A laboratory gas monitoring method based on internet of things data acquisition

Abstract

The application discloses a laboratory gas monitoring method based on Internet of Things data collection and relates to the technical field of laboratory gas monitoring.The method comprises the following steps: collecting a gas sampling frame according to an Internet of Things collection task, checking integrity and aligning a sampling time, generating an aligned gas sequence, positioning a calibration interval according to an environment sampling value of the aligned gas sequence, interpolating a drift amount of adjacent calibration points, deducting an environment drift amount of an original gas response value, and generating an environment calibration sequence; establishing a fast drift window and a slow drift window around the environment calibration sequence, calculating a calibration residual error according to a median value in the fast drift window, extracting a residual fast drift amount according to a synchronous relationship between the environment sampling value and the calibration residual error, eliminating a concentration rising section in the slow drift window, extracting an aging slow drift amount along a residual concentration median track, deducting the residual fast drift amount and the aging slow drift amount point by point, and freezing the aging slow drift amount correction when the concentration continuously rises, and generating a double-scale concentration sequence; and the application improves target gas recognition reliability and alarm directivity.

Description

Technical Field

[0001] This invention relates to the field of laboratory gas monitoring technology, and in particular to a laboratory gas monitoring method based on Internet of Things (IoT) data acquisition. Background Technology

[0002] With the development of IoT communication, low-power microprocessors, and gas-sensitive devices, laboratory gas monitoring has gradually evolved from independent audible and visual alarms to intelligent sensing systems. Monitoring nodes typically collect gas concentrations and environmental parameters such as temperature, humidity, and air pressure at fixed intervals, and upload them to the monitoring platform via wireless links. The node and platform ends then compare the sampled values ​​with set thresholds to complete over-limit alarms, ventilation linkage, gas source cutoff, and historical records. Some solutions also incorporate calibration data to compensate for environmental impacts, thus forming a continuous monitoring technology system for long-term unattended scenarios.

[0003] Existing alarm solutions primarily rely on static concentration responses under fixed operating conditions. When gas-sensitive devices operate for extended periods and are subjected to changes in temperature and humidity, residual environmental drift, device aging, and cross-response from non-target gases, even after environmental calibration and threshold correction, the calibrated readings may still contain drift components at different time scales. Furthermore, different gases may produce similar response amplitudes under a single operating condition. Especially in scenarios where solvent evaporation, air conditioning mode switching, and slow accumulation of target gases coexist in the laboratory, monitoring nodes struggle to identify the true response location of the target gas from the concentration increase. This is because static amplitudes fail to characterize the phased response differences and recovery features of gas-sensitive devices as operating conditions change, thus lacking a stable basis for subsequent concentration extraction and alarm confirmation. Summary of the Invention

[0004] In view of the aforementioned existing problems, the present invention is proposed.

[0005] Therefore, this invention provides a laboratory gas monitoring method based on Internet of Things (IoT) data acquisition to solve the problem of difficulty in accurately identifying target gases and reliably alarming under drift and cross-interference conditions.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution: This invention provides a laboratory gas monitoring method based on Internet of Things (IoT) data acquisition, comprising: acquiring gas sampling frames according to IoT acquisition tasks, verifying integrity and aligning sampling times, generating an aligned gas sequence, locating a calibration interval using environmental sampling values ​​of the aligned gas sequence, interpolating the drift of adjacent calibration points, subtracting the environmental drift of the original gas response value, and generating an environmental calibration sequence. Fast drift window and slow drift window are established around the environmental calibration sequence. The calibration residual is calculated using the median value of the fast drift window. The residual fast drift amount is extracted according to the synchronization relationship between the environmental sampling value and the calibration residual. The concentration rise segment is removed in the slow drift window. The aging slow drift amount is extracted along the median trajectory of the remaining concentration. The residual fast drift amount and the aging slow drift amount are deducted point by point. The aging slow drift amount is frozen and corrected when the concentration continues to rise, generating a dual-scale concentration sequence. The short-term increment and long-term trend of the dual-scale concentration sequence are verified in the same direction to determine the excitation trigger state. When the excitation trigger state is established, the gas-sensitive device is controlled to switch the working conditions along the stepped excitation trajectory. The transient response under each working condition is collected. The excitation stage is divided according to the switching time and the time axis is registered. The stage response difference is determined from the initial response and the stable response of each excitation stage. The recovery ratio is determined from the response drop in the recovery stage and the response difference of the previous excitation stage, thus forming a gas dynamic fingerprint. Based on the target gas response location marked by the gas dynamic fingerprint, the corresponding concentration value in the dual-scale concentration sequence is extracted, and the alarm threshold and confirmation time are determined according to the concentration change status to generate a gas monitoring record.

[0007] As a preferred embodiment of the laboratory gas monitoring method based on IoT data acquisition described in this invention, the specific steps for generating the aligned gas sequence are as follows: Verify the structural integrity and frame sequence continuity of the gas sampling frames, remove abnormal frames and mark missing positions, and generate valid gas sampling frames; The local sampling time of the effective gas sampling frame is corrected according to the unified sampling period and matched to the task sampling time within the allowable deviation to generate an aligned gas sequence.

[0008] As a preferred embodiment of the laboratory gas monitoring method based on Internet of Things data acquisition described in this invention, the specific steps for generating the environmental calibration sequence are as follows: Environmental sampling values ​​are extracted one by one from the aligned gas sequence. The environmental sampling values ​​are compared with the corresponding calibration nodes to determine the upper and lower boundary nodes. When the current environmental sampling value is equal to the maximum calibration node, the calibration interval is defined by the maximum calibration node and the previous calibration node. When the boundary is exceeded, the environmental sampling value is clamped to the nearest boundary node to generate a calibration interval record. Based on the calibration interval records, adjacent calibration points are extracted, the distance ratio between the environmental sampling values ​​and the upper and lower boundary nodes is calculated, and boundary constraints are applied to the distance ratio exceeding the boundary to generate the environmental interpolation ratio. The drift of adjacent calibration points is linearly interpolated according to the environmental interpolation ratio to determine the environmental drift. The environmental drift is then subtracted from the original gas response values ​​in the aligned gas sequence and arranged according to the task sampling time to generate an environmental calibration sequence.

[0009] As a preferred embodiment of the laboratory gas monitoring method based on IoT data acquisition described in this invention, the steps of establishing a fast drift window and a slow drift window around the environmental calibration sequence, calculating the calibration residual using the median value of the fast drift window, and extracting the residual fast drift amount according to the synchronization relationship between the environmental sampling value and the calibration residual are as follows: Using the current sampling position of the environmental calibration sequence as the end of the window, continuous sampling positions are truncated forward according to the short-time response time of the environment and the aging tracking time of the gas-sensitive device, respectively. The window is moved synchronously with the task sampling time and invalid sampling positions are excluded to generate the current fast drift window and the current slow drift window. The median value of the environmental calibration values ​​within the fast drift window is calculated. The calibration residual is determined by the deviation of each environmental calibration value from the median value. The environmental sampling values ​​are moved according to the candidate response delay and the synchronous change positions are screened. After the abnormal positions are removed, the fast drift response ratio is determined. The environmental sampling change amount corresponding to the fast drift response ratio is scaled and limited by the calibration residual amplitude to generate the residual fast drift amount sequence.

[0010] As a preferred embodiment of the laboratory gas monitoring method based on IoT data acquisition described in this invention, the steps of removing concentration-increasing segments within the slow drift window, extracting the aged slow drift amount along the median trajectory of the remaining concentration, deducting the residual fast drift amount and the aged slow drift amount point by point, and freezing the aged slow drift amount for correction when the concentration continues to rise, thereby generating a dual-scale concentration sequence, are as follows: The residual fast drift amount is deducted point by point from the environmental calibration value within the slow drift window. The concentration rise segment is marked according to the case where the adjacent concentration increment exceeds the normal fluctuation boundary and continuously covers the response time of the gas-sensitive device. After removing the concentration rise segment, the median value of the remaining concentration is calculated in segments, and the adjacent median values ​​are connected in time sequence to generate the median trajectory of the remaining concentration. The adjacent offsets that do not exceed the aging change range are extracted along the median trajectory of the remaining concentration and accumulated sequentially to generate an aging slow drift sequence. When the concentration enters a state of continuous increase at the current sampling position, the aging slow drift before the starting point of the rise is fixed, and the freeze is lifted after the subsequent complete fast drift window recovers stability. According to the sampling time of the task, the residual fast drift and aging slow drift are deducted point by point from the environmental calibration sequence. For positions where the concentration continues to rise, the aging slow drift is frozen and the positive and negative change directions of the calibration results are retained to generate a dual-scale concentration sequence.

