An asynchronous co-sampling control system and method of a multi-modal gas sensor

Abstract

The application relates to the technical field of gas sensor detection, and discloses an asynchronous cooperative sampling control system and method of a multimodal gas sensor, which comprises the following steps: acquiring real-time original sampling values of each electrochemical sentinel channel, extracting the occurrence time of a signal change amount and a signal deviation change amount first exceeding a noise base threshold value to construct a two-dimensional time sequence fingerprint vector, combining a corrected cross interference characteristic domain to complete attribution determination and generate a trigger request signal; after receiving the trigger request signal, activating a surge immunity window, executing dynamic compensation on the sampling values in the immunity window, comparing the ratio of a trigger frequency in a sliding time window and a dynamic upper limit determined according to a battery residual capacity, and adaptively switching the working mode of a high-power infrared sensor; the application can realize low-power stable operation of the multimodal gas sensor, prolong the equipment endurance time and the service life of the device.

Description

Technical Field

[0001] This invention relates to the field of gas sensor detection technology, and more specifically, to an asynchronous cooperative sampling control system and method for a multimodal gas sensor. Background Technology

[0002] Portable multi-gas detection devices are core devices for safety protection in confined space operations in municipal, petrochemical, and mining fields. They often adopt a multi-modal detection architecture that combines low-power electrochemical sensors with high-power infrared sensors. The electrochemical sensor acts as a normally open sentinel channel for continuous monitoring, while the infrared sensor is activated on demand as a target gas detection unit, thus balancing the device's detection response performance and battery life.

[0003] Among existing related technologies, Chinese patent application CN119881223A discloses an intelligent fault diagnosis method and system for multi-channel gas sensors. This scheme obtains an effective sensor set by performing random geometric analysis and constructing response-limited regions on a multi-channel gas sensor array. The optimal sensor combination is then selected through correlation analysis and a sensitivity-accuracy dilution metric function. Feature data matrix is ​​obtained by extracting features from periodic sampling data. A fault diagnosis trigger signal is generated through deep Q-network analysis, and finally, a bidirectional long short-time memory network is used to determine the fault type, achieving accurate identification and classification of different types of faults in multi-channel gas sensors. Chinese patent application CN110441471A discloses a monitoring method and system for multi-sensor collaborative operation. This scheme categorizes sensors into low-power and high-power sensors based on their power consumption. After the system is activated, both types of sensors perform sampling at preset sampling periods. When the measured value of a low-power sensor reaches or exceeds its set alarm threshold, real-time sampling by both low-power and high-power sensors is triggered until the system returns to normal, thereby reducing system energy consumption and improving the comprehensiveness, accuracy, and timeliness of environmental monitoring.

[0004] However, in the asynchronous collaborative architecture of multimodal gas sensors that uses electrochemical sensors as low-power normally open sentinel channels and infrared sensors as high-power detection units, electrochemical sensors generally exhibit cross-sensitivity to non-target gases. When multiple electrochemical sentinel sensors share the same power supply rail, the abnormal sentinel channel signal caused by cross-interference will trigger the high-power infrared sensor to start. The surge current generated during the infrared sensor startup process will cause a transient drop in the power supply rail voltage, which in turn causes fluctuations in the bias voltage of the electrochemical sensor and a step shift in the sampling baseline. This shift will be identified by the sampling logic of the sentinel channel as a new gas abnormality signal, forming a closed positive feedback loop of "trigger-voltage drop-baseline shift-re-trigger", causing the system to fall into a signal loop self-locking state. In existing technologies, the triggering logic of the sentry channel only sets independent thresholds for the amplitude or rate of change of each channel signal. It cannot distinguish whether the change in the sampled signal originates from a real gas event or a false response derived from cross-interference, nor can it distinguish whether the signal fluctuation originates from changes in gas concentration or electrical artifacts caused by power supply disturbances. As a result, in closed operation scenarios containing trace amounts of non-target interfering gas, the infrared sensor is repeatedly triggered even when there is no target gas leakage. This causes the battery power to be excessively consumed in a short period of time, which cannot meet the long-term use requirements of the equipment. At the same time, the thermal shock generated by the frequent power-on of the infrared sensor will accelerate the fatigue aging of optical components and shorten the effective service life of the device. Summary of the Invention

[0005] This invention relates to portable multi-gas detection equipment applicable to confined space operations such as municipal underground operations, petrochemical pipeline inspections, and mining. It can be used for sampling control and power management of multimodal gas sensing systems that employ low-power electrochemical sensors as normally open monitoring units and high-power infrared gas sensors as target gas detection units. To overcome the aforementioned deficiencies of existing technologies, this invention provides an asynchronous collaborative sampling control system and method for multimodal gas sensors. A two-dimensional temporal fingerprint vector is constructed by extracting the signal change and the event time when the signal deviation exceeds the noise floor from the electrochemical sentinel channel. This vector is then combined with a cross-interference feature domain corrected for environmental parameters to determine attribution, generating a trigger request signal for the high-power infrared sensor. Upon receiving the trigger request signal, a surge immunity window is activated, and dynamic compensation is performed on the sampled values ​​within the immunity window. Simultaneously, based on the trigger frequency within the sliding time window and a dynamic upper limit determined by the remaining battery power, the operating mode of the high-power infrared sensor is adaptively switched. This invention enables low-power stable operation of multimodal gas sensors, avoids positive feedback loops formed by cross-interference and power supply disturbance coupling, and extends the device's battery life and component lifespan.

