environmental parameters on toxic gas detection is taken into consideration in S3, and the environmental parameter data is associated and fused with the toxic gas concentration data to improve the accuracy and reliability of the data.

[0015] Preferably, in S4, toxic gas sensor detection data, environmental parameter data and corresponding actual calibration result data under different environmental conditions and at different time points are collected, and these data are divided into a training set, a validation set and a test set.

[0016] Preferably, in said S5, the original detection data of the toxic gas sensor is calibrated in real time to trigger an alarm logic, and an alarm is issued if the original detection data is greater than a threshold value.

[0017] Preferably, the detection error evaluation cycle in S6 is once every 24 hours, or the alarm is automatically triggered after the alarm continuously runs for more than a preset time.

[0018] Preferably, if the detection error in S6 exceeds a preset threshold, recalibration is performed, and the joint calibration model is adjusted and optimized to ensure that the detection accuracy of the toxic gas detection alarm always meets the requirements.

[0019] Compared with the prior art, the present invention has the following beneficial effects:

[0020] This AI-powered joint calibration method for toxic gas detection alarms, based on multi-sensor data fusion, integrates data from multiple toxic gas sensors using a weighted fusion algorithm. It also incorporates environmental parameters such as temperature, humidity, and air pressure for correlation and fusion, effectively eliminating single-sensor errors and environmental interference. For example, sensor weights are dynamically assigned based on historical detection accuracy and stability, reducing errors in the fused data. This addresses the inability of traditional single-sensor calibration to handle complex environments.

[0021] 2. This artificial intelligence toxic gas detection alarm joint calibration method based on multi-sensor data fusion uses a neural network-based joint calibration model to automatically learn the mapping relationship between data features and calibration parameters through deep learning of historical calibration data. Compared with traditional manual calibration, the calibration error is reduced after model training, and momentum optimization and time decay factors are used to prioritize adaptation to recent environmental changes, allowing the system to maintain stable detection accuracy when sensors age or the environment changes suddenly.

[0022] 3. This AI-powered joint calibration method for toxic gas detection alarms based on multi-sensor data fusion regularly calculates the absolute error (AE) and relative error (RE) by comparing the concentration of standard gases. When the error exceeds a preset threshold, recalibration is automatically triggered. Combined with a model parameter fine-tuning mechanism, this method ensures that detection accuracy always meets industrial safety requirements. Compared with traditional periodic calibration modes, this method can reduce the frequency of manual intervention while improving calibration efficiency and consistency. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0024] Figure 1 This is a system architecture diagram of the present invention;

[0025] Figure 2 A diagram showing the steps of the method of the present invention. DETAILED DESCRIPTION

[0026] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0027] See also Figure 1-Figure 2 The present invention provides a technical solution: a joint calibration method for an artificial intelligence toxic gas detection alarm based on multi-sensor data fusion, comprising the following steps:

[0028] S1: Data acquisition: The data acquisition module collects in real time the toxic gas concentration data detected by multiple toxic gas sensors, as well as environmental parameter data such as temperature, humidity, and air pressure detected by environmental parameter sensors. Different types of toxic gas sensors can detect a variety of toxic gases to obtain more comprehensive detection information.

[0029] S2: Data preprocessing: The collected data is transferred to the data processing module for preprocessing, including removing abnormal data, filling missing data, and normalizing the data. By analyzing the statistical characteristics of the data, such as mean and variance, abnormal data that deviates significantly from the normal range is identified and removed.

[0030] For missing data, linear interpolation is used to fill in the gaps, and normalization is used to map the data to a specific interval for subsequent data processing and analysis.

[0031] S3: Data fusion: The pre-processed data is transmitted to the artificial intelligence calibration module, and the detection data of multiple toxic gas sensors and environmental parameter data are fused using a weighted fusion algorithm. A corresponding weight is assigned to each sensor based on factors such as its historical detection accuracy and stability, and the weighted average is calculated to obtain the fused toxic gas concentration data.

[0032] Considering the impact of environmental parameters on toxic gas detection, the environmental parameter data is associated and fused with the toxic gas concentration data to improve the accuracy and reliability of the data.

[0033] First, the data of each toxic gas sensor during the historical detection process is collected, and the detection accuracy index and stability index of each sensor are calculated.

[0034] Calculation of detection accuracy index: Assume that the measurement value of a sensor in the i-th detection is x i , the true value is y i , perform n detections. Then the detection accuracy index A of the sensor can be expressed by the inverse of the mean absolute error (MAE), as follows:

[0035] Stability index calculation: Calculate the standard deviation σ of the sensor detection data to measure stability. The formula is: in is the average value of n test data. The stability index S can be expressed as Then, the weight w is assigned to each sensor based on the comprehensive detection accuracy index and stability index. j (j=1,2,…,m), m is the number of toxic gas sensors). Using the normalized method, the formula is: Among them, α and β are weight coefficients, which are adjusted according to actual needs to balance the impact of detection accuracy and stability on the weight.