[0011] As a preferred embodiment of the laboratory gas monitoring method based on Internet of Things data acquisition described in this invention, the specific steps for verifying the short-term increment and long-term trend of the dual-scale concentration sequence in the same direction to determine the excitation trigger state are as follows: Using the current sampling position of the dual-scale concentration sequence as the end, the current short-time segment and the short-time reference segment are truncated, invalid sampling positions are removed and isolated jump values ​​are replaced, the difference between the median concentration values ​​of the two segments is calculated, and the short-time increment is generated. Long-term trend segments covering continuous short time intervals are extracted from dual-scale concentration sequences. Linear fitting is performed on the effective concentration values. Based on the fitting change direction and the long-term trend boundary, the corresponding rising, stable and falling states are marked to generate a long-term trend. Verify the direction of change of short-term increment and long-term trend. When the short-term increment exceeds the short-term increment boundary and the long-term trend continues to rise to reach the minimum effective response time of the gas-sensitive device, determine that the excitation trigger state is established and generate a stepped excitation trigger record.

[0012] As a preferred embodiment of the laboratory gas monitoring method based on IoT data acquisition described in this invention, when the excitation trigger state is established, the gas-sensitive device is controlled to switch operating conditions along a stepped excitation trajectory, and the transient response under each operating condition is collected. The excitation stages are divided according to the switching time and time axis registration is performed. The specific steps are as follows: Based on the stepped excitation trigger record, the reference response before stepped excitation is determined, the gas-sensitive device is controlled to switch the heating temperature step by step along the stepped excitation trajectory, and the temperature stabilization time is recorded after the actual heating temperature enters the target temperature allowable deviation range and remains stable, thus generating a working condition switching record. Based on the working condition switching record, gas-sensitive response values ​​are continuously collected starting from the first working condition switching moment. The excitation stage is divided by adjacent switching moments. The response range after returning to normal recovery temperature is included in the recovery stage. The thermal transition zone at the beginning of each stage is marked to generate the excitation stage. Using the temperature stabilization moment of each excitation stage as the stage zero moment, the gas-sensitive response value is converted to the relative sampling moment and registered according to a uniform sampling interval. For positions that exceed the allowable interpolation time, a missing response mark is retained to generate the time axis registration excitation stage.

[0013] As a preferred embodiment of the laboratory gas monitoring method based on Internet of Things data acquisition described in this invention, the steps of determining the stage response difference between the initial response and the stable response of each excitation stage, and determining the recovery ratio between the response drop in the recovery stage and the response difference of the previous excitation stage to form a gas dynamic fingerprint are as follows: The initial response is extracted from the thermal transition zone after the time axis registration excitation stage, and the stable response is extracted from the response range that continuously satisfies the stable fluctuation condition at the end of the stage. The stage response difference is determined by the change of the stable response relative to the initial response. Extract the response fall-off from the recovery phase, determine the recovery ratio when the response difference of the previous excitation phase exceeds the minimum effective response boundary, arrange the phase response differences and associate them with the recovery ratio according to the working conditions, retain the corresponding marks for invalid phases and unrecovered positions, and form a gas dynamic fingerprint.

[0014] As a preferred embodiment of the laboratory gas monitoring method based on Internet of Things data acquisition described in this invention, the specific steps of marking the target gas response position according to the gas dynamic fingerprint are as follows: matching the gas dynamic fingerprint with the target gas reference fingerprint position by position according to the fixed fingerprint position; confirming the target gas dynamic fingerprint when the stage response difference and recovery ratio fall within the corresponding reference range; and finding the stable segment forward according to the step excitation trigger record and extending along the concentration change direction to obtain the target gas response position.

[0015] As a preferred embodiment of the laboratory gas monitoring method based on IoT data acquisition described in this invention, the steps of extracting corresponding concentration values ​​from the dual-scale concentration sequence, determining the alarm threshold and confirmation duration based on the concentration change status, and generating a gas monitoring record are as follows: Based on the target gas response location, the task sampling time in the dual-scale concentration sequence is matched, the concentration extraction isolation section corresponding to the step excitation is excluded, the dual-scale concentration values ​​are extracted and arranged according to the task sampling time to generate the target gas concentration trajectory. The effective alarm threshold and confirmation duration are determined based on the change status of the target gas concentration trajectory. The effective over-limit period is accumulated from the first over-limit location. When the accumulated period reaches the confirmation duration, the alarm is confirmed, and the target gas response location, concentration value and alarm result are associated to generate a gas monitoring record.

[0016] The beneficial effects of this invention are as follows: By controlling the gas-sensitive device to switch the heating temperature along the stepped excitation trajectory after the excitation trigger state is established, the excitation stage is divided and registered according to the temperature stabilization time, and the initial response, stable response, stage response difference and recovery ratio are extracted to form a gas dynamic fingerprint that characterizes the response law of the target gas. This enables the intelligent sensing system to verify the identity of the source of the concentration rise, and limits the target gas response position and concentration trajectory accordingly, preventing drift components and non-target gas responses from directly entering the alarm judgment, thereby improving the reliability of target gas identification and the directionality of alarm. Attached Figure Description

[0017] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1This is a flowchart of a laboratory gas monitoring method based on Internet of Things (IoT) data acquisition.

[0019] Figure 2 This is a flowchart for gas sampling verification and environmental calibration.

[0020] Figure 3 This is a flowchart for dual-scale drift compensation and concentration sequence generation.

[0021] Figure 4 This is a flowchart of the stepped excitation triggering and gas dynamic fingerprint formation.

[0022] Figure 5 The time-axis registration response curves for different gases under stepped excitation are shown.

[0023] Figure 6 The graph shows the response decline curves for different gas recovery stages. Detailed Implementation

[0024] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0025] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0026] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0027] Reference Figures 1-6 This is one embodiment of the present invention, which provides a laboratory gas monitoring method based on Internet of Things (IoT) data acquisition, comprising the following steps: S1. Collect gas sampling frames according to the IoT acquisition task, verify the integrity and align the sampling time, generate an aligned gas sequence, locate the calibration interval with the environmental sampling value of the aligned gas sequence, interpolate the drift of adjacent calibration points, subtract the environmental drift of the original gas response value, and generate an environmental calibration sequence.

[0028] It should also be noted that, according to the IoT data acquisition task, the gas-sensitive devices that need to be monitored in the laboratory and their corresponding sampling locations are determined. When the task sampling time arrives, the original gas response value output by the gas-sensitive device is collected. At the same time, the environmental sampling value, local sampling time, task sampling time, sampling location identifier, frame number and acquisition status mark corresponding to the original gas response value are recorded. Records formed at the same sampling location and at the same task sampling time are encapsulated into a gas sampling frame.

[0029] S1.1 Verify the structural integrity and frame sequence continuity of the gas sampling frames, remove abnormal frames and mark missing positions, and generate valid gas sampling frames.

[0030] Furthermore, the gas sampling frames corresponding to the same sampling location are arranged according to the frame sequence number. The specified fields are checked to see if they are complete and if the format is consistent. If a field is missing or the format is abnormal, an abnormal frame is marked. If the frame sequence number is skipped, a missing position is marked. If the frame sequence number is reversed, it is rearranged according to the local sampling time. If the records corresponding to the same frame sequence number are inconsistent, a conflicting frame is marked. The gas sampling frames that have passed the verification are collected to generate a valid gas sampling frame.

[0031] S1.2 Correct the local sampling time of the effective gas sampling frame according to the unified sampling period, and match it to the task sampling time within the allowable deviation to generate an aligned gas sequence.

[0032] Furthermore, the local sampling time of the effective gas sampling frame is corrected with a unified sampling period, and the corrected local sampling time is matched with the task sampling time whose time difference falls within the allowable deviation range. When the matching fails, the missing mark is retained. When multiple frames correspond to the same task sampling time, the effective gas sampling frame with the smallest time difference is retained. After arranging according to the task sampling time, an aligned gas sequence is generated.

[0033] It should also be noted that the unified sampling period is obtained through the sampling period field in the IoT data acquisition task.

[0034] S1.3 Extract environmental sampling values ​​from the aligned gas sequence one by one, compare the environmental sampling values ​​with the corresponding calibration nodes, determine the upper and lower boundary nodes, when the current environmental sampling value is equal to the maximum calibration node, limit the calibration interval with the maximum calibration node and the previous calibration node, when the boundary is exceeded, clamp the environmental sampling value to the nearest boundary node, and generate a calibration interval record.

[0035] Furthermore, environmental sampling values ​​are used sequentially from the aligned gas sequence according to the task sampling time. Each environmental sampling value is limited to a single environmental parameter sampled by the IoT data acquisition task, such as ambient temperature. When the IoT data acquisition task records multiple environmental parameters, calibration intervals are located separately according to the environmental parameter identifiers, and environmental sampling values ​​of different dimensions are not merged. The environmental sampling values ​​are compared with the corresponding calibration nodes in ascending order. When an environmental sampling value is between adjacent calibration nodes, the calibration node with the closest value and not higher than the environmental sampling value is used as the lower boundary node, and the calibration node with the closest value and not lower than the environmental sampling value is used as the upper boundary node. When the environmental sample value is equal to the maximum calibration node, the calibration interval is defined by the previous calibration node and the maximum calibration node. When the environmental sample value is lower than the minimum calibration node, it is clamped to the minimum calibration node, and when it is higher than the maximum calibration node, it is clamped to the maximum calibration node. When there are fewer than two calibration nodes, the calibration node order is abnormal, or the environmental parameter identifiers are inconsistent, the abnormal marker is retained and the interpolation process is stopped. The environmental sample value, upper and lower boundary nodes, and boundary states are associated with the task sampling time to generate a calibration interval record.