[0006] To achieve the above objectives, the present invention provides the following technical solution:

[0007] An asynchronous cooperative sampling control method for a multimodal gas sensor includes:

[0008] This is used to acquire real-time raw sampled values ​​of each electrochemical sentinel channel, extract the time of occurrence of the first occurrence of the signal change and signal deviation change exceeding the noise floor threshold of each electrochemical sentinel channel, construct a two-dimensional time-series fingerprint vector based on the time of occurrence of the first occurrence of the signal change and signal deviation change exceeding the noise floor threshold, acquire the corrected cross-interference feature domain, determine the attribution of the two-dimensional time-series fingerprint vector and the corrected cross-interference feature domain, and generate a trigger request signal based on the attribution determination result.

[0009] Upon receiving a trigger request signal, the surge immunity window is activated. During the surge immunity window, the original sampled values ​​of each electrochemical sentinel channel are acquired. Dynamic compensation is performed on the original sampled values ​​of each electrochemical sentinel channel during the immunity window. After the surge immunity window is closed, the current trigger frequency within the sliding time window is counted. Based on the comparison between the current trigger frequency and the dynamic upper limit of the trigger frequency determined by the current battery power, the working mode of the high-power infrared sensor is adaptively switched.

[0010] The method for detecting the event where the signal deviation change first exceeds the noise floor threshold includes:

[0011] The analog-to-digital conversion of multiple electrochemical sentinel channels is performed synchronously at a preset sentinel sampling period to acquire the real-time raw sample values ​​of each electrochemical sentinel channel. The real-time raw sample values ​​are written into the sliding data buffer corresponding to each electrochemical sentinel channel. Statistical calculations are performed on the sample values ​​in the sliding data buffer to obtain the current baseline estimate and current noise floor value of each electrochemical sentinel channel.

[0012] The current signal deviation change of each electrochemical sentinel channel is calculated based on the current baseline estimate. The event in which the signal deviation change of the electrochemical sentinel channel first exceeds the noise floor threshold is detected based on the current signal deviation change, and the time of occurrence of the event in which the signal deviation change first exceeds the noise floor threshold is recorded.

[0013] The method for detecting the event where the deviation of the electrochemical sentinel channel signal first exceeds the noise floor threshold is as follows:

[0014] The current signal deviation change of each electrochemical sentinel channel is compared with the corresponding noise floor threshold. When the current signal deviation change of the electrochemical sentinel channel exceeds the noise floor threshold and the electrochemical sentinel channel was previously in a silent state, it is determined that the electrochemical sentinel channel has experienced a signal deviation change exceeding the noise floor threshold for the first time.

[0015] The noise floor threshold is obtained by multiplying the current noise floor value by a preset noise multiple threshold;

[0016] The silent state refers to the situation where, in the previous several consecutive sentinel sampling cycles, the deviation of the current signal of the electrochemical sentinel channel has not exceeded the noise floor threshold of the corresponding electrochemical sentinel channel.

[0017] Within a preset correlation time window, the number of electrochemical sentinel channels that experience an event where the signal deviation change first exceeds the noise floor threshold is statistically determined, and the signal change of each electrochemical sentinel channel is extracted.

[0018] The method for statistically determining the number of electrochemical sentinel channels that first exceed the noise floor threshold within a preset correlation time window includes:

[0019] Within a preset associated time window, determine whether at least two electrochemical sentinel channels have experienced an event where the signal deviation change amount first exceeds the noise floor threshold. If so, read the occurrence time of the event where the signal deviation change amount first exceeds the noise floor threshold and calculate the signal change amount of each electrochemical sentinel channel. Construct a two-dimensional time-series fingerprint vector based on the signal change amount and the occurrence time of the event where the signal deviation change amount first exceeds the noise floor threshold.

[0020] If only a single electrochemical sentinel channel experiences a signal deviation change that exceeds the noise floor threshold for the first time, the decision on whether to directly generate a trigger request signal is based on the comparison between the signal change of that electrochemical sentinel channel and the preset alarm threshold.

