An industrial gas leakage automatic detection and alarm device

Abstract

The application discloses an industrial gas leakage automatic detection and alarm device, comprising a drift temperature determination module, a drift amount calculation module and a drift amount regulation module; the drift temperature determination module integrates historical temperature and output characteristics of the alarm device, establishes a temperature-detection concentration curve, determines and verifies the drift temperature range of the alarm device by dividing a window interval, calculating a slope and K-means clustering; the drift amount calculation module compares concentration output data at different temperatures according to the drift temperature range to obtain the drift amount, combines curve data, uses a linear regression algorithm, adds an interaction term, sets a regularization coefficient and the like to establish a drift amount prediction model; the drift amount regulation module collects data in real time to predict the drift amount, verifies the reliability of the drift amount through the temperature range, adjusts the threshold according to the relationship between the detection concentration and the alarm threshold for the reliable drift amount, corrects the influence of the temperature on the detection concentration and improves the detection accuracy.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of gas leakage detection, and particularly relates to an industrial gas leakage automatic detection and alarm device. BACKGROUND

[0002] In the field of industrial gas leakage detection, the detection accuracy of the alarm device is directly related to production safety; however, the existing detection device is easily affected by environmental temperature, resulting in drift of the detection concentration output, and causing false alarm or missed alarm risks.

[0003] Specifically, extreme temperature can make the sensor performance unstable, causing deviation between the detected concentration and the actual concentration, and the traditional method lacks systematic analysis of the temperature drift range and real-time compensation mechanism.

[0004] In the prior art, the temperature influence analysis of the alarm device mainly depends on a single index or simple threshold judgment, and the correlation between historical temperature data and device output characteristics is not fully combined, so it is difficult to accurately identify the temperature sensitive interval.

[0005] At the same time, for the drift amount calculation caused by temperature, there is a lack of data-driven prediction model, which cannot dynamically adapt to complex environmental changes; in addition, in the real-time regulation link, the traditional device mainly adopts a fixed threshold strategy, and does not intelligently adjust the direction, amplitude and continuous trend of the drift amount.

[0006] Therefore, the application provides an industrial gas leakage automatic detection and alarm device. SUMMARY

[0007] In order to make up for the deficiencies of the prior art and solve at least one technical problem proposed in the background art.

[0008] The technical scheme adopted by the application to solve the technical problems is:

[0009] The application provides an industrial gas leakage automatic detection and alarm device, characterized by comprising:

[0010] The drift temperature determination module: according to the historical temperature data of the alarm device and the output characteristics of the alarm device, a temperature-output characteristic table is integrated, and a temperature-detection concentration curve is established according to the temperature-output characteristic table, and the temperature drift range of the alarm device is screened out by using the temperature-detection concentration curve;

[0011] The drift amount calculation module: according to the drift temperature range of the alarm device and the concentration output amount of different alarm devices in the same concentration atmosphere gas at normal temperature, the drift amount of the alarm device is calculated; and according to the calculated drift amount of the alarm device and the temperature-detection concentration curve, a drift amount prediction model is established by using a linear regression algorithm;

[0012] Drift amount regulation module: based on the drift amount prediction model, the drift amount of the alarm device at the current time is predicted in real time, the reliability of the drift amount at the current time is verified, if it is unreliable, an alarm device failure signal is generated, and if it is reliable, the drift amount at the current time is regulated in real time.

[0013] As a further improved scheme of the present application, the specific process of integrating the temperature-output characteristic table is characterized in that:

[0014] The historical working temperature data of the alarm device and the output characteristics of the alarm device at the corresponding time are deeply associated and integrated, each temperature data is matched with the output characteristics at the same time node through data matching technology, and a structured temperature-output characteristic table is formed.

[0015] As a further improved scheme of the present application, the specific process of establishing the temperature-detection concentration curve is characterized in that:

[0016] The temperature-output characteristic table obtained by integration is used to establish a temperature-detection concentration curve with temperature as the abscissa and real-time detection concentration output as the ordinate.

[0017] As a further improved scheme of the present application, the specific process of screening the temperature drift range of the alarm device is characterized in that:

[0018] The wave crest and wave trough in the temperature-detection concentration curve are used as division anchor points to divide the temperature-detection concentration curve into a plurality of continuous and independent window intervals; in each window interval, the temperature-detection concentration curve presents a single monotonicity; that is, strictly monotonically increasing or strictly monotonically decreasing, to ensure the consistency of the data change trend in any window interval; window intervals not in the preset range are removed; and the temperature-detection concentration slope of each window interval is calculated based on the remaining window intervals after removal.

[0019] The K-means clustering algorithm is used to find out and mark the data with large temperature-detection concentration slope jumps, the marked data with large temperature-detection concentration slope jumps are integrated and analyzed to obtain the alarm device drift temperature range.