[0036] It should also be noted that the environmental sampling values ​​were collected synchronously by laboratory environmental sensors at the time of the task sampling.

[0037] Environmental parameter identifiers are determined through environmental monitoring projects in IoT data collection tasks.

[0038] The calibration nodes are determined by the calibration positions corresponding to different environmental parameter values ​​in the calibration test records of the gas-sensitive device.

[0039] S1.4 Extract adjacent calibration points based on the calibration interval records, calculate the distance ratio between the upper and lower boundary nodes of the environmental sampling values, and apply boundary constraints to the distance ratios that exceed the boundaries to generate the environmental interpolation ratio.

[0040] Furthermore, the lower and upper boundary nodes corresponding to the same task sampling time in the calibration interval record are taken as adjacent calibration points. The distance from the environmental sampling value to the lower boundary node is represented by the difference between the environmental sampling value after boundary constraints and the value of the lower boundary node. The total distance between the upper and lower boundary nodes is represented by the difference between the values ​​of the upper and lower boundary nodes. The proportion of the aforementioned distance to the total distance is determined as the distance ratio, expressed as: ; in: It is the first The distance ratio corresponding to each task sampling time; It is the first Environmental sample values ​​after boundary constraints at each task sampling time; It is the first The value of the lower boundary node corresponding to the sampling time of each task; It is the first The values ​​of the upper boundary nodes corresponding to the sampling time of each task; It is the sequence number of the task sampling time; When the total distance between the upper and lower boundary nodes is zero, stop determining the distance ratio at the current task sampling time and retain the abnormal environmental interpolation ratio flag; when the distance ratio is lower than zero, limit the distance ratio to zero; when the distance ratio is higher than one, limit the distance ratio to one; when the distance ratio is between zero and one, retain the original distance ratio. Then, associate the distance ratio that completes the boundary constraints with the corresponding task sampling time to generate the environmental interpolation ratio.

[0041] It should also be noted that, The first one is obtained from the aligned gas sequence. The environmental sampling values ​​at each task sampling time are determined after being clamped at the calibrated node boundaries.

[0042] Select records that are no higher than the specified range from the calibration interval. And the value is closest It was determined after the calibration nodes were established.

[0043] Select records of no less than the specified range from the calibration interval. And the value is closest It was determined after the calibration nodes were established.

[0044] S1.5. Based on the environmental interpolation ratio, perform linear interpolation on the drift of adjacent calibration points to determine the environmental drift. Subtract the environmental drift from the original gas response values ​​in the aligned gas sequence and arrange them according to the task sampling time to generate an environmental calibration sequence.

[0045] Furthermore, the environmental interpolation ratio is associated with adjacent calibration points corresponding to the sampling time of the same task. The difference between the upper boundary node drift and the lower boundary node drift is allocated according to the environmental interpolation ratio, and the allocation result is superimposed on the lower boundary node drift to form the environmental drift. When the environmental interpolation ratio is zero, the lower boundary node drift is used; when the environmental interpolation ratio is one, the upper boundary node drift is used; when the environmental interpolation ratio has an anomaly marker, no environmental drift is formed and the calibration anomaly marker is retained.

[0046] Subsequently, the environmental drift is matched to the original gas response value according to the task sampling time. When the match is consistent, the environmental drift is subtracted from the original gas response value to form the environmental calibration response value; when the match is inconsistent, the original gas response value is retained and a calibration anomaly mark is added. The gas identifier, gas-sensitive device identifier, and normal heating temperature corresponding to the environmental calibration response value are matched with the gas-sensitive device calibration test record to extract the response concentration calibration relationship under consistent conditions. Adjacent response calibration points are selected within the response calibration interval corresponding to the environmental calibration response value. According to the position ratio of the environmental calibration response value between adjacent response calibration points, the concentration calibration values ​​corresponding to adjacent response calibration points are linearly interpolated to form the environmental calibration concentration value. When the environmental calibration response value is equal to the response calibration boundary, the corresponding boundary concentration calibration value is used. When the environmental calibration response value exceeds the response calibration range, the matching conditions are inconsistent, or the response concentration calibration relationship is missing, a concentration conversion anomaly mark is retained, and no valid environmental calibration concentration value is formed. The environmental calibration concentration values ​​are arranged according to the task sampling time to generate an environmental calibration sequence.

[0047] It should also be noted that the response concentration calibration relationship is formed by contacting the gas-sensitive device with target gases of different known concentrations at normal heating temperature, recording the stable gas response values ​​and corresponding concentration calibration values, and associating them according to the order of the stable gas response values.

[0048] The upper boundary node drift is the response offset formed by extracting the stable gas response value corresponding to the upper boundary node from the calibration test record of the gas-sensitive device and comparing it with the stable response value of the reference environment at the same gas concentration, the same gas-sensitive device identification, and the same heating temperature.

[0049] The lower boundary node drift is the response offset formed by extracting the stable gas response value corresponding to the lower boundary node from the calibration test record of the gas-sensitive device and comparing it with the stable response value of the reference environment under the same gas concentration, the same gas-sensitive device identification, and the same heating temperature.

[0050] S2. Establish fast drift windows and slow drift windows around the environmental calibration sequence. Calculate the calibration residual using the median value of the fast drift window. Extract the residual fast drift amount according to the synchronization relationship between the environmental sampling value and the calibration residual. Remove the concentration increase segment within the slow drift window. Extract the aging slow drift amount along the median trajectory of the remaining concentration. Deduct the residual fast drift amount and the aging slow drift amount point by point. Freeze the aging slow drift amount correction when the concentration continues to rise to generate a dual-scale concentration sequence.

[0051] S2.1. Using the current sampling position of the environmental calibration sequence as the end of the window, continuously sampled positions are extracted forward according to the short-time response time of the environment and the aging tracking time of the gas-sensitive device, respectively. The window is moved synchronously with the task sampling time and invalid sampling positions are excluded to generate the current fast drift window and the current slow drift window.

[0052] Furthermore, using the current sampling position of the environmental calibration sequence as the end of the window, the environmental calibration concentration value is truncated forward according to the number of task sampling moments covered by the short-term environmental response time, forming the current fast drift window; at the same time, the environmental calibration concentration value is truncated forward according to the number of task sampling moments covered by the aging tracking time of the gas-sensitive device, forming the current slow drift window.

[0053] As the current sampling position moves backward along the task sampling time, the current fast drift window and the current slow drift window move synchronously. Sampling positions with missing markers, calibration abnormal markers, and invalid acquisition status markers are not included in the window statistics. When the window's starting position has not yet covered the specified duration, the window incompleteness marker is retained. After the number of valid sampling positions within the window reaches the required number for the corresponding duration, the current fast drift window and the current slow drift window are generated.

[0054] It should also be noted that the short-term environmental response time is determined by the duration during which the environmental calibration concentration value completes a short-term follow-up change after the environmental sample value changes; the value is determined based on the duration during which the environmental calibration concentration value completes a short-term follow-up change after the environmental sample value changes, with an example value of 60s.

[0055] The aging tracking time of the gas-sensitive device is determined by the shortest period of time in the long-term stable operation record of the gas-sensitive device that can continuously reflect slow drift changes; the example value is 24h.

[0056] The current fast drift window is a set of environmental calibration concentration values ​​that covers a short-term response time of the environment and retains only the valid sampling locations, with the current sampling location as the end point. It is used to characterize the response fluctuations in a short period of time and to provide a sampling range for determining the median value, calibration residual, and residual fast drift amount of the fast drift window.

[0057] The current slow drift window is a set of environmental calibration concentration values ​​that ends at the current sampling position, covers the aging tracking time of the gas-sensitive device, and retains only the effective sampling positions. It is used to observe slow changes over a longer period of time and to provide a sampling interval for eliminating concentration rise segments, forming the median trajectory of the remaining concentration, and extracting the aging slow drift amount.

[0058] S2.2 Calculate the median value of the environmental calibration concentration values ​​within the fast drift window. Determine the calibration residual based on the deviation of each environmental calibration concentration value from the median value. Move the environmental sampling values ​​according to the candidate response delay and screen the synchronous change positions. After removing abnormal positions, determine the fast drift response ratio. Scale the environmental sampling change amount corresponding to the fast drift response ratio and limit the amplitude with the calibration residual value to generate the residual fast drift amount sequence.

[0059] Furthermore, the environmental calibration concentration values ​​within the current fast drift window are arranged in numerical order, with the environmental calibration concentration value in the middle position serving as the median value of the fast drift window. Then, the median value of the fast drift window is subtracted from each environmental calibration concentration value within the current fast drift window, and the values ​​are arranged according to the task sampling time to form calibration residuals. Starting from zero latency, the candidate response delay is gradually increased according to a uniform sampling period until the short-term response duration of the environment is covered. Under any candidate response delay, the environmental sampling value is shifted backward along the task sampling time, and the change in environmental sampling after the shift is correlated with the change in calibration residual at the same task sampling time. When the change directions of both are consistent and the change amplitudes are both higher than the median absolute deviation corresponding to the stable sampling position, the corresponding task sampling time is marked as the synchronous change position. If the change directions are inconsistent or the change amplitudes are insufficient, the synchronous change position is not recorded.