[0021] The method for constructing the two-dimensional temporal fingerprint vector includes:

[0022] The ratio of signal changes between channels is calculated based on the signal changes of each electrochemical sentinel channel, and this ratio is used as the first dimension component of the two-dimensional time-series fingerprint vector. The time difference between channels is calculated based on the time when the signal deviation change of each electrochemical sentinel channel first exceeds the noise floor threshold, and this time difference is used as the second dimension component of the two-dimensional time-series fingerprint vector. The first dimension component and the second dimension component of the two-dimensional time-series fingerprint vector are combined to form the two-dimensional time-series fingerprint vector.

[0023] The method for obtaining the corrected cross-interference feature domain includes:

[0024] Load the cross-interference feature domain reference parameters, which include a first dimension reference interval and a second dimension reference interval;

[0025] Read the current ambient temperature and humidity values, and perform dynamic boundary correction on the first and second dimension reference intervals in the cross-interference feature domain reference parameters based on the current ambient temperature and humidity values, respectively, to obtain the corrected cross-interference feature domain composed of the first and second dimension correction intervals.

[0026] The method for determining the attribution of the two-dimensional temporal fingerprint vector and the corrected cross-interference feature domain includes:

[0027] Determine whether the first dimension component of the two-dimensional temporal fingerprint vector falls within the first dimension correction interval of the corrected cross-interference feature domain, and simultaneously determine whether the second dimension component of the two-dimensional temporal fingerprint vector falls within the second dimension correction interval of the corrected cross-interference feature domain.

[0028] Based on the attribution determination result, cross-interference events and independent gas events are distinguished, and a trigger request signal is generated when the event is determined to be an independent gas event.

[0029] The activation time of the surge immune window is:

[0030] A wake-up command is sent to the power supply control switch of the high-power infrared sensor and the time of the wake-up command is recorded. The surge immunity window is activated from the time the wake-up command is sent.

[0031] The method for dynamically compensating the original sampled values ​​during the immune window of each electrochemical sentinel channel includes:

[0032] Based on the wake-up command issuance time, the baseline offset estimate of each electrochemical sentinel channel at each sampling time is calculated. The original sampled value during the immune window period of each electrochemical sentinel channel is subtracted from the corresponding baseline offset estimate to obtain the sampled value after compensation during the immune window period.

[0033] The method for adaptively switching the operating mode of the high-power infrared sensor includes:

[0034] When the current trigger frequency does not exceed the dynamic upper limit of the trigger frequency, the high-power infrared sensor maintains the on-demand trigger-and-sleep mode. When the current trigger frequency exceeds the dynamic upper limit of the trigger frequency, the high-power infrared sensor is switched to the continuous low duty cycle cruise mode.

[0035] An asynchronous cooperative sampling control system for a multimodal gas sensor, used to implement the aforementioned asynchronous cooperative sampling control method for a multimodal gas sensor, the system comprising:

[0036] Trigger request generation module: used to acquire real-time raw sampled values ​​of each electrochemical sentinel channel, extract the occurrence time of the first time the signal change and signal deviation change of each electrochemical sentinel channel exceed the noise floor threshold, construct a two-dimensional time-series fingerprint vector based on the occurrence time of the first time the signal change and signal deviation change exceed the noise floor threshold; acquire the corrected cross-interference feature domain, determine the attribution of the two-dimensional time-series fingerprint vector and the corrected cross-interference feature domain, and generate a trigger request signal based on the attribution determination result;

[0037] Infrared mode switching module: Used to receive trigger request signals, activate surge immunity window, acquire raw sampled values ​​of each electrochemical sentinel channel during the surge immunity window, perform dynamic compensation on the raw sampled values ​​of each electrochemical sentinel channel during the immunity window, and count the current trigger frequency within the sliding time window after the surge immunity window is closed. Based on the comparison between the current trigger frequency and the dynamic upper limit of the trigger frequency determined by the current battery power, adaptively switch the working mode of the high-power infrared sensor.

[0038] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0039] This invention constructs a two-dimensional temporal fingerprint vector based on the moment when the signal change and signal deviation change of the electrochemical sentinel channel first exceed the noise floor threshold. This vector, combined with a modified cross-interference feature domain, is used to determine the attribution of the signal and generate a trigger request signal. This achieves attribute discrimination of the source of the electrochemical sentinel channel signal change, distinguishing between real gas events and cross-interference-derived events. It avoids invalid wake-up of high-power infrared sensors caused by cross-interference at the trigger signal generation stage. By activating the surge immunity window after receiving the trigger request signal and performing dynamic compensation on the original sampled values ​​during the immunity window period, it eliminates the power supply issues during the high-power infrared sensor startup process. Electrical disturbances interfere with the sampling baseline of the electrochemical sentinel channel, blocking the positive feedback loop formed by the coupling of gas cross-sensitivity and power supply disturbances, thus avoiding signal loop self-locking in the system. By statistically analyzing the current trigger frequency within the sliding time window and combining it with the dynamic upper limit of the trigger frequency determined by the current remaining battery power, a comparison is completed. Based on the comparison results, the operating mode of the high-power infrared sensor is adaptively switched. This enables the matching and control of the infrared sensor's operating mode with the battery power status and the trigger frequency of gas events. While maintaining the gas detection response performance, the device power consumption is reasonably allocated, reducing the impact of frequent power-ups on the infrared sensor on device aging and ensuring the stability of the device's endurance. Attached Figure Description

[0040] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. 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.