[0020] As a further improved scheme of the present application, the specific process of integrating and analyzing the alarm device drift temperature range is characterized in that:

[0021] The window interval corresponding to each slope data and the temperature range in the window interval are determined, the temperature intervals corresponding to all the marked temperature-detection concentration jump large slope data are collected and arranged, and the collected temperature intervals are combined, if two temperature intervals have an overlapping part, the two temperature intervals are combined into a larger temperature interval, and after the combination, the remaining temperature interval is the alarm device drift temperature range.

[0022] As a further improved scheme of the present application, the specific process of calculating the alarm device drift amount is characterized in that:

[0023] Based on the temperature-detection concentration curve and the alarm device drift temperature range, the concentration output data corresponding to each moment of the temperature-detection concentration curve in the alarm device drift temperature interval is accurately extracted; the concentration output data of different alarm devices at each moment under the normal working temperature state in the same concentration atmosphere gas environment is collected; the concentration output values in the drift temperature range and the concentration output values under the normal temperature are subjected to difference operation, and the drift amount of the alarm device at different moments is obtained.

[0024] As a further improved scheme of the present application, the specific process of establishing the drift amount prediction model is characterized in that:

[0025] Using a linear regression algorithm, the device drift amount is taken as a drift amount model prediction target, a data matrix is input, the dimension of the data matrix is set to N*6, wherein N is the number of data samples, an output vector y=Δc is set, wherein Δc represents the alarm device drift amount;

[0026] In order to capture the nonlinear coupling effect, a temperature-slope interaction term is added in the data sample; in the optimization process, a regularization coefficient is set for the loss function to suppress the overfitting phenomenon;

[0027] According to the temperature interval, a training set and a verification set are set, key variables are screened through feature significance test, and the drift amount prediction model is trained through the above training process.

[0028] As a further improved scheme of the present application, the process of verifying the reliability of the drift amount is characterized in that:

[0029] The drift temperature value corresponding to the current moment drift amount is obtained, the drift temperature value is compared with the alarm device drift temperature range, and whether the result of the current moment drift amount is reliable is judged;

[0030] If the drift temperature value is in the alarm device drift temperature range, the result of the current moment drift amount output by the drift amount prediction model is reliable;

[0031] If the drift temperature value is not within the alarm device drift temperature range, the result of the current drift amount output by the drift amount prediction model is unreliable.

[0032] As a further improved aspect of the present application, the specific process of generating the alarm device fault signal is characterized in that:

[0033] If the result of the current drift amount is unreliable, the temperature-detection concentration curve corresponding to the current drift amount at the current time is obtained, and the detection output value corresponding to the current drift amount at the current time is found on the temperature-detection concentration curve; and the detection output value corresponding to the current drift amount at the current time is subtracted from the detection output value of the different alarm devices under normal temperature to obtain an output difference value.

[0034] If the output difference value is within the preset output difference value interval, the detection output value corresponding to the current drift amount at the current time is the real output value of the alarm device, and no further processing is performed.

[0035] If the output difference value is not within the preset output difference value interval, an alarm device fault signal is generated.

[0036] As a further improved aspect of the present application, the specific process of real-time regulation and control of the current drift amount is characterized in that:

[0037] If the current drift amount causes the alarm device detection concentration output value to be greater than or equal to the alarm threshold value, and the same direction drift amount is detected for three consecutive times, the alarm threshold value is increased by 1.5 times the current drift amount.

[0038] If the current drift amount causes the alarm device detection concentration output value to be less than the alarm threshold value, and the same direction drift amount is detected for three consecutive times, the alarm threshold value is reduced by 0.5 times the current drift amount.

[0039] The beneficial effects of the present application are as follows:

[0040] 1. By integrating historical temperature data and device output characteristics, a temperature-detection concentration curve is constructed, a monotonic window interval is divided using wave peaks and troughs, and a slope is calculated, and a K-means clustering algorithm is used to screen out temperature sensitive intervals; this method effectively distinguishes the significant influence area of temperature change on detection concentration, avoids false positives caused by short-term noise or small temperature fluctuations, and after historical false positive data verification, can accurately define the effective drift temperature range, reducing the risk of false positives caused by temperature interference from the source.

[0041] 2. A drift amount prediction model is established based on a linear regression algorithm, a nonlinear coupling effect is captured by adding a temperature-slope interaction term, overfitting is inhibited by regularization, and the drift amount is calculated by combining the concentration difference under normal temperature and drift temperature. This model can dynamically learn the complex relationship between temperature and detection concentration, adapt to the individual differences of equipment installed in different positions, provide accurate drift amount prediction for real-time regulation, and solve the problem that traditional methods rely on a single indicator and cannot adapt to complex environmental changes.