[0036] Assume that the toxic gas concentration data detected by the jth toxic gas sensor is c j , then the fused toxic gas concentration data c fusion The calculation formula is:

[0037] Assuming that the environmental parameters include temperature T, humidity H, and air pressure P, a correlation model between environmental parameters and toxic gas concentrations can be established. A multiple linear regression model can be used, assuming that the correlation model is: C final =C fusion +γ1T+γ2H+γ3P+ε, where γ1, γ2, and γ3 are regression coefficients obtained through regression analysis of historical data; ε is the error term. This formula is used to correlate and fuse environmental parameter data with toxic gas concentration data to obtain more accurate and reliable data C for calibration. final .

[0038] S4: Establishing a calibration model: The AI ​​calibration module is based on deep learning algorithms, such as neural network algorithms, and uses a large amount of historical calibration data and fused data to establish a joint calibration model. During the model training process, the error between the model output and the actual calibration results is minimized by adjusting the parameters of the neural network, such as weights and thresholds, to improve the calibration accuracy and generalization ability of the model.

[0039] Collect toxic gas sensor detection data, environmental parameter data, and corresponding actual calibration result data under different environmental conditions and at different time points. Divide this data into training set, validation set, and test set. Generally speaking, the training set is used to learn model parameters, the validation set is used to adjust model hyperparameters and prevent overfitting, and the test set is used to evaluate the final performance of the model. Assume that the training set data is represented as in, is the feature vector containing the fused toxic gas concentration data and environmental parameter data, is the corresponding actual calibration result; the validation set data is The test set data is

[0040] Taking the multi-layer perceptron neural network as an example, a network structure is constructed that includes an input layer, several hidden layers, and an output layer. Assume that the number of neurons in the input layer is n in , corresponding to the dimension of the input feature vector X; there are L hidden layers, and the number of neurons in the lth layer (l=1,2,…,L) is n l ; The number of neurons in the output layer is n out , corresponding to the number of calibration parameters.

[0041] In the network, the input layer to the hidden layer, the hidden layer to the hidden layer, and the hidden layer to the output layer are connected by the weight matrix W and the bias vector b. For example, the weight matrix from the l-1 layer to the l layer is W l ∈R nl×nl-1 , the bias vector is b l ∈R nl .

[0042] For the input sample X, the input z at layer l l and output a l The calculation is as follows:

[0043] z l =W l a l-1 +b l , a l =f(z l ), where a 0 =X, f(·) is the activation function, such as ReLU function f(x)=max(0,x), Sigmoid function Etc. Through layer-by-layer calculation, the output of the output layer is finally obtained These are the calibration parameters for the model predictions.

[0044] The loss function L(θ) is used to measure the model prediction results The difference between the actual calibration result y, the commonly used loss function is the mean square error (MSE), the formula is: Where θ represents all the parameters of the neural network (weight W and bias b), and N is the number of samples.

[0045] The gradient of the loss function with respect to each parameter is calculated by the back propagation algorithm to update the parameter. According to the chain rule, the gradient is calculated. and Then the parameters are updated using the stochastic gradient descent optimization algorithm: Where η is the learning rate, which controls the step size of parameter updates.

[0046] S5: Real-time calibration: The fused data is input into the established joint calibration model. The model outputs calibration parameters based on the input data, and the toxic gas detection alarm is calibrated in real time. New data is continuously collected to update and optimize the joint calibration model to adapt to the ever-changing detection environment and sensor performance changes.

[0047] The fused toxic gas concentration data C final (including the environmental parameter association fusion result) input joint calibration model (such as a trained neural network), the model outputs the calibration parameter vector in: It is the offset calibration parameter, which is used to correct the zero point deviation of the sensor; It is the scaling factor calibration parameter, which is used to correct the sensor range deviation.

[0048] The original detection data of the toxic gas sensor C raw Perform real-time calibration, the formula is: Among them, C c is the calibrated toxic gas concentration value, which is used to trigger the alarm logic (if C c >threshold value, an alarm is issued).

[0049] Continuously collect new calibration data x t is the new input feature vector, is the standard calibration result), and incremental learning is used to update the model parameters θ:

[0050] Calculate the gradient of the current data through the loss function (such as mean square error)

[0051] Introducing the momentum coefficient β to accumulate historical gradient, the formula is: (momentum accumulation); (Parameter update), where η is the learning rate, ε is the minimum value to prevent the denominator from being zero, and v t is the current momentum vector.

[0052] The historical gradient is weighted by the time decay factor γ(0,1), giving priority to adapting to recent data: Among them, k is the number of historical time steps traced back. The smaller γ is, the faster the model forgets historical data.

[0053] S6: Calibration result evaluation: Regularly evaluate the detection accuracy of the calibrated toxic gas detection alarm, calculate the detection error by comparing it with the standard gas concentration, and if the detection error exceeds the preset threshold, recalibrate and adjust and optimize the joint calibration model to ensure that the detection accuracy of the toxic gas detection alarm always meets the requirements.