[0060] The ratio between the calibration residual change and the environmental sampling change is derived from the synchronous change locations. Synchronous change locations that deviate from the median of the ratio by more than a set multiple of the median absolute deviation are marked as abnormal locations and discarded. The median of the remaining ratios is used as the fast drift response ratio corresponding to the current candidate response delay. When the number of synchronous change locations is insufficient, invalid candidate response delays are retained. When there are valid synchronous change locations, the candidate response delay with the largest number of synchronous change locations and the lowest ratio dispersion is selected. The environmental sampling change under the corresponding delay is scaled by the fast drift response ratio. When the absolute value of the scaling result is higher than the absolute value of the calibration residual at the same task sampling time, the scaling result is limited by the absolute value of the calibration residual, while retaining the direction of change of the scaling result. The residual fast drift amount sequence is generated after arranging the results according to the task sampling time.

[0061] It should also be noted that the candidate response delay is formed by increasing from zero delay according to a uniform sampling period to the short-time response duration of the environment.

[0062] Stable sampling locations are determined from the environmental calibration sequence by selecting continuous and valid sampling locations where the adjacent changes in environmental sampling values ​​and environmental calibration concentration values ​​do not exceed the corresponding normal fluctuation boundaries and do not have concentration increase markers.

[0063] S2.3. Subtract the residual fast drift amount point by point from the environmental calibration concentration value within the slow drift window. Mark the concentration rise segment according to the case where the adjacent concentration increment exceeds the normal fluctuation boundary and continuously covers the response time of the gas-sensitive device. After removing the concentration rise segment, calculate the median value of the remaining concentration segment by segment, and perform time-series connection on the adjacent median values ​​to generate the median trajectory of the remaining concentration.

[0064] Furthermore, the residual fast drift is deducted point by point from the environmental calibration concentration value within the current slow drift window according to the task sampling time, and then the increment between the corresponding concentration values ​​at adjacent task sampling times is compared; when the adjacent concentration increments are continuously higher than the normal fluctuation boundary, and the continuous period reaches the response time of the gas sensor, the first out-of-bounds position of the continuous period is marked as the concentration rise start point, and the end of the continuous period is marked as the concentration rise end point, forming a concentration rise segment; when the out-of-bounds duration does not reach the response time of the gas sensor, the corresponding position is retained as a short-term fluctuation and the concentration rise segment is not marked.

[0065] Exclude concentration-rising segments from the current slow drift window, and then divide the remaining sampling segments according to the task sampling time between adjacent concentration-rising segments. When there are enough valid sampling positions in the remaining sampling segments, use the median value of the remaining concentrations as the representative concentration of the corresponding sampling segment. When there are not enough valid sampling positions in the remaining sampling segments, retain the invalid segment marker. When both adjacent representative concentrations are valid, perform linear connection according to the task sampling time corresponding to the adjacent representative concentrations. When there are invalid segments between adjacent representative concentrations, stop the cross-segment connection and restart the connection from the first subsequent valid representative concentration to generate the median trajectory of the remaining concentration.

[0066] It should also be noted that the normal fluctuation boundary is determined by arranging the absolute values ​​of adjacent concentration increments in order of magnitude, calculating their median and median absolute deviation, and using the sum of the median and three times the median absolute deviation. This ensures that normal random fluctuations fall within the boundary, while concentration increments that continuously exceed the normal fluctuation level can be identified as concentration increases.

[0067] The response time of a gas-sensitive device is determined by the duration during which the gas-sensitive response value reaches a stable change after a change in the concentration of the target gas. It is determined based on the duration from the change in the concentration of the target gas to the stable change in the gas-sensitive response value. An example value could be 60 seconds.

[0068] S2.4 Extract adjacent offsets that have not exceeded the aging change range along the median trajectory of the remaining concentration and accumulate them sequentially to generate an aging slow drift sequence. When the current sampling position enters a state of continuous concentration increase, fix the aging slow drift before the starting point of the rise, and unfreeze it after the subsequent complete fast drift window recovers stability.

[0069] Furthermore, the difference between the median values ​​of adjacent task sampling times in the median trajectory of the remaining concentration is used as the adjacent offset. The absolute value of the adjacent offset is compared with the aging change range. When the adjacent offset does not exceed the aging change range, the direction of the adjacent offset is retained. When the adjacent offset exceeds the aging change range, it is not included in the accumulation. An aging offset anomaly mark is retained at the corresponding task sampling time.

[0070] The initial reference position is set to the first valid task sampling time of the median concentration trajectory within the initially formed complete slow drift window. The aging slow drift amount at the initial reference position is set to zero, and adjacent offsets are accumulated sequentially along subsequent valid task sampling times. As the slow drift window moves with the current sampling position, the position with the latest task sampling time and valid aging slow drift amount that has already formed is found in the overlapping section between the previous and current slow drift windows, and this position is used as the connection reference position. The aging slow drift amount already formed in the overlapping section remains unchanged, and adjacent offsets are accumulated only from the newly added task sampling time after the connection reference position. If there is no valid connection reference position in the overlapping section, the aging slow drift amount is not reset to zero, the slow drift connection anomaly mark is retained, and the addition of new aging slow drift amounts is paused until a valid task sampling time that can be connected with the existing aging slow drift amount sequence reappears in the subsequent slow drift window, forming an aging slow drift amount sequence.

[0071] The effective aging slow drift amount corresponding to the nearest task sampling time before the start of the concentration increase segment is used as the frozen aging slow drift amount, and this amount remains unchanged during the continuous concentration increase. After the continuous concentration increase ends, the concentration change is checked from the subsequent complete fast drift window. When there are no concentration increase markers within the complete fast drift window and adjacent concentration increments do not exceed the normal fluctuation boundary, the freeze is lifted, and the frozen aging slow drift amount is used as the recovery benchmark. Effective adjacent offsets are then re-established from the unfrozen position. If there are still out-of-bounds increments within the complete fast drift window, the frozen aging slow drift amount continues to be used until the unfrozen condition is met.

[0072] It should also be noted that the aging variation range is determined by the median value of the offset of adjacent concentration medians in the long-term stable operation record of the gas-sensitive device and the set multiple of the median absolute deviation.

[0073] S2.5. Based on the task sampling time, the residual fast drift and aging slow drift are deducted point by point from the environmental calibration sequence. For positions where the concentration continues to rise, the aging slow drift is frozen, and the positive and negative change directions of the calibration results are retained to generate a dual-scale concentration sequence.

[0074] Furthermore, the environmental calibration concentration values ​​in the environmental calibration sequence are correlated with the residual fast drift sequence according to the task sampling time. When the task sampling time is consistent, the residual fast drift formed based on the short-time environmental response scale is subtracted from the environmental calibration concentration values ​​to obtain the short-time drift correction concentration value. Then, the short-time drift correction concentration value is correlated with the aging slow drift sequence according to the same task sampling time. The aging slow drift formed based on the aging tracking scale of the gas-sensitive device is subtracted from the short-time drift correction concentration value to obtain the dual-scale drift correction concentration value.

[0075] For environmental calibration concentration values ​​that are in a continuously rising concentration position, the correction under the aging tracking scale of the gas-sensitive device is completed using the frozen aging slow drift amount determined before the concentration rise begins. For environmental calibration concentration values ​​that are not in a continuously rising concentration position, the aging slow drift amount corresponding to the task sampling time in the aging slow drift amount sequence is used. When the residual fast drift amount or the aging slow drift amount is missing, the calibration anomaly mark corresponding to the task sampling time is retained, and adjacent drift amounts are not used for replacement. When the dual-scale drift correction concentration value is positive, it is retained as a positive value, and when it is negative, it is retained as a negative value. Zero value truncation is not performed. Then, the dual-scale drift correction concentration values ​​are arranged according to the task sampling time to generate a dual-scale concentration sequence.

[0076] S3. Verify the short-term increment and long-term trend of the dual-scale concentration sequence in the same direction to determine the excitation trigger state. When the excitation trigger state is established, control the gas-sensitive device to switch the working conditions along the stepped excitation trajectory, collect the transient response under each working condition, divide the excitation stage according to the switching time and perform time axis registration, determine the stage response difference between the initial response and the stable response of each excitation stage, determine the recovery ratio between the response drop in the recovery stage and the response difference of the previous excitation stage, and form a gas dynamic fingerprint.

[0077] S3.1. Using the current sampling position of the dual-scale concentration sequence as the end, extract the current short-time segment and the short-time reference segment, remove invalid sampling positions and replace isolated jump values, calculate the difference between the median concentration values ​​of the two segments, and generate the short-time increment.