[0042] 3. By collecting temperature and concentration data in real time, the reliability of the drift amount is verified and dynamic threshold adjustment is implemented: the alarm threshold is adjusted in proportion to the continuous same direction drift to avoid threshold oscillation caused by short-term fluctuations; the temperature-detection concentration curve is corrected by combining the positive and negative drift amounts, and the influence of temperature on detection concentration is directly compensated; this strategy effectively deals with the performance drift of the sensor under extreme temperature, so that the equipment can still maintain detection accuracy in high or low temperature environment, reduces the risk of false negatives, and improves the safety and stability of industrial gas leakage detection. BRIEF DESCRIPTION OF DRAWINGS

[0043] The application will be further described below with reference to the accompanying drawings.

[0044] Figure 1 is a system module diagram of embodiment 1 of the application;

[0045] Figure 2 is a step flowchart of embodiment 2 of the application. DETAILED DESCRIPTION

[0046] In order to make the technical means, creative features, purposes and effects realized by the application easy to understand, the application will be further described below with reference to the specific embodiments.

[0047] Embodiment 1

[0048] As shown in Figure 1 , the industrial gas leakage automatic detection alarm device described in the embodiment of the application comprises:

[0049] Drift temperature determination module: a temperature-output characteristic table is obtained by combining the historical temperature data of the alarm device with the output characteristics of the alarm device, and a temperature-detection concentration curve is established according to the temperature-output characteristic table, and the temperature drift range of the alarm device is screened out by using the temperature-detection concentration curve;

[0050] Specifically, the historical working temperature data of the alarm device is obtained, which refers to the environmental temperature data of the alarm device when it is working, and a temperature-output characteristic table is obtained by combining the environmental temperature data with the output characteristics of the alarm device;

[0051] The output feature of the alarm device represents the characteristic information output by the alarm device according to the preset collection frequency or monitoring frequency of the alarm device in a monitoring period; more specifically, the output feature includes: real-time detection concentration output, alarm time, alarm level, alarm location;

[0052] The specific process of integrating the ambient temperature data and the output feature of the alarm device to obtain the temperature-output feature table is as follows:

[0053] The historical working temperature data of the alarm device and the output feature of the alarm device at the corresponding time are deeply associated and integrated, each temperature data is matched with the output feature at the same time node through data matching technology, and a structured temperature-output feature table is constructed;

[0054] A temperature-detection concentration curve is established by taking temperature as the abscissa and real-time detection concentration output as the ordinate in the temperature-output feature table obtained by integration;

[0055] The temperature drift range of the alarm device in the temperature-detection concentration curve is screened out;

[0056] The specific process of screening the temperature drift range of the alarm device is as follows:

[0057] Taking the wave crest and wave trough in the temperature-detection concentration curve as the division anchor point, the temperature-detection concentration curve is divided into a plurality of continuous and independent window intervals; in each window interval, the temperature-detection concentration curve presents a single monotonicity; that is, strictly monotonically increasing or strictly monotonically decreasing, to ensure the consistency of the data change trend in any window interval;

[0058] The window intervals not in the preset range are removed;

[0059] It should be noted that the window intervals not in the preset range refer to in some scenarios, the temperature change amplitude in the environment is small, and the concentration detection output of the corresponding alarm device is small, which reflects that the output change of the alarm device is weak and has no reference value;

[0060] Based on the remaining window intervals after removal, the temperature-detection concentration slope of each window interval is calculated;

[0061] It should be noted that the temperature-detection concentration slope refers to the detection concentration change caused by unit temperature change, and in the gas leakage detection scene, a high slope value means that a slight fluctuation of the environment temperature will cause a significant change in the detection concentration reading, and attention should be paid to the false alarm risk caused by temperature interference;

[0062] It needs to be further explained that the detection concentration output of the alarm device is fixed under the ideal ambient temperature state, but in the actual application process, the detection concentration output will be affected by the extreme ambient temperature, resulting in a certain deviation between the detection concentration output and the actual concentration output, that is, there is a certain drift amount;

[0063] Based on the calculated temperature-detection concentration slope of each window interval, the temperature-detection concentration slope is analyzed by using the K-means clustering algorithm. In general, the temperature-detection concentration slope data will be clustered in a main category, and the data with large temperature-detection concentration slope fluctuations will be divided into smaller categories. By using the K-means clustering algorithm, the data with large temperature-detection concentration slope fluctuations are found and marked;

[0064] The marked data with large temperature-detection concentration slope fluctuations are integrated and analyzed to obtain the alarm device drift temperature range;

[0065] The specific process of the integration analysis to obtain the alarm device drift temperature range is as follows:

[0066] The window interval corresponding to each slope data and the temperature range in the window interval are determined. The temperature intervals corresponding to all marked temperature-detection concentration slope data with large fluctuations are collected and sorted, and the collected temperature intervals are merged. If two temperature intervals have overlapping parts, the two temperature intervals are merged into a larger temperature interval;

[0067] After merging, the remaining temperature interval is the alarm device drift temperature range;

[0068] For example, if the temperature interval [50, 60] and the temperature interval [55, 65] are merged to obtain the temperature interval [50, 65], it is considered that when the temperature is in the temperature interval [50, 65], the detection concentration output of the alarm device has a drift phenomenon, and the temperature interval [50, 65] is the alarm device drift temperature range;

[0069] The event data marked as temperature interference false alarm in the historical alarm record are sorted, the marked drift temperature range is spatiotemporally matched with the ambient temperature of the historical false alarm event, the coincidence rate is calculated, and if the coincidence rate is higher than a preset threshold and the overlapping part obeys a normal distribution, it is considered that the alarm device drift temperature range is an effective drift temperature range;

[0070] The drift amount calculation module: according to the alarm device drift temperature range combined with the concentration output of different alarm devices in the same concentration atmosphere gas at normal temperature, the alarm device drift amount is calculated; and according to the calculated alarm device drift amount combined with the temperature-detection concentration curve, a drift amount prediction model is established using a linear regression algorithm;

[0071] Specifically, the specific process of calculating the alarm device drift amount is as follows:

[0072] Based on the temperature-detection concentration curve and the alarm device drift temperature range, the concentration output data corresponding to each time in the temperature-detection concentration curve in the alarm device drift temperature range is accurately extracted; at the same time, the concentration output data of different alarm devices at each time under normal working temperature state in the same concentration atmosphere gas environment is synchronously collected; the concentration output values in the drift temperature range and the concentration output values under normal temperature are subjected to difference operation, thereby obtaining the drift amount of the alarm device at different times;

[0073] It should be particularly noted that the concentration output of different alarm devices under normal temperature in the same concentration atmosphere gas environment is different; this is because in the same system, the installation positions of the alarm devices are different, resulting in a deviation in the actual working environment temperature, and further affecting the concentration output value under normal temperature;

[0074] The calculated alarm device drift amount is combined with the temperature-detection concentration curve to establish a drift amount prediction model;

[0075] The specific establishment process is as follows:

[0076] Using a linear regression algorithm, the device drift amount is used as the drift amount model prediction target, the data matrix is input, the dimension of the data matrix is set to N*6, wherein N is the number of data samples, the output vector y=Δc is set, wherein Δc represents the alarm device drift amount;

[0077] At the same time, in order to capture the nonlinear coupling effect, the temperature-slope interaction term is added in the data sample; in the optimization process, the regularization coefficient is set for the loss function to suppress the overfitting phenomenon;

[0078] According to the temperature interval, the training set and the validation set are set, and the key variables are screened through the feature significance test;

[0079] Using a linear regression algorithm, the drift amount prediction model is trained through the above training process;

[0080] The drift amount regulation module: based on the drift amount prediction model, the drift amount of the alarm device at the current time is predicted in real time, the reliability of the drift amount at the current time is verified, if it is unreliable, an alarm device fault signal is generated, if it is reliable, the drift amount at the current time is regulated in real time;

[0081] The temperature data of the environment where the alarm device is located is acquired in real time by a temperature sensor, the gas concentration data is collected in real time by a detection module of the alarm device, the collected temperature and concentration data are preprocessed in real time, so as to meet the input requirements of the drift amount prediction model;

[0082] Features related to the drift amount are extracted from the preprocessed data, and the extracted feature data is input into the trained drift amount prediction model, and the drift amount prediction model predicts the drift amount at the current time according to the input data;

[0083] Based on the model-predicted drift amount at the current time, the reliability of the drift amount at the current time is verified;

[0084] The specific reliability verification process is as follows:

[0085] The drift temperature value corresponding to the drift amount at the current time is obtained, and the drift temperature value is compared with the alarm device drift temperature range to determine whether the result of the drift amount at the current time is reliable;

[0086] If the drift temperature value is within the alarm device drift temperature range, it means that the drift amount at the current time is caused by the influence of temperature on the detection output of the alarm device, that is, the result of the drift amount at the current time output by the drift amount prediction model is reliable;

[0087] If the drift temperature value is not within the alarm device drift temperature range, it means that the drift amount at the current time is not affected by temperature, and the detection output of the alarm device has not drifted, that is, the result of the drift amount at the current time output by the drift amount prediction model is unreliable;