[0041] Figure 1 A flowchart illustrating an asynchronous cooperative sampling control method for a multimodal gas sensor provided in this embodiment of the invention;

[0042] Figure 2 This is a schematic diagram of the sensor layout of a portable four-in-one detector provided in an embodiment of the present invention;

[0043] Figure 3 This is a schematic diagram illustrating the cross-interference phenomenon of H2S gas on the CO sensor provided in an embodiment of the present invention;

[0044] Figure 4 This is a flowchart of the event detection method for the first time when the signal deviation change exceeds the noise floor threshold, provided in an embodiment of the present invention.

[0045] Figure 5 This is a schematic diagram comparing two operating modes of the high-power infrared sensor provided in an embodiment of the present invention;

[0046] Figure 6 This is a functional block diagram of an asynchronous cooperative sampling control system for a multimodal gas sensor provided in an embodiment of the present invention. Detailed Implementation

[0047] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0048] Example 1

[0049] Please see Figure 1 As shown, this embodiment provides an asynchronous cooperative sampling control method for a multimodal gas sensor, including:

[0050] Step S10: Obtain the real-time raw sampling values ​​of each electrochemical sentinel channel, extract the occurrence time of the first time the signal change and signal deviation change of each electrochemical sentinel channel exceed the noise floor threshold, construct a two-dimensional time-series fingerprint vector based on the occurrence time of the first time the signal change and signal deviation change exceed the noise floor threshold; obtain the corrected cross-interference feature domain, determine the attribution of the two-dimensional time-series fingerprint vector and the corrected cross-interference feature domain, and generate a trigger request signal based on the attribution determination result;

[0051] Specifically, step S10 constructs a two-dimensional time-series fingerprint vector by extracting the temporal sequence and amplitude ratio of signal changes in each electrochemical sentinel channel. This two-dimensional time-series fingerprint vector is then compared with the cross-interference feature domain corrected for environmental temperature and humidity to determine the signal source attribute. (See also...) Figure 2 This is a schematic diagram of the sensor layout of the portable four-in-one detector provided in the embodiments of this application. Figure 2 The diagram illustrates the overall hardware architecture of a portable four-in-one detector. Its core detection unit is divided into two main functional modules: a normally open electrochemical sentinel channel and a high-power infrared sensor. Both modules are powered by a shared battery. The electrochemical sentinel channel refers to a gas detection channel that uses an electrochemical sensor as its sensing element. In a portable four-in-one detector, this typically includes a CO channel for detecting carbon monoxide, an H2S channel for detecting hydrogen sulfide, and an O2 channel for detecting oxygen. Figure 2 As shown, the three gas detection channels are all integrated into the electrochemical sentinel channel module. The three electrochemical sentinel channels operate continuously as low-power normally open sentinels, and their power consumption is much lower than that of the infrared sensor used to detect methane. Figure 2 The document also clarifies the differences in operating modes between the two types of detection modules: the electrochemical sentinel channel operates in a low-power continuous mode, while the module used for methane detection... The infrared sensor has a built-in infrared light source and operates in a high-power, on-demand wake-up mode. Cross-interference refers to the phenomenon where an electrochemical sensor responds to a non-target gas; see [link to relevant documentation]. Figure 3 This is a schematic diagram illustrating the cross-interference phenomenon of H2S gas on the CO sensor provided in the embodiments of this application. Figure 3The diagram illustrates the difference in response of a single H2S gas source to different electrochemical sensors. The H2S gas source produces a direct response to the H2S sensor (target gas is H2S), forming a true response; however, it causes cross-interference to the CO sensor (target gas is CO), forming a spurious response. For example, the electrochemical CO sensor, in addition to being sensitive to CO gas, also exhibits positive cross-sensitivity to H2S gas. When H2S gas is present in the environment, the CO channel also produces a signal response. This response does not originate from the actual presence of CO gas, but rather from the electrochemical reaction of H2S gas molecules on the surface of the CO sensor's electrolyte. The two-dimensional temporal fingerprint vector is a two-dimensional vector formed by combining the ratio of signal changes and the difference in the timing of the first occurrence of the signal deviation exceeding the noise floor threshold. The ratio of signal changes reflects the relative relationship of the response intensity of each electrochemical sentinel channel, and the difference in the timing of the first occurrence of the signal deviation exceeding the noise floor threshold reflects the temporal order of the responses of each electrochemical sentinel channel. The cross-interference feature domain is a rectangular area on a two-dimensional plane jointly defined by the first-dimensional correction interval and the second-dimensional correction interval. Two-dimensional time-series fingerprint vectors falling within this rectangular area correspond to cross-interference events, while two-dimensional time-series fingerprint vectors falling outside this rectangular area correspond to independent gas events. The trigger request signal is a control signal output from step S10 to step S20, used to indicate that the high-power infrared sensor needs to be woken up to perform methane concentration sampling.