[0078] Furthermore, using the current sampling position of the dual-scale concentration sequence as the end of the current short-time interval, continuous sampling positions are drawn forward according to the short-time observation duration. Continuous sampling positions with the same duration immediately preceding the start of the current short-time interval are included in the short-time reference interval. Sampling positions marked with calibration anomalies, missing responses, or invalid acquisition status are not included in the formation of the median concentration value. If either the current short-time interval or the short-time reference interval does not cover the short-time observation duration, the incomplete short-time interval marker is retained, and no short-time increment is generated. The effective dual-scale concentration value is compared with the local median concentration value corresponding to the adjacent sampling position. If a single dual-scale concentration value significantly deviates from the local median concentration value and there is no continuous change in the same direction at adjacent positions, the single dual-scale concentration value is replaced with the local median concentration value. If continuous changes in the same direction occur at consecutive positions, the original dual-scale concentration value is retained. The median concentration value of the current short-time interval is compared temporally with the median concentration value of the short-time reference interval, and the differences and directions of change between the median concentration values ​​are recorded to generate a short-time increment.

[0079] It should also be noted that the short-term observation duration is determined by the shortest duration that can stably reflect short-term concentration changes in the historical response record of the gas-sensitive device.

[0080] S3.2 Extract a long-term trend segment covering a continuous short time interval from the dual-scale concentration sequence, perform linear fitting on the effective concentration value, and mark the corresponding rising, stable and falling states according to the fitting change direction and the long-term trend boundary to generate a long-term trend.

[0081] Furthermore, from the dual-scale concentration sequence, taking the current sampling position as the end, a long-term trend segment covering a continuous short time interval is extracted, excluding sampling positions with calibration anomaly markers and invalid acquisition status markers. When the long-term trend segment does not cover the long-term observation period, the incomplete long-term trend segment marker is retained, and no long-term trend corresponding to the current sampling position is formed. The effective concentration values ​​within the long-term trend segment are arranged according to the task sampling time, and a least-squares linear fitting is used to form a concentration change line. The fitting slope is determined according to the following formula: ; ; ; in: Indicates the current sampling position The corresponding long-term trend fitting slope; This indicates the current sampling position number in the dual-scale concentration sequence; Indicates the valid sampling position number within the long-term trend segment; Indicates the current sampling position This represents the long-term trend segment at the end; Indicates sampling position The corresponding task sampling time; Indicates sampling position The corresponding effective dual-scale concentration values; This represents the arithmetic mean of the sampling times within the long-term trend segment. This represents the arithmetic mean of the effective dual-scale concentration values ​​within a long-term trend segment. This indicates the number of valid sampling locations within a long-term trend segment; Indicates sampling position This belongs to a long-term trend segment; This indicates the deviation of the task sampling time from the average value of the task sampling time. This indicates the deviation of the effective two-scale concentration value from the average effective two-scale concentration value; When the fitting slope is higher than the long-term trend boundary, it is marked as an upward state; when the fitting slope is between the positive and negative long-term trend boundaries, it is marked as a stable state; when the fitting slope is lower than the negative long-term trend boundary, it is marked as a downward state. When the effective concentration value is insufficient, the task sampling time is repeated, or the fitting result cannot form a stable direction, the invalid trend is retained. The effective trend states are arranged according to the task sampling time to generate the long-term trend.

[0082] It should also be noted that the long-term observation period is determined by the shortest duration in the historical response record of the gas-sensitive device that can distinguish between continuous concentration changes and short-term concentration fluctuations.

[0083] Arrange the absolute values ​​of each fitted slope in order of magnitude, calculate the median and median absolute deviation, and use the median absolute deviation plus a set multiple as the positive long-term trend boundary, and the opposite of the median as the negative long-term trend boundary. The set multiple can be determined based on the coverage of normal fluctuation slopes in historical stable operation records, so that most of the fitted slopes in the stable operation state fall between the positive and negative boundaries. For example, three times the median absolute deviation can be used as the judgment criterion to obtain the long-term trend boundary.

[0084] S3.3 Verify the direction of change of short-term increment and long-term trend. When the short-term increment exceeds the short-term increment boundary and the long-term trend continues to rise to the minimum effective response time of the gas-sensitive device, the excitation trigger state is determined to be established, and a stepped excitation trigger record is generated.

[0085] Furthermore, the short-term increment is compared with the long-term trend corresponding to the same task sampling time to verify the direction of change. If the short-term increment is positive and the long-term trend is in an upward state, the direction of change is confirmed to be consistent. If the short-term increment is negative and the long-term trend is in a downward state, the direction of change is confirmed to be consistent. If the direction of the short-term increment is inconsistent with the state of the long-term trend, the direction is marked as inconsistent and monitoring continues. The short-term increment is compared with the short-term increment boundary. If the short-term increment does not exceed the short-term increment boundary, the duration verification is not performed. If the short-term increment exceeds the positive short-term increment boundary and the long-term trend continues to rise, the duration is accumulated from the starting task sampling time of the continuous rising state. When the duration reaches the minimum effective response time of the gas-sensitive device, the current task sampling time is registered as the excitation trigger time, and the short-term increment, long-term trend, and duration are associated to determine that the excitation trigger state is established. The current accumulation is terminated when the continuous rising state is interrupted or the short-term increment falls back to within the short-term increment boundary, the excitation trigger state is determined to be invalid, and a stepped excitation trigger record is generated.

[0086] It should also be noted that when there is no change in the target gas concentration, only environmental disturbances and normal measurement fluctuations, the absolute values ​​of each short-time increment are extracted. All short-time increment absolute values ​​are arranged according to their numerical values, and their median value and median absolute deviation are calculated. The sum of the median value of the short-time increment absolute value and the median absolute deviation of a set multiple is used as the boundary of the short-time increment. For example, three times the median absolute deviation can be used as the criterion for judging abnormal changes. Short-time increments exceeding this boundary are considered to be outside the normal fluctuation range.

[0087] Under different known target gas concentrations, record the duration of the gas-sensitive response value changing from the initial state to the stable response state. The shortest duration that allows the gas-sensitive response to reach a stable change and obtain consistent response characteristics in repeated tests is taken as the minimum effective response time of the gas-sensitive device. For example, if the gas-sensitive response can reach a stable and repeatable change state within 30s after continuous action of the target gas, then 30s can be taken as the minimum effective response time.

[0088] S3.4. Determine the reference response before step excitation based on the step excitation trigger record, control the gas-sensitive device to switch the heating temperature step by step along the step excitation trajectory, and register the temperature stabilization time after the actual heating temperature enters the target temperature allowable deviation range and remains stable, and generate the working condition switching record.

[0089] Furthermore, starting from the excitation trigger time corresponding to the stepped excitation trigger record, a continuous stable segment without a concentration increase marker is searched backwards. The median value of the gas-sensitive response within the continuous stable segment is used as the baseline response before stepped excitation. If the continuous stable segment does not cover the baseline observation time, the baseline response invalidation marker is retained, and the current stepped excitation is stopped. The target heating temperatures in the stepped excitation trajectory are sent to the gas-sensitive device step by step according to the operating conditions. After the previous target heating temperature is stabilized and confirmed, the process switches to the next target heating temperature to avoid overlap of transient responses corresponding to adjacent operating conditions.

[0090] The actual heating temperature is continuously compared with the current target heating temperature. Once the actual heating temperature enters the allowable deviation range of the target temperature, the temperature holding time is accumulated. When the temperature holding time reaches the temperature stabilization time, the starting time of the accumulated task sampling is recorded as the temperature stabilization time. If the actual heating temperature exceeds the allowable deviation range of the target temperature during the accumulation period, the current accumulation is cleared, and accumulation restarts from the point where it re-enters the allowable deviation range of the target temperature. If the target heating temperature is not confirmed to be stable within the duration of the working conditions, an abnormal working condition switching flag is retained, and subsequent heating temperature switching is stopped. When stability is confirmed, the target heating temperature, the working condition switching time, and the temperature stabilization time are associated to generate a working condition switching record.

[0091] It should also be noted that the stepped excitation trajectory is the result of the arrangement of working conditions determined by the allowable heating temperature range of the gas-sensitive device and the target gas response test records.

[0092] The allowable deviation of the target temperature is determined by combining the heating temperature acquisition resolution and the heating temperature interval of adjacent targets, and is kept below half of the heating temperature interval of adjacent targets.

[0093] Repeatedly perform heating and isothermal tests at each target heating temperature, calculate the duration required for the actual heating temperature to remain within the allowable deviation range of the target temperature and for the temperature change amplitude to not exceed the stable fluctuation boundary, and take the upper limit of the duration obtained from repeated tests as the temperature stabilization duration of the corresponding target heating temperature.

[0094] Based on the continuous stable range before the step excitation is triggered and not affected by the change in target gas concentration, the fluctuation range of the gas-sensitive response value under different observation durations is statistically analyzed. When the fluctuation of the response value within the continuous observation duration is within the boundary of the stable response fluctuation and the median response value no longer changes significantly, the shortest duration that meets the stability requirement is determined as the benchmark observation duration.