[0088] If the result of the drift amount at the current time is unreliable, the temperature-detection concentration curve corresponding to the drift amount at the current time is obtained, and the detection output value corresponding to the drift amount at the current time is found on the temperature-detection concentration curve. The difference between the detection output value corresponding to the drift amount at the current time and the detection output value of different alarm devices under normal temperature is obtained to obtain an output difference value. If the output difference value is within a preset output difference value interval, the detection output value of the alarm device is normal, and it is further judged that the alarm device is not affected by temperature. The detection output value corresponding to the drift amount at the current time is the real output value of the alarm device, and no further processing is performed. If the output difference value is not within the preset output difference value interval, an alarm device fault signal is generated;

[0089] It should be noted that the preset output difference value interval is a reference value set by professional technical personnel in the field according to experience, which is used to judge the fault information of the alarm device;

[0090] Based on the result of the current drift amount being reliable, the current drift amount is regulated in real time;

[0091] The drift amount at the current moment is obtained, and the alarm threshold of the alarm device is controlled in real time according to the size of the drift amount.

[0092] The specific adjustment process is as follows:

[0093] If the drift amount at the current moment causes the concentration output value detected by the alarm device to be greater than or equal to the alarm threshold, and the same direction drift amount is detected for three times at a period of 1 minute / time, then the alarm threshold is increased by 1.5 times of the drift amount at the current moment; the adjusted threshold needs to meet ≤ initial threshold * 3, and if it exceeds, it is fixed as 3 times the initial threshold, and a threshold high limit log is recorded and a time stamp is marked;

[0094] If the drift amount at the current moment causes the concentration output value detected by the alarm device to be less than the alarm threshold, and the same direction drift amount is detected for three times at a period of 1 minute / time, then the alarm threshold is reduced by 0.5 times of the drift amount at the current moment; the adjusted threshold needs to meet ≥ initial threshold * 0.2, and if it exceeds, it is fixed as 0.2 times the initial threshold, and a threshold low limit log is recorded and a time stamp is marked;

[0095] After the threshold is adjusted, a reference concentration is generated immediately by calling the built-in standard gas calibration module according to the threshold low limit log or the threshold high limit log, and if the deviation between the detected concentration and the reference concentration is greater than 15% for three times in a row, the original threshold is restored and an adjustment failure alarm is generated, prompting manual troubleshooting of the sensor or environmental abnormalities;

[0096] It should be noted that the same direction drift amount refers to the same positive or negative trend of the detected concentration deviation caused by temperature influence in a plurality of continuous monitoring periods, and the same direction means that the drift amount is positive or negative for three times in a row, indicating that the detected concentration is continuously high or low due to the influence of temperature.

[0097] If the single drift amount fluctuation triggers the threshold adjustment, the threshold will change repeatedly due to short-term noise, causing system instability, and only when the drift amount is consistent in multiple periods can accidental noise be excluded.

[0098] The technical scheme of the embodiment of the application is: a drift temperature determination module obtains historical working temperature data of the alarm device, integrates the output characteristics into a temperature-output characteristic table, and draws a temperature-detection concentration curve; the monotonic window interval is divided according to the curve peak and valley, the invalid interval is removed, the temperature-detection concentration slope of each interval is calculated, the abnormal slope is marked by using K-means clustering, the drift temperature range is determined by integrating the corresponding temperature interval, and the effectiveness is verified by matching the historical temperature interference false alarm data in space and time.

[0099] The drift amount calculation module obtains the drift amount of the alarm device by comparing the concentration output data at the drift temperature range with that at the normal temperature; the drift amount is taken as the prediction target, the temperature-detection concentration curve data are combined, the linear regression algorithm is used for modeling, the temperature-slope interaction term is added to capture the nonlinear effect, the regularization is used to inhibit overfitting, the model training is completed after the training set and the validation set are divided, and the variables are screened through the feature significance test;

[0100] The drift amount regulation module collects the temperature and gas concentration data in real time through the temperature sensor and the detection module, extracts the features after preprocessing, inputs the drift amount prediction model to obtain the drift amount at the current time. The prediction reliability is verified by judging whether the drift temperature value is within the drift temperature range, if not, further comparison of the detection output is made to determine whether the device is malfunctioning; for reliable drift amount, the alarm threshold is adjusted according to the relationship between the detection concentration and the alarm threshold caused by the drift amount, such as increasing the threshold by 1.5 times when the detection concentration is greater than or equal to the threshold, and decreasing the threshold by 0.5 times when the detection concentration is less than the threshold; at the same time, the temperature data in the temperature-detection concentration curve is adjusted according to the positive and negative of the drift amount, so as to correct the influence of temperature on the detection concentration;

[0101] Embodiment 2

[0102] Based on embodiment 1, the application further provides an industrial gas leakage automatic detection and alarm method, comprising the following steps:

[0103] S1: According to the historical temperature data of the alarm device and the output characteristics of the alarm device, a temperature-output characteristic table is integrated, and a temperature-detection concentration curve is established according to the temperature-output characteristic table, and the temperature-detection concentration curve is used to screen out the temperature drift range of the alarm device;

[0104] Specifically, the historical working temperature data of the alarm device is obtained, which is the environmental temperature data of the alarm device during working, and the environmental temperature data is integrated with the output characteristics of the alarm device to obtain a temperature-output characteristic table;

[0105] The output characteristics of the alarm device represent the characteristic information output by the alarm device according to the preset acquisition frequency or monitoring frequency of the alarm device in the monitoring period; more specifically, the output characteristics include: real-time detection concentration output, alarm time, alarm level, alarm position;

[0106] The specific process of integrating the environmental temperature data with the output characteristics of the alarm device to obtain the temperature-output characteristic table is as follows:

[0107] The historical working temperature data of the alarm device and the output characteristics of the alarm device at the corresponding time are deeply associated and integrated, each temperature data is matched with the output characteristics at the same time node through data matching technology, and a structured temperature-output characteristic table is constructed.

[0108] The temperature-output characteristic table obtained by integration is used to establish a temperature-detection concentration curve with temperature as the abscissa and real-time detection concentration output as the ordinate;

[0109] The temperature drift range of the alarm device in the temperature-detection concentration curve is screened out;

[0110] The temperature-detection concentration curve is divided into multiple continuous and independent window intervals using the wave peaks and wave troughs in the temperature-detection concentration curve as division anchor points; in each window interval, the temperature-detection concentration curve presents a single monotonicity; that is, strictly monotonically increasing or strictly monotonically decreasing, ensuring the consistency of the data trend in any window interval;

[0111] The window intervals that are not within the preset range are removed;

[0112] Based on the remaining window intervals after removal, the temperature-detection concentration slope of each window interval is calculated;

[0113] Based on the calculated temperature-detection concentration slope of each window interval, K-means clustering algorithm is used to perform clustering analysis on the temperature-detection concentration slope; in general cases, the temperature-detection concentration slope data will be clustered in a main category, while the temperature-detection concentration slope data with large fluctuations will be divided into smaller categories; by using K-means clustering algorithm, the temperature-detection concentration slope data with large fluctuations are found and marked;

[0114] The temperature-detection concentration slope data with large fluctuations that are marked are integrated and analyzed to obtain the alarm device drift temperature range;

[0115] The specific process of the integration and analysis to obtain the alarm device drift temperature range is as follows:

[0116] The window interval corresponding to each slope data and the temperature range in the window interval are determined, the temperature intervals corresponding to all marked temperature-detection concentration slope data with large fluctuations are collected and sorted, and the collected temperature intervals are merged; if two temperature intervals overlap, the two temperature intervals are merged into a larger temperature interval;

[0117] The event data in the historical alarm records that are explicitly marked as temperature interference false alarms are sorted, the marked drift temperature range is spatiotemporally matched with the environmental temperature at which the historical false alarm events occurred, the coincidence rate is calculated, and if the coincidence rate is higher than a preset threshold and the overlapping part obeys a normal distribution, the alarm device drift temperature range is considered to be an effective drift temperature range;

[0118] S2: Calculate the drift of the alarm device according to the concentration output of the alarm device at normal temperature in the same concentration atmosphere combined with the drift temperature range of the alarm device; and establish a drift prediction model based on the calculated drift of the alarm device combined with the temperature-detection concentration curve using a linear regression algorithm;

[0119] Based on the temperature-detection concentration curve and the drift temperature range of the alarm device, the concentration output data corresponding to each time in the drift temperature range of the alarm device is accurately extracted; at the same time, the concentration output data of different alarm devices at each time under normal working temperature in the same concentration atmosphere is synchronously collected; the concentration output values in the drift temperature range and the concentration output values at normal temperature are subjected to difference operation, thereby obtaining the drift of the alarm device at different times;

[0120] The calculated drift of the alarm device is combined with the temperature-detection concentration curve to establish a drift prediction model;

[0121] Using a linear regression algorithm, the device drift is taken as the drift prediction model target, the data matrix is input, the dimension of the data matrix is set to N x 6, where N is the number of data samples, and the output vector y = Δc is set, where Δc represents the drift of the alarm device;

[0122] At the same time, in order to capture the nonlinear coupling effect, a temperature-slope interaction term is added to the data sample; in the optimization process, a regularization coefficient is set for the loss function to suppress the phenomenon of overfitting;

[0123] According to the temperature interval, the training set and the validation set are set, and the key variables are screened through feature significance test;

[0124] Using a linear regression algorithm, the drift prediction model is trained through the above training process;