[0052] Step S10 employs a two-dimensional temporal fingerprint vector matching method for cross-interference identification. Compared to existing technologies that only set independent thresholds for the amplitude or rate of change of signals from each electrochemical sentinel channel, this method utilizes the correlation characteristics between the changes in signals from each electrochemical sentinel channel for determination. The independent threshold determination method only focuses on whether the signal of a single electrochemical sentinel channel exceeds a preset threshold, failing to acquire the temporal sequence and amplitude ratio relationships between the changes in signals from each electrochemical sentinel channel. Therefore, it cannot distinguish whether the signal changes originate from cross-interference from the same gas source or from real gas events generated independently by multiple gases. The two-dimensional temporal fingerprint vector integrates the temporal sequence and amplitude ratio relationships into a single two-dimensional feature, making cross-interference events and independent gas events exhibit distinguishable spatial distributions on a two-dimensional plane. The two-dimensional temporal fingerprint vectors of cross-interference events cluster within a specific region determined by the sensor's cross-sensitivity characteristics, while the two-dimensional temporal fingerprint vectors of independent gas events are distributed outside this specific region. Step S10 changes the judgment criterion from single-channel independent threshold comparison to region attribution judgment on a two-dimensional plane, enabling the system to accurately identify the signal source attribute when signal changes occur simultaneously in multiple electrochemical sentinel channels. Trigger request signals are only generated when the event is determined to be an independent gas event, thereby avoiding cross-interference that causes high-power infrared sensors to be frequently woken up. This judgment logic can effectively reduce the ineffective power consumption of the shared battery and extend the overall battery life of the portable four-in-one detector.

[0053] Further, step S10 includes:

[0054] Step S11: Simultaneously perform analog-to-digital conversion on multiple electrochemical sentinel channels at a preset sentinel sampling period to acquire the real-time raw sampling values ​​of each electrochemical sentinel channel. Write the real-time raw sampling values ​​into the sliding data buffer corresponding to each electrochemical sentinel channel. Perform statistical calculations on the sampling values ​​in the sliding data buffer to obtain the current baseline estimate and current noise floor value of each electrochemical sentinel channel.

[0055] Specifically, step S11 synchronously performs analog-to-digital conversion (ADC) acquisition on the CO, H2S, and O2 channels at a preset sentinel sampling period to obtain real-time raw sampled values ​​for each electrochemical sentinel channel, forming a real-time raw sampled value sequence. The sentinel sampling period refers to the time interval between the system performing one ADC acquisition on each electrochemical sentinel channel. This time interval must meet two constraints: first, the sentinel sampling period should be short enough to capture the dynamic process of gas concentration changes, ensuring the system can promptly detect the event where the signal deviation first exceeds the noise floor threshold; second, the sentinel sampling period should be long enough to control the acquisition power consumption of the electrochemical sentinel channels, avoiding excessively frequent ADCs that lead to rapid battery consumption. For example, the sentinel sampling period can be set to a fixed value within the range of hundreds of milliseconds to several seconds, with the specific value determined based on the response speed characteristics of the electrochemical sensor and the power consumption budget of the detector. ADC acquisition refers to the process of converting the analog current or voltage signal output by the electrochemical sensor into a digital quantity. This process is achieved through the built-in ADC of the detector, and the resolution of the ADC determines the quantization accuracy of the sampled values.

[0056] The system maintains a sliding data buffer for each electrochemical sentinel channel. This sliding data buffer is a first-in, first-out (FIFO) data storage structure, with its length set to cover the number of sampling points for dozens to hundreds of complete sentinel sampling cycles. The length of the sliding data buffer must consider the following factors: if the length is too short, the number of sampling values ​​in the buffer will be insufficient to reflect the noise statistics of the electrochemical sensor, leading to large fluctuations in the calculated noise baseline value; if the length is too long, the buffer may contain sampling values ​​from earlier times. If the environmental conditions at that time differ from the current time, the current noise baseline value cannot accurately reflect the noise level under the current environmental conditions. For example, the length of the sliding data buffer can be set to cover the number of sampling points for dozens to hundreds of sentinel sampling cycles, with the specific value determined based on the noise characteristics of the electrochemical sensor and the rate of change of environmental conditions. When each new sampling value arrives, the system writes the new sampling value to the end of the sliding data buffer for the corresponding electrochemical sentinel channel, while simultaneously removing the oldest sampling value from the sliding data buffer, ensuring that the sliding data buffer always stores sampling values ​​from the most recent dozens to hundreds of sentinel sampling cycles.