[0095] S3.5. Based on the working condition switching record, continuously collect gas-sensitive response values ​​starting from the first working condition switching moment, divide the excitation stage by adjacent switching moments, classify the response range after returning to normal recovery temperature into the recovery stage, and mark the thermal transition zone at the beginning of each stage to generate the excitation stage.

[0096] Furthermore, starting from the first working condition switching moment in the working condition switching record, gas-sensitive response values ​​are continuously collected. The gas-sensitive response values ​​from the current working condition switching moment to the next working condition switching moment are classified as the current excitation phase, and the response period from the end of the last target heating temperature back to the normal recovery temperature is classified as the recovery phase. If a working condition switching moment is missing, an adjacent working condition switching moment is not used as a substitute, and a missing phase boundary marker is retained at the corresponding position; if a subsequent working condition switching moment is missing, the current excitation phase is extended to the end of the gas-sensitive response value collection moment, while retaining an incomplete phase end marker.

[0097] The gas-sensitive response values ​​between the operating condition switching point and the corresponding temperature stabilization point are marked as the thermal transition zone. Gas-sensitive response values ​​after the temperature stabilization point are retained as valid responses for that stage. If the temperature stabilization point is missing, the corresponding excitation stage is marked as an invalid stage and does not proceed to the stage response difference formation step. The recovery stage uses the operating condition switching point of returning to normal recovery temperature as the starting point of the recovery stage and continues to collect data until the gas-sensitive response values ​​meet the recovery stabilization conditions and the recovery observation duration. The excitation stages and recovery stages are arranged in the order of operating conditions to generate the excitation stage.

[0098] It should also be noted that the normal recovery temperature is determined by the heating temperature used to keep the gas-sensitive device running stably before the start of stepped excitation.

[0099] The recovery stabilization condition is determined by recording the stable operation of the gas-sensitive device at the normal recovery temperature. When the maximum fluctuation of the gas-sensitive response value within the continuous response range at the end of the recovery phase does not exceed the stable fluctuation boundary, and adjacent gas-sensitive response values ​​do not show a continuous change in the same direction, the gas-sensitive response value is determined to meet the recovery stabilization condition.

[0100] The recovery observation time is determined by repeated recovery test records after the gas-sensitive device completes step excitation and returns to the normal recovery temperature. The duration during which the gas-sensitive response value remains stable after first meeting the recovery stabilization conditions is statistically analyzed, and the upper limit of the duration obtained from each effective recovery test is taken as the recovery observation time.

[0101] S3.6. Using the temperature stabilization time of each excitation stage as the stage zero time, the gas-sensitive response value is converted to the relative sampling time and registered according to the uniform sampling interval. For positions that exceed the allowable interpolation time, a response missing mark is retained to generate the time axis registration excitation stage.

[0102] Furthermore, the temperature stabilization moment corresponding to each excitation phase is taken as the zero moment of the phase. The task sampling moment within the phase is converted into a relative sampling moment relative to the temperature stabilization moment. The thermal transition zone before the temperature stabilization moment corresponds to the negative relative sampling moment, the temperature stabilization moment corresponds to the zero moment, and the temperature stabilization moment corresponds to the positive relative sampling moment. Excitation phases with missing temperature stabilization moments retain invalid phase markers and do not perform time axis registration.

[0103] The relative sampling times are arranged according to a uniform sampling interval to form the registration times. When the registration time matches the original relative sampling time, the original gas-sensitive response value is retained. When the registration time is between adjacent valid responses and the time interval between adjacent valid responses does not exceed the allowable interpolation duration, linear interpolation is used to supplement the gas-sensitive response value corresponding to the registration time. When the time interval between adjacent valid responses exceeds the allowable interpolation duration, cross-interval supplementation is not performed, and a response missing marker is retained at the corresponding registration time. The registered gas-sensitive response values ​​are arranged according to the relative sampling times to generate the time-axis registration excitation stage.

[0104] It should also be noted that the allowable interpolation duration is determined by combining the uniform sampling interval and the allowable continuous missing duration of the gas-sensitive response value, and is kept within the duration during which the short-term response changes of the gas-sensitive device can be continuously identified.

[0105] like Figure 5As shown, the normalized gas-sensitive response values ​​of acetone, ethanol, and isopropanol are quite similar in the normal operating stage, indicating that it is difficult to effectively distinguish them based solely on the steady-state response under fixed operating conditions. After the gas-sensitive device successively enters the excitation stages of 300℃, 350℃, and 400℃, the curves of the three gases exhibit different characteristics of rise, fall, and stabilization. Among them, the response of acetone continuously increases with increasing temperature, ethanol reaches a high response in the middle stage and then falls, while the subsequent increase of isopropanol is relatively small. By using the temperature stabilization moment as the zero moment of the stage for time axis registration, the influence of heating transition differences on the response position can be eliminated, providing a reliable basis for extracting stage response differences and forming gas dynamic fingerprints.

[0106] S3.7 Extract the initial response from the thermal transition zone after the time axis registration excitation stage, extract the stable response from the response range that continuously satisfies the stable fluctuation condition at the end of the stage, and determine the stage response difference by the change of the stable response relative to the initial response.

[0107] Furthermore, the thermal transition zone is excluded from the time-axis registration excitation stage. Effective gas-sensitive response values ​​are retained within the initial observation section after the temperature stabilizes, and the median value of the gas-sensitive response values ​​within the initial observation section is used as the initial response. If the effective gas-sensitive response values ​​within the initial observation section are insufficient, the corresponding excitation stage is marked as an invalid stage, thus preventing the formation of stage response differences. The initial observation section begins at the temperature stabilizes and avoids registration positions still marked with missing responses.

[0108] The end of the excitation phase is divided into continuous stable observation sections. The maximum fluctuation range of the gas-sensitive response value and the variation range of adjacent responses within each stable observation section are checked. If the maximum fluctuation range of the continuous stable observation section does not exceed the stable fluctuation boundary, and adjacent response changes do not show a continuous unidirectional change, the continuous response range near the end of the excitation phase is taken as the stable response range, and the median value of the gas-sensitive response within the stable response range is taken as the stable response. If no stable response range is formed at the end of the excitation phase, the stage instability marker is retained, and no stage response difference is formed. When both the initial response and the stable response are valid, the stage response difference is determined by the change in the stable response relative to the initial response.

[0109] It should also be noted that the gas-sensitive response value difference corresponding to the sampling time of adjacent tasks within the stable operation phase is extracted to form a response fluctuation amplitude sequence. The response fluctuation amplitudes are arranged according to their numerical values, and their statistical distribution characteristics are calculated. The combination of the median value of the fluctuation amplitude and the median absolute deviation is used as the stable fluctuation boundary. For example, the median value of the fluctuation amplitude plus three times the median absolute deviation is used to determine the stable fluctuation boundary, which is used to distinguish between normal response fluctuations and unstable response changes at a fixed temperature.

[0110] S3.8 Extract the response fall-off from the recovery phase. When the response difference of the previous excitation phase exceeds the minimum effective response boundary, determine the recovery ratio. Arrange the phase response differences and associate them with the recovery ratio according to the working conditions. Retain the corresponding marks for invalid phases and unrecovered positions to form a gas dynamic fingerprint.

[0111] Furthermore, the thermal transient response after returning to normal recovery temperature is excluded from the recovery phase, and the same stable fluctuation determination method as in the excitation phase is used to find the range of stable recovery response. The median value of the gas-sensitive response within the range of stable recovery response is taken as the stable recovery response. If a stable recovery response range is not formed in the recovery phase, the unrecovered position is marked, and no recovery ratio is formed. The direction of change of the stable recovery response is verified with the stable response of the previous excitation phase, and the response drop is determined according to the following formula: ; In the formula, Indicates the first The corresponding response drop after the end of each incentive phase; Indicates the first The stage response difference corresponding to each incentive stage; Indicates the first The median value of the gas-sensitive response within the stable response range corresponding to each excitation stage; This represents the median value of the gas-sensitive response within the recovery stability response range corresponding to the recovery phase.

[0112] When the response fallback is less than zero, an abnormal recovery direction marker is retained. The absolute value of the response difference from the previous excitation stage is compared with the minimum effective response boundary. If the response difference from the previous excitation stage does not exceed the minimum effective response boundary, an insufficient stage response marker is retained, and no recovery ratio is formed. If the response difference from the previous excitation stage exceeds the minimum effective response boundary, the recovery ratio is determined according to the following formula: ; In the formula, Indicates the first The recovery ratio corresponds to each excitation stage; the response fall-off and the stage response difference use the same response dimension, and the recovery ratio is a dimensionless number.

[0113] When the recovery ratio is higher than the corresponding ratio for full recovery, the recovery overshoot feature is retained; when the recovery ratio is lower than zero, the recovery direction anomaly marker is retained, and the recovery ratio is not truncated. The stage response differences are arranged in order of operating conditions, and the recovery ratio is associated with the excitation stage before returning to normal recovery temperature. Corresponding markers are retained for invalid stages and unrecovered positions to form a gas dynamic fingerprint.