[0125] The drift regulation module: based on the drift prediction model, the drift of the alarm device at the current time is predicted in real time, the reliability of the drift at the current time is verified, if it is unreliable, an alarm device fault signal is generated, if it is reliable, the drift at the current time is real-time regulated;

[0126] The temperature data of the environment where the alarm device is located is acquired in real time through the temperature sensor, the gas concentration data is collected in real time using the detection module of the alarm device, and the real-time collected temperature and concentration data are preprocessed to meet the input requirements of the drift prediction model;

[0127] The features related to the drift are extracted from the preprocessed data, the extracted feature data is input into the trained drift prediction model, and the drift prediction model predicts the drift at the current time according to the input data;

[0128] based on the model prediction of the drift amount at the current time, verifying the reliability of the drift amount at the current time;

[0129] obtaining a drift temperature value corresponding to the drift amount at the current time, comparing the drift temperature value with the alarm device drift temperature range, and determining whether the result of the drift amount at the current time is reliable;

[0130] If the drift temperature value is within the alarm device drift temperature range, it means that the drift amount at the current time is based on the influence of temperature, causing the detection output of the alarm device to drift, i.e. the result of the drift amount at the current time predicted by the drift amount prediction model is reliable;

[0131] If the drift temperature value is not within the alarm device drift temperature range, it means that the drift amount at the current time is not affected by temperature, and the detection output of the alarm device has not drifted, i.e. the result of the drift amount at the current time predicted by the drift amount prediction model is unreliable;

[0132] If the result of the drift amount at the current time is unreliable, obtain the temperature-detection concentration curve corresponding to the drift amount at the current time, find the detection output value corresponding to the drift amount at the current time on the temperature-detection concentration curve, and calculate the output difference value by subtracting the detection output value of the alarm device at normal temperature from the detection output value corresponding to the drift amount at the current time. If the output difference value is within the preset output difference value interval, the detection output value of the alarm device is normal, and it is further determined that the alarm device is not affected by temperature. Therefore, the detection output value corresponding to the drift amount at the current time is the real output value of the alarm device, and no further processing is needed. If the output difference value is not within the preset output difference value interval, a fault signal of the alarm device is generated;

[0133] based on the result of the current drift amount being reliable, real-time regulation of the current drift amount is performed;

[0134] obtaining the drift amount at the current time, and real-time regulating the alarm threshold of the alarm device according to the size of the drift amount;

[0135] If the drift amount at the current time causes the detection concentration output value of the alarm device to be greater than or equal to the alarm threshold, and the same direction of drift amount is detected for three consecutive times, the alarm threshold is increased by 1.5 times of the current drift amount;

[0136] If the drift amount at the current time causes the detection concentration output value of the alarm device to be less than the alarm threshold, and the same direction of drift amount is detected for three consecutive times, the alarm threshold is decreased by 0.5 times of the current drift amount;

[0137] If the single drift fluctuation triggers the threshold adjustment, the threshold will change repeatedly due to short-term noise, causing system instability. Only when the drift is consistent in multiple cycles can accidental noise be excluded.

[0138] The above shows and describes the basic principles, main features and advantages of the present application. Those skilled in the art should understand that the present application is not limited to the above embodiments, and the above embodiments and descriptions in the specification are only to illustrate the principles of the present application. Without departing from the spirit and scope of the present application, various changes and improvements can be made to the present application, and these changes and improvements all fall within the scope of the claimed present application. The scope of protection of the present application is defined by the appended claims and their equivalents.

Claims

1. An automatic detection and alarm device for industrial gas leaks, characterized in that: include: Drift temperature determination module: Based on the historical temperature data of the alarm device and the output characteristics of the alarm device, a temperature-output characteristic table is obtained, and a temperature-detection concentration curve is established based on the temperature-output characteristic table. The temperature-detection concentration curve is used to screen out the temperature drift range of the alarm device. The specific process for filtering out the temperature drift range of the alarm device is as follows: Using the peaks and troughs in the temperature-detection concentration curve as anchor points, the temperature-detection concentration curve is divided into multiple continuous and independent window intervals; within each window interval, the temperature-detection concentration curve exhibits a single monotonicity; that is, strictly monotonically increasing or strictly monotonically decreasing, ensuring the consistency of the data change trend within any window interval. Remove window ranges that are not within the preset range; Based on the remaining window intervals after rejection, calculate the temperature-detection concentration slope for each window interval; The K-means clustering algorithm is used to find data with large fluctuations in the temperature-detection concentration slope and these data are then labeled. The labeled data with large fluctuations in the temperature-detection concentration slope are then integrated and analyzed to obtain the drift temperature range of the alarm device. Drift Calculation Module: Based on the drift temperature range of the alarm device and the concentration output of different alarm devices in the same concentration atmosphere gas at normal temperature, the drift of the alarm device is calculated; and based on the calculated drift of the alarm device and the temperature-detection concentration curve, a drift prediction model is established using a linear regression algorithm. The specific process for establishing the drift prediction model is as follows: Using a linear regression algorithm, the equipment drift is taken as the prediction target of the drift model. The input data matrix is ​​set to have a dimension of 1. Where N is the number of data samples, and the output vector is set. ,in, Indicates the amount of drift of the alarm device; To capture the nonlinear coupling effect, a temperature-slope interaction term was added to the data samples; during the optimization process, a regularization coefficient was set for the loss function to suppress overfitting. Training and validation sets are set according to temperature ranges, and key variables are screened through feature significance testing. The drift prediction model is obtained through the above training process. Drift control module: Based on the drift prediction model, it predicts the drift of the alarm device in real time, verifies the reliability of the drift at the current moment, generates a fault signal for the alarm device if it is unreliable, and adjusts the drift at the current moment in real time if it is reliable.