[0057] After each sentinel sampling cycle, the system performs statistical operations on all sampled values ​​stored in the sliding data buffer of each electrochemical sentinel channel to obtain the current baseline estimate and the current noise floor value. The current baseline estimate is calculated as the arithmetic mean of all sampled values ​​in the sliding data buffer. This arithmetic mean smooths out random fluctuations in the sampled values, providing a stable baseline reference point. The current noise floor value is calculated as the arithmetic mean of the absolute values ​​of the differences between each sampled value in the sliding data buffer and the current baseline estimate. The physical meaning of the current noise floor value is the average deviation of the output signal of the electrochemical sentinel channel from the current baseline estimate; it reflects the background fluctuation amplitude of the electrochemical sentinel channel under current environmental temperature and humidity conditions and current sensor status. The current noise baseline value has significant advantages compared to traditional fixed factory noise parameters: the noise characteristics of electrochemical sensors change with temperature and sensor aging, and fixed factory noise parameters cannot adapt to this change, which may lead to underestimation of noise at high temperatures, resulting in false detections, or overestimation of noise at low temperatures, resulting in missed real gas events. Using a sliding data buffer for dynamic updates ensures that the current noise baseline value always reflects the true noise level of the sensor under its current operating conditions. The current baseline estimate and current noise baseline value of each electrochemical sentinel channel are stored in the corresponding baseline register and noise baseline register of the electrochemical sentinel channel, respectively, for subsequent reading and use in step S12. The baseline register and noise baseline register are internal data storage units of the system, and their storage capacity only needs to hold one value, being overwritten by the new calculation result after each sentinel sampling cycle.

[0058] Step S11 uses a sliding data buffer to dynamically calibrate the current noise baseline value, enabling subsequent step S12 to obtain a noise reference that is adaptively updated as environmental conditions change and the sensor ages. Without the dynamic calibration in step S11, the comparison between the signal deviation change and the noise baseline threshold in step S12 will lack an accurate reference. This could lead to false detections of the first signal deviation change exceeding the noise baseline threshold in a calm environment without gas events because the current noise baseline value is lower than the actual background fluctuation amplitude of the electrochemical sentinel channel under the current environmental conditions. Conversely, in the presence of gas events, the first signal deviation change exceeding the noise baseline threshold event might be missed because the current noise baseline value is higher than the actual background fluctuation amplitude of the electrochemical sentinel channel under the current environmental conditions. This would affect the accuracy of the two-dimensional temporal fingerprint vector construction in step S13 and the reliability of the attribution determination in step S15.

[0059] See Figure 4Step S12: Calculate the current signal deviation change of each electrochemical sentinel channel based on the current baseline estimate. Detect the event where the signal deviation change of the electrochemical sentinel channel first exceeds the noise baseline threshold based on the current signal deviation change, and record the occurrence time of this event. Within a preset correlation time window, statistically determine the number of electrochemical sentinel channels where the signal deviation change first exceeds the noise baseline threshold, and extract the signal change of each electrochemical sentinel channel.

[0060] The method for detecting the event that the change in signal deviation of an electrochemical sentinel channel exceeds the noise floor threshold for the first time is as follows: the current change in signal deviation of each electrochemical sentinel channel is compared with the noise floor threshold obtained by multiplying the corresponding current noise floor value by a preset noise multiple threshold. When the current change in signal deviation of an electrochemical sentinel channel exceeds the noise floor threshold and the electrochemical sentinel channel was previously in a silent state, it is determined that the electrochemical sentinel channel has experienced the event that the change in signal deviation exceeds the noise floor threshold for the first time.

[0061] The method for statistically determining the number of electrochemical sentinel channels that have experienced a signal deviation change exceeding the noise floor threshold for the first time within a preset associated time window includes: determining whether at least two electrochemical sentinel channels have experienced a signal deviation change exceeding the noise floor threshold for the first time within the preset associated time window; if so, reading the occurrence time of the event and calculating the signal change of each electrochemical sentinel channel; transmitting the signal change and the occurrence time of the event to step S13; and constructing a two-dimensional temporal fingerprint vector based on the signal change and the occurrence time of the event; if only a single electrochemical sentinel channel has experienced a signal deviation change exceeding the noise floor threshold for the first time, determining whether to directly generate a trigger request signal based on the comparison result of the signal change of the electrochemical sentinel channel with a preset alarm threshold.

[0062] The signal change refers to the signed difference between the current new sample value of each electrochemical sentinel channel and the current baseline estimate of the corresponding electrochemical sentinel channel. The current new sample value refers to the value written to the sliding data buffer by each electrochemical sentinel channel at the current sampling time. Under normal acquisition conditions, the current new sample value is the real-time raw sample value obtained in step S11, and during the surge immune window, the current new sample value is the compensated sample value during the immune window period output in step S22.