[0054] Every fixed period (such as T hours), use the standard gas concentration C s Test the calibrated alarm to obtain the actual concentration value C test The absolute error (AE) or relative error (RE) is used to measure the accuracy. The formula is: Absolute error AE = |C test -C s |, relative error Preset error threshold (such as absolute error ±5% or relative error ±3%), if

[0055] If AE is greater than the error threshold or RE > the error threshold, a recalibration process is triggered, using the latest data to fine-tune the model and adjust the neural network parameters or weight distribution strategy. If the error continues to exceed the threshold, manual intervention is prompted to check the sensor hardware status.

[0056] It should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprises," "comprising," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or device comprising a series of elements includes not only those elements, but also other elements not explicitly listed, or elements inherent to such process, method, article, or device. In the absence of further limitations, an element defined by the phrase "comprising a..." does not exclude the presence of additional identical elements in the process, method, article, or device comprising the element.

[0057] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A joint calibration method for artificial intelligence toxic gas detection alarm based on multi-sensor data fusion, characterized in that: The following steps are involved: S1: Data acquisition: The data acquisition module collects in real time the toxic gas concentration data detected by multiple toxic gas sensors, as well as environmental parameter data such as temperature, humidity, and air pressure detected by environmental parameter sensors. Different types of toxic gas sensors can detect a variety of toxic gases to obtain more comprehensive detection information. S2: Data preprocessing: The collected data is transferred to the data processing module for preprocessing, including removing abnormal data, filling missing data, and normalizing the data. By analyzing the statistical characteristics of the data, such as mean and variance, abnormal data that deviates significantly from the normal range is identified and removed. S3: Data fusion: The pre-processed data is transmitted to the artificial intelligence calibration module, and the detection data of multiple toxic gas sensors and environmental parameter data are fused using a weighted fusion algorithm. A corresponding weight is assigned to each sensor based on factors such as its historical detection accuracy and stability, and the weighted average is calculated to obtain the fused toxic gas concentration data. S4: Establishing a calibration model: The AI ​​calibration module is based on a neural network algorithm and uses a large amount of historical calibration data and fused data to establish a joint calibration model. During the model training process, the error between the model output and the actual calibration results is minimized by adjusting the neural network parameters, such as weights and thresholds, to improve the model's calibration accuracy and generalization ability. S5: Real-time calibration: The fused data is input into the established joint calibration model. The model outputs calibration parameters based on the input data, and the toxic gas detection alarm is calibrated in real time. New data is continuously collected to update and optimize the joint calibration model to adapt to the ever-changing detection environment and sensor performance changes. S6: Calibration result evaluation: Regularly evaluate the detection accuracy of the calibrated toxic gas detection alarm, and calculate the detection error by comparing it with the standard gas concentration.

2. The method for joint calibration of an artificial intelligence toxic gas detection alarm based on multi-sensor data fusion according to claim 1 is characterized in that: The outlier elimination in S2 is performed by analyzing statistical characteristics of the data such as the mean and variance to identify and remove data that are significantly deviated from the normal range.

3. The method for joint calibration of an artificial intelligence toxic gas detection alarm based on multi-sensor data fusion according to claim 1 is characterized in that: In S2, missing data are filled by linear interpolation, and normalization is used to map the data to a specific interval for subsequent data processing and analysis.

4. The method for joint calibration of an artificial intelligence toxic gas detection alarm based on multi-sensor data fusion according to claim 1 is characterized in that: In S3, the influence of environmental parameters on toxic gas detection is taken into consideration, and the environmental parameter data is associated and fused with the toxic gas concentration data to improve the accuracy and reliability of the data.

5. The method for joint calibration of an artificial intelligence toxic gas detection alarm based on multi-sensor data fusion according to claim 1 is characterized in that: In the S4, toxic gas sensor detection data, environmental parameter data and corresponding actual calibration result data under different environmental conditions and at different time points are collected, and these data are divided into a training set, a validation set and a test set.

6. The method for joint calibration of an artificial intelligence toxic gas detection alarm based on multi-sensor data fusion according to claim 1 is characterized in that: In the above S5, the original detection data of the toxic gas sensor is calibrated in real time to trigger the alarm logic. If the original detection data is greater than the threshold value, an alarm is issued.

7. The method for joint calibration of an artificial intelligence toxic gas detection alarm based on multi-sensor data fusion according to claim 1, characterized in that: The detection error evaluation cycle in S6 is once every 24 hours, or the alarm is automatically triggered after the alarm continuously runs for more than a preset time.

8. The method for joint calibration of an artificial intelligence toxic gas detection alarm based on multi-sensor data fusion according to claim 1, characterized in that: If the detection error exceeds a preset threshold in S6, recalibration is performed, and the joint calibration model is adjusted and optimized to ensure that the detection accuracy of the toxic gas detection alarm always meets the requirements.

Images