2. The automatic detection and alarm device for industrial gas leaks according to claim 1, characterized in that: The specific process for integrating the temperature-output characteristic table is as follows: The historical operating temperature data of the alarm device is deeply correlated and integrated with the output characteristics of the alarm device at the corresponding time. Through data matching technology, each temperature data is matched one-to-one with the output characteristics at the same time point to construct a structured temperature-output characteristic table.

3. The automatic detection and alarm device for industrial gas leaks according to claim 1, characterized in that: The specific process for establishing the temperature-concentration detection curve is as follows: In the integrated temperature-output feature table, a temperature-detection concentration curve is established with temperature as the horizontal axis and real-time detection concentration output as the vertical axis.

4. The automatic detection and alarm device for industrial gas leaks according to claim 1, characterized in that: The specific process for obtaining the drift temperature range of the alarm device through integrated analysis is as follows: Define the window interval corresponding to each slope data and the temperature range within that window interval. Collect and organize the temperature intervals corresponding to all marked temperature-detection concentration jump slope data, and merge the collected temperature intervals. If two temperature intervals overlap, merge them into a larger temperature interval. After merging, the remaining temperature interval is the drift temperature range of the alarm device.

5. The automatic detection and alarm device for industrial gas leakage according to claim 1, characterized in that: The specific process for calculating the drift amount of the alarm device is as follows: Based on the temperature-detection concentration curve and the drift temperature range of the alarm device, the concentration output data corresponding to each moment of the temperature-detection concentration curve within the drift temperature range of the alarm device is accurately extracted; the concentration output data of different alarm devices at each moment under normal operating temperature conditions are collected under the same concentration atmosphere gas environment; the difference between the concentration output value within the drift temperature range and the concentration output value at normal temperature is calculated to obtain the drift amount of the alarm device at different moments.

6. The automatic detection and alarm device for industrial gas leaks according to claim 1, characterized in that: The process of verifying the reliability of the drift is as follows: Obtain the drift temperature value corresponding to the drift amount at the current moment, compare the drift temperature value with the drift temperature range of the alarm device, and determine whether the result of the drift amount at the current moment is reliable. If the drift temperature value is within the drift temperature range of the alarm device, then the drift amount prediction model outputting the drift amount at the current moment is reliable. If the drift temperature value is not within the drift temperature range of the alarm device, the result of the current drift amount output by the drift amount prediction model is unreliable.

7. The automatic detection and alarm device for industrial gas leaks according to claim 1, characterized in that: The specific process for generating alarm device fault signals is as follows: If the current drift result is unreliable, obtain the temperature-detection concentration curve corresponding to the drift at the current moment, and find the detection output value corresponding to the drift at the current moment in the temperature-detection concentration curve. The output difference is obtained by subtracting the current drift value from the output value of different alarm devices at normal temperature. If the output difference is within the preset output difference range, the detection output value corresponding to the drift amount at the current moment is the actual output value of the alarm device, and no processing is performed. If the output difference is not within the preset output difference range, an alarm device fault signal will be generated.

8. The automatic detection and alarm device for industrial gas leaks according to claim 1, characterized in that: The specific process of real-time adjustment of the drift amount at the current moment is as follows: If the drift amount at the current moment causes the alarm device to detect a concentration output value greater than or equal to the alarm threshold, and drift amounts in the same direction are detected three times consecutively, then the alarm threshold will be increased by 1.5 times the drift amount at the current moment. If the drift amount at the current moment causes the alarm device's detected concentration output value to be less than the alarm threshold, and drift amounts in the same direction are detected three times consecutively, then the alarm threshold will be reduced by 0.5 times the drift amount at the current moment.

Images