[0063] Step S12 calculates the current signal deviation and the change in current signal deviation for each electrochemical sentinel channel upon arrival of each new sampled value. The current signal deviation is calculated by taking the absolute value of the difference between the current new sampled value and the current baseline estimate of the electrochemical sentinel channel. The current baseline estimate is read from the baseline register of the electrochemical sentinel channel in step S11. The current signal deviation reflects the degree to which the output signal of the electrochemical sentinel channel deviates from the baseline at the current sampling time. Taking the absolute value ensures that the current signal deviation only reflects the degree of deviation without distinguishing the direction of deviation, because regardless of whether the signal deviates upwards or downwards, a change in the degree of deviation indicates a possible change in gas concentration. Simultaneously, the current signal deviation is stored in the signal deviation register of the electrochemical sentinel channel for use in the next sentinel sampling cycle.

[0064] The current signal deviation change is calculated as follows: the difference between the current signal deviation of the electrochemical sentinel channel and the signal deviation stored in the previous sentinel sampling period is used to obtain the current signal deviation change of the electrochemical sentinel channel. The current signal deviation change represents the magnitude of the signal deviation change between adjacent sampling periods. Since the sentinel sampling period is a fixed constant, the difference in signal deviation between adjacent sampling periods is strictly proportional to the rate of change of signal deviation per unit time. Therefore, the current signal deviation change can be directly compared with the noise floor threshold derived from the current noise floor value using the same dimensions, reflecting the rate of change in the degree of signal deviation from the baseline. Using the change in signal deviation, rather than the signal deviation itself, as the detection criterion for the first time the change in signal deviation exceeds the noise floor threshold has the following advantages: when the gas concentration changes slowly, the signal deviation may remain near a certain value for a long time, making it difficult to determine whether there is a meaningful gas event based on the deviation at a single moment; the change in signal deviation can capture the dynamic change process of the signal deviation, producing a significant numerical change the instant the gas concentration begins to rise or fall, enabling the system to detect the occurrence of gas events earlier.

[0065] The current signal deviation change of the electrochemical sentinel channel is compared with the noise floor threshold of the electrochemical sentinel channel. The noise floor threshold is calculated by multiplying the current noise floor value of the electrochemical sentinel channel by a preset noise multiple threshold. The current noise floor value is read from the noise floor register of the electrochemical sentinel channel in step S11. The noise multiple threshold is a preset constant used to define the boundary between meaningful signal changes and noise fluctuations. The setting of the noise multiple threshold needs to consider the following factors: if the noise multiple threshold is too small, the noise floor threshold will be too low, and normal noise fluctuations may be misjudged as the first time the signal deviation change exceeds the noise floor threshold, leading to an increase in false detections; if the noise multiple threshold is too large, the noise floor threshold will be too high, and the signal deviation change of low-concentration gas events may not exceed the noise floor threshold, leading to an increase in omissions. For example, the noise multiple threshold can be set to a value greater than 1 and less than 10, and the specific value is determined according to the signal-to-noise ratio of the electrochemical sensor and the sensitivity requirements of the detector.

[0066] If the current signal deviation of the electrochemical sentinel channel exceeds the noise floor threshold of the electrochemical sentinel channel, and the electrochemical sentinel channel was previously in a silent state, the system determines that the signal deviation of the electrochemical sentinel channel has exceeded the noise floor threshold for the first time. A silent state refers to a situation where the current signal deviation of the electrochemical sentinel channel has not exceeded the noise floor threshold for several consecutive sentinel sampling periods (e.g., 3 to 10). For example, the determination of a silent state can use 5 consecutive sentinel sampling periods as the determination window, with the specific value determined based on the response speed characteristics of the electrochemical sensor and the fluctuation frequency of the background noise. The purpose of determining a previously silent state is to distinguish between the event of the signal deviation exceeding the noise floor threshold for the first time and the event of the signal deviation continuously exceeding the noise floor threshold, ensuring that the recorded event is the starting moment when the signal deviation jumps from the noise level to exceeding the noise floor threshold, rather than an intermediate moment when the signal deviation remains above the noise floor threshold. Accurate recording of the moment when the signal deviation change first exceeds the noise floor threshold is a prerequisite for constructing the second dimension component of the two-dimensional temporal fingerprint vector in subsequent step S13. If the recorded moment is not the first exceedance moment but an intermediate moment, the time difference between each electrochemical sentinel channel will deviate from its true temporal order, causing distortion of the second dimension component of the two-dimensional temporal fingerprint vector. After determining that the signal deviation change first exceeds the noise floor threshold, the system records the precise moment of this event in the change time register of the electrochemical sentinel channel.