[0114] It should also be noted that the gas dynamic fingerprint is a dynamic response record composed of stage response difference, recovery ratio, invalid stage marker and unrecovered position marker arranged in the order of working conditions. It is used to characterize the response change characteristics and recovery ability of the target gas at different heating temperatures.

[0115] The absolute values ​​of the stage response differences under the action of target gas of different known concentrations are extracted. All stage response difference absolute values ​​are arranged according to their numerical values, and their distribution characteristics are statistically analyzed. The lowest response amplitude that can distinguish between normal response fluctuations and effective gas response changes is taken as the minimum effective response boundary.

[0116] Extract the response difference of the previous excitation stage and the response drop in the recovery stage from each valid test, calculate the ratio of their absolute values, screen test records that recover stable response and return to the allowable fluctuation range of the baseline response before stepped excitation, and determine the corresponding proportion of complete recovery based on the median value of the corresponding ratio.

[0117] like Figure 6 As shown, after the final excitation stage ends and the temperature returns to normal, the normalized gas-sensitive response values ​​of acetone, ethanol, and isopropanol all gradually decrease. However, the magnitude, rate, and stable recovery position of the three gases differ significantly. Acetone exhibits a larger response decrease, ethanol shows a moderate recovery, and isopropanol recovers more slowly in the later stages. Each curve also retains small, continuous fluctuations, consistent with the actual sampling characteristics during the recovery process of gas-sensitive devices. By comparing the stable recovery response with the stable response of the previous excitation stage, and determining the recovery ratio based on the response decrease and the difference between the stage responses, gases with similar static responses can be further distinguished, improving the reliability of dynamic fingerprint matching of target gases.

[0118] S4. Based on the target gas response location marked by the gas dynamic fingerprint, extract the corresponding concentration value from the dual-scale concentration sequence, determine the alarm threshold and confirmation time according to the concentration change status, and generate a gas monitoring record.

[0119] S4.1 Arrange the stage response difference, recovery ratio, invalid stage marker, and unrecovered position marker in the gas dynamic fingerprint according to the working conditions, and correspond them item by item with the fixed fingerprint positions in the target gas reference fingerprint that are arranged in the same order. For the stage response difference position, verify whether the stage response difference is between the lower and upper reference boundaries, and for the recovery position, verify whether the recovery ratio is between the lower and upper reference boundaries. If all the values ​​corresponding to the fixed fingerprint positions fall within the upper and lower reference boundaries, and the fixed fingerprint positions do not have invalid stage markers, unrecovered position markers, or abnormal recovery direction markers, then the gas dynamic fingerprint is confirmed to belong to the target gas dynamic fingerprint. If the fixed fingerprint positions are missing, the arrangement order is inconsistent, or the values ​​exceed the upper and lower reference boundaries, retain the fingerprint matching failure marker, and do not mark the target gas response position.

[0120] Extract the excitation trigger time from the stepped excitation trigger record corresponding to the dynamic fingerprint of the target gas. Locate the same task sampling time in the dual-scale concentration sequence. Then, search backward from the excitation trigger time for the nearest continuous segment where the adjacent concentration increment does not exceed the normal fluctuation boundary and the long-term trend remains stable. Take the first valid sampling position after the end of the nearest continuous segment as the starting position of the target gas response. Using the concentration change direction corresponding to the excitation trigger time as the extension direction, continuously include sampling positions with the same concentration change direction and stable after rising along the task sampling time. When the long-term trend turns to the opposite direction, or when the dual-scale concentration value falls back to the concentration distribution boundary corresponding to the nearest continuous segment, stop extending the position and register the end position of the target gas response. If the stopping condition is not met before the end of the dual-scale concentration sequence, take the last valid sampling position as the temporary end position and retain the unclosed response segment mark to obtain the target gas response position.

[0121] It should also be noted that the target gas reference fingerprint is formed by statistically analyzing the distribution boundaries of the stage response difference and recovery ratio based on repeated response test records of the target gas under the same step excitation trajectory, according to the working conditions.

[0122] S4.2 Match the task sampling time in the dual-scale concentration sequence according to the target gas response location, exclude the concentration extraction isolation section corresponding to the step excitation, extract the dual-scale concentration values ​​and arrange them according to the task sampling time to generate the target gas concentration trajectory.

[0123] Furthermore, the task sampling times corresponding to the start and end positions of the target gas response are matched to the dual-scale concentration sequence one by one. When the task sampling times are consistent, the corresponding dual-scale concentration values ​​are extracted. When the task sampling time is missing or the dual-scale concentration value has a calibration anomaly mark, the concentration missing mark is retained and adjacent concentration values ​​are not used as substitutes.

[0124] The sampling time from the first working condition switching moment to the recovery phase to form a stable recovery response is included in the concentration extraction isolation zone, and the dual-scale concentration values ​​in the concentration extraction isolation zone are excluded from the target gas response location; if the stable recovery response is not formed, the sampling time after returning to the normal recovery temperature is retained in the concentration extraction isolation zone, and an unclosed isolation zone mark is added.

[0125] The effective dual-scale concentration values ​​outside the isolation zone are arranged according to the mission sampling time, and the target gas response start position, target gas response end position and concentration missing marker are associated to generate the target gas concentration trajectory.

[0126] It should also be noted that the concentration extraction isolation section starts at the first working condition switching moment and ends at the task sampling moment when the gas-sensitive device returns to normal recovery temperature and forms a stable recovery response. This is used to exclude dual-scale concentration values ​​during heating temperature switching, thermal transition, and response recovery. If a stable recovery response has not been formed, the isolation section is extended to the end of the gas-sensitive response value acquisition and the isolation section is marked as not closed.

[0127] S4.3. Compare adjacent dual-scale concentration values ​​from the target gas concentration trajectory according to the task sampling time. Mark a rising concentration state when adjacent concentration increments are continuously higher than the normal fluctuation boundary, a stable concentration state when adjacent concentration increments are within the normal fluctuation boundary, and a falling concentration state when adjacent concentration increments are continuously lower than the negative normal fluctuation boundary. Match the target gas identifier and the current concentration change state with the gas identifier and status identifier in the target gas alarm parameter record. When the identifiers match, extract the corresponding alarm threshold and confirmation duration, and register them as the effective alarm threshold and confirmation duration used in the current over-limit process. If no matching status identifier is found, the alarm threshold is missing, or the confirmation duration is invalid, retain the alarm parameter abnormality mark and do not proceed to alarm confirmation.

[0128] The first valid dual-scale concentration value exceeding the effective alarm threshold in the target gas concentration trajectory is taken as the initial exceedance location. The duration of exceedance between adjacent valid task sampling times is accumulated from this initial exceedance location. Accumulation continues as the dual-scale concentration value remains above the effective alarm threshold, terminating when the dual-scale concentration value falls below the effective alarm threshold, a concentration missing marker appears, or the concentration extraction isolation zone is entered. When the accumulated valid exceedance period reaches the confirmation duration, the task sampling time reaching the confirmation duration is recorded as the alarm confirmation time, and the alarm result is marked as confirmed. When the accumulated valid exceedance period does not reach the confirmation duration, the alarm result is marked as pending confirmation, and after the exceedance state terminates, the alarm result is marked as unconfirmed. The target gas identifier, target gas response location, initial exceedance location, effective alarm threshold, confirmation duration, alarm confirmation time, corresponding concentration value, and alarm result are associated according to the task sampling time to generate a gas monitoring record.

[0129] In summary, this invention achieves the following: after the excitation trigger state is established, the gas-sensitive device is controlled to switch the heating temperature along the stepped excitation trajectory. The excitation stages are divided and registered based on the temperature stabilization time. The initial response, stable response, stage response difference, and recovery ratio are extracted to form a gas dynamic fingerprint characterizing the response law of the target gas. This enables the intelligent sensing system to verify the identity of the source of the concentration rise and, based on this, limit the target gas response position and concentration trajectory, preventing drift components and non-target gas responses from directly entering the alarm judgment, thereby improving the reliability of target gas identification and the directionality of alarms.

[0130] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A laboratory gas monitoring method based on Internet of Things (IoT) data acquisition, characterized in that, include: According to the IoT acquisition task, gas sampling frames are collected, integrity is verified and sampling time is aligned, an aligned gas sequence is generated, the environmental sampling value of the aligned gas sequence is used to locate the calibration interval, the drift of adjacent calibration points is interpolated, the environmental drift of the original gas response value is subtracted, and an environmental calibration sequence is generated. Fast drift window and slow drift window are established around the environmental calibration sequence. The calibration residual is calculated using the median value of the fast drift window. The residual fast drift amount is extracted according to the synchronization relationship between the environmental sampling value and the calibration residual. The concentration rise segment is removed in the slow drift window. The aging slow drift amount is extracted along the median trajectory of the remaining concentration. The residual fast drift amount and the aging slow drift amount are deducted point by point. The aging slow drift amount is frozen and corrected when the concentration continues to rise, generating a dual-scale concentration sequence. The short-term increment and long-term trend of the dual-scale concentration sequence are verified in the same direction to determine the excitation trigger state. When the excitation trigger state is established, the gas-sensitive device is controlled to switch the working conditions along the stepped excitation trajectory. The transient response under each working condition is collected. The excitation stage is divided according to the switching time and the time axis is registered. The stage response difference is determined from the initial response and the stable response of each excitation stage. The recovery ratio is determined from the response drop in the recovery stage and the response difference of the previous excitation stage, thus forming a gas dynamic fingerprint. Based on the target gas response location marked by the gas dynamic fingerprint, the corresponding concentration value in the dual-scale concentration sequence is extracted, and the alarm threshold and confirmation time are determined according to the concentration change status to generate a gas monitoring record.

2. The laboratory gas monitoring method based on Internet of Things data acquisition as described in claim 1, characterized in that, The specific steps for generating the aligned gas sequence are as follows: Verify the structural integrity and frame sequence continuity of the gas sampling frames, remove abnormal frames and mark missing positions, and generate valid gas sampling frames; The local sampling time of the effective gas sampling frame is corrected according to the unified sampling period and matched to the task sampling time within the allowable deviation to generate an aligned gas sequence.

3. The laboratory gas monitoring method based on Internet of Things data acquisition as described in claim 2, characterized in that, The specific steps for generating the environmental calibration sequence are as follows: Environmental sampling values ​​are extracted one by one from the aligned gas sequence. The environmental sampling values ​​are compared with the corresponding calibration nodes to determine the upper and lower boundary nodes. When the current environmental sampling value is equal to the maximum calibration node, the calibration interval is defined by the maximum calibration node and the previous calibration node. When the boundary is exceeded, the environmental sampling value is clamped to the nearest boundary node to generate a calibration interval record. Based on the calibration interval records, adjacent calibration points are extracted, the distance ratio between the environmental sampling values ​​and the upper and lower boundary nodes is calculated, and boundary constraints are applied to the distance ratio exceeding the boundary to generate the environmental interpolation ratio. The drift of adjacent calibration points is linearly interpolated according to the environmental interpolation ratio to determine the environmental drift. The environmental drift is then subtracted from the original gas response values ​​in the aligned gas sequence and arranged according to the task sampling time to generate an environmental calibration sequence.

4. The laboratory gas monitoring method based on Internet of Things data acquisition as described in claim 1, characterized in that, The process involves establishing fast drift windows and slow drift windows around the environmental calibration sequence, calculating the calibration residual using the median value of the fast drift window, and extracting the residual fast drift amount based on the synchronization relationship between the environmental sampling values ​​and the calibration residual. The specific steps are as follows: Using the current sampling position of the environmental calibration sequence as the end of the window, continuous sampling positions are truncated forward according to the short-time response time of the environment and the aging tracking time of the gas-sensitive device, respectively. The window is moved synchronously with the task sampling time and invalid sampling positions are excluded to generate the current fast drift window and the current slow drift window. The median value of the environmental calibration values ​​within the fast drift window is calculated. The calibration residual is determined by the deviation of each environmental calibration value from the median value. The environmental sampling values ​​are moved according to the candidate response delay and the synchronous change positions are screened. After the abnormal positions are removed, the fast drift response ratio is determined. The environmental sampling change amount corresponding to the fast drift response ratio is scaled and limited by the calibration residual amplitude to generate the residual fast drift amount sequence.

5. The laboratory gas monitoring method based on Internet of Things data acquisition as described in claim 1 or 4, characterized in that, The process involves removing concentration-increasing segments within the slow-drift window, extracting the aged slow-drift amount along the median trajectory of the remaining concentration, deducting the residual fast-drift amount and the aged slow-drift amount point by point, and freezing the aged slow-drift amount for correction when the concentration continues to rise, thereby generating a dual-scale concentration sequence. The specific steps are as follows: The residual fast drift amount is deducted point by point from the environmental calibration value within the slow drift window. The concentration rise segment is marked according to the case where the adjacent concentration increment exceeds the normal fluctuation boundary and continuously covers the response time of the gas-sensitive device. After removing the concentration rise segment, the median value of the remaining concentration is calculated in segments, and the adjacent median values ​​are connected in time sequence to generate the median trajectory of the remaining concentration. The adjacent offsets that do not exceed the aging change range are extracted along the median trajectory of the remaining concentration and accumulated sequentially to generate an aging slow drift sequence. When the concentration enters a state of continuous increase at the current sampling position, the aging slow drift before the starting point of the rise is fixed, and the freeze is lifted after the subsequent complete fast drift window recovers stability. According to the sampling time of the task, the residual fast drift and aging slow drift are deducted point by point from the environmental calibration sequence. For positions where the concentration continues to rise, the aging slow drift is frozen and the positive and negative change directions of the calibration results are retained to generate a dual-scale concentration sequence.

6. The laboratory gas monitoring method based on Internet of Things data acquisition as described in claim 5, characterized in that, The steps for verifying the short-term increments and long-term trends of the dual-scale concentration sequence in the same direction to determine the excitation trigger state are as follows: Using the current sampling position of the dual-scale concentration sequence as the end, the current short-time segment and the short-time reference segment are truncated, invalid sampling positions are removed and isolated jump values ​​are replaced, the difference between the median concentration values ​​of the two segments is calculated, and the short-time increment is generated. Long-term trend segments covering continuous short time intervals are extracted from dual-scale concentration sequences. Linear fitting is performed on the effective concentration values. Based on the fitting change direction and the long-term trend boundary, the corresponding rising, stable and falling states are marked to generate a long-term trend. Verify the direction of change of short-term increment and long-term trend. When the short-term increment exceeds the short-term increment boundary and the long-term trend continues to rise to reach the minimum effective response time of the gas-sensitive device, determine that the excitation trigger state is established and generate a stepped excitation trigger record.

7. The laboratory gas monitoring method based on Internet of Things data acquisition as described in claim 6, characterized in that, When the excitation trigger state is established, the gas-sensitive device is controlled to switch operating conditions along the stepped excitation trajectory, and the transient response under each operating condition is collected. The excitation stages are divided according to the switching time and time axis registration is performed. The specific steps are as follows: Based on the stepped excitation trigger record, the reference response before stepped excitation is determined, the gas-sensitive device is controlled to switch the heating temperature step by step along the stepped excitation trajectory, and the temperature stabilization time is recorded after the actual heating temperature enters the target temperature allowable deviation range and remains stable, thus generating a working condition switching record. Based on the working condition switching record, gas-sensitive response values ​​are continuously collected starting from the first working condition switching moment. The excitation stage is divided by adjacent switching moments. The response range after returning to normal recovery temperature is included in the recovery stage. The thermal transition zone at the beginning of each stage is marked to generate the excitation stage. Using the temperature stabilization moment of each excitation stage as the stage zero moment, the gas-sensitive response value is converted to the relative sampling moment and registered according to a uniform sampling interval. For positions that exceed the allowable interpolation time, a missing response mark is retained to generate the time axis registration excitation stage.

8. The laboratory gas monitoring method based on Internet of Things data acquisition as described in claim 7, characterized in that, The process of determining the stage response difference between the initial response and the stable response of each excitation stage, and determining the recovery ratio between the response drop-off in the recovery stage and the response difference of the previous excitation stage to form a gas dynamic fingerprint, is as follows: The initial response is extracted from the thermal transition zone after the time axis registration excitation stage, and the stable response is extracted from the response range that continuously satisfies the stable fluctuation condition at the end of the stage. The stage response difference is determined by the change of the stable response relative to the initial response. Extract the response fall-off from the recovery phase, determine the recovery ratio when the response difference of the previous excitation phase exceeds the minimum effective response boundary, arrange the phase response differences and associate them with the recovery ratio according to the working conditions, retain the corresponding marks for invalid phases and unrecovered positions, and form a gas dynamic fingerprint.

9. The laboratory gas monitoring method based on Internet of Things data acquisition as described in claim 8, characterized in that, The specific steps for marking the target gas response position based on the gas dynamic fingerprint are as follows: matching the gas dynamic fingerprint with the target gas reference fingerprint position by position according to the fixed fingerprint position; confirming the target gas dynamic fingerprint when the stage response difference and recovery ratio fall within the corresponding reference range; and searching forward for the stable segment and extending along the concentration change direction according to the step excitation trigger record to obtain the target gas response position.

10. The laboratory gas monitoring method based on Internet of Things data acquisition as described in claim 9, characterized in that, The steps for extracting the corresponding concentration values ​​from the dual-scale concentration sequence, determining the alarm threshold and confirmation duration based on the concentration change status, and generating a gas monitoring record are as follows: Based on the target gas response location, the task sampling time in the dual-scale concentration sequence is matched, the concentration extraction isolation section corresponding to the step excitation is excluded, the dual-scale concentration values ​​are extracted and arranged according to the task sampling time to generate the target gas concentration trajectory. The effective alarm threshold and confirmation duration are determined based on the change status of the target gas concentration trajectory. The effective over-limit period is accumulated from the first over-limit location. When the accumulated period reaches the confirmation duration, the alarm is confirmed, and the target gas response location, concentration value and alarm result are associated to generate a gas monitoring record.

Images