[0067] The system statistically determines the number of electrochemical sentinel channels whose signal deviation exceeds the noise baseline threshold for the first time within a preset correlation time window. The correlation time window refers to the time range used to determine whether multiple electrochemical sentinel channels exhibit coordinated changes; its length is set to cover the possible delay time of the electrochemical sensor's cross-interference response. The setting of the correlation time window length needs to consider the following factors: if the length is too short, multi-channel coordinated changes that should be cross-interference may be judged as independent single-channel changes, because cross-interference responses require a certain amount of time to appear on the interfering channels; if the length is too long, multiple independent gas events that are indeed unrelated in time may be misjudged as coordinated changes, causing the two-dimensional temporal fingerprint vector constructed in subsequent step S13 to have no physical meaning. For example, the length of the correlation time window can be set to a fixed value within the range of several seconds to tens of seconds, with the specific value determined based on the cross-interference response delay characteristics of the electrochemical sensor.

[0068] When at least two electrochemical sentinel channels experience a signal deviation change exceeding the noise baseline threshold for the first time within a preset associated time window, the system reads the occurrence time of the first signal deviation change exceeding the noise baseline threshold stored in the change time register of each electrochemical sentinel channel, and calculates the signal change of each electrochemical sentinel channel. The signal change is calculated as follows: the signed value of the difference between the current new sample value of each electrochemical sentinel channel and the current baseline estimate of that electrochemical sentinel channel is taken as the signal change of that electrochemical sentinel channel. Unlike the current signal deviation, the signal change retains sign information; a positive value indicates that the signal has shifted upward relative to the baseline, and a negative value indicates that the signal has shifted downward relative to the baseline. The purpose of retaining the sign information is that when constructing the first dimension component of the two-dimensional time-series fingerprint vector in step S13, it is necessary to calculate the ratio of signal changes between channels. The sign of this ratio can reflect the consistency or oppositeity of the signal change direction of each electrochemical sentinel channel, thus providing additional judgment criteria. The system transmits the moment when the signal change of each electrochemical sentinel channel and the moment when the signal deviation change of each electrochemical sentinel channel first exceeds the noise floor threshold to step S13.

[0069] If, within the associated time window, only a single electrochemical sentinel channel experiences a signal deviation exceeding the noise baseline threshold for the first time, the system determines that the signal change originates from an independent response of a single gas source, eliminating the need for cross-interference discrimination. The processing logic in this case is as follows: A trigger request signal is generated based directly on whether the signal change of the single electrochemical sentinel channel exceeds its preset alarm threshold. If it exceeds the preset alarm threshold, a trigger request signal is generated and passed to step S20. If it does not exceed the preset alarm threshold but exceeds the preset attention threshold of the electrochemical sentinel channel, the event is marked as a single-channel attention event and monitoring continues without generating a trigger request signal. The preset alarm threshold is the critical value for the signal change that triggers an alarm for each electrochemical sentinel channel. The preset attention threshold is an intermediate critical value below the preset alarm threshold but above the noise level. Setting the preset attention threshold allows the system to maintain attention on signal changes that have not yet reached the alarm condition but are significantly above the noise level. For example, for the CO electrochemical sentinel channel, the preset alarm threshold can be set to the signal change corresponding to a CO concentration of 24 ppm to 35 ppm, and the preset attention threshold can be set to the signal change corresponding to a CO concentration of 10 ppm to 20 ppm. For the H2S electrochemical sentinel channel, the preset alarm threshold can be set to the signal change corresponding to a H2S concentration of 10 ppm to 15 ppm, and the preset attention threshold can be set to the signal change corresponding to a H2S concentration of 4 ppm to 8 ppm. For the O2 electrochemical sentinel channel, the preset alarm threshold can be set to the signal change corresponding to an oxygen volume fraction deviating from the normal atmospheric concentration by 2% to 3%, and the preset attention threshold can be set to the signal change corresponding to an oxygen volume fraction deviating from the normal atmospheric concentration by 0.5% to 1.5%. The specific values ​​are determined based on the sensitivity characteristics of the electrochemical sensor and the alarm level requirements of the detector. The method of processing single-channel independent changes directly into the threshold determination process avoids performing unnecessary cross-interference analysis calculations on single-channel events, because the occurrence of cross-interference necessarily involves the linkage changes of at least two electrochemical sentinel channels, and there is no need for cross-interference determination for single-channel changes. Step S12 uses a time window-based conditional judgment to initially separate multi-channel linked changes from single-channel independent changes. Multi-channel linked changes proceed to the cross-interference identification process in step S13, while single-channel independent changes directly enter the threshold determination process. This separation mechanism concentrates the system's computing resources on multi-channel linked change events that require cross-interference identification, while employing a simplified processing flow for single-channel independent change events that do not require cross-interference identification. This reduces the average computational load of the system while ensuring the integrity of the detection function.

[0070] Step S13: Calculate the ratio of signal changes between channels based on the signal changes of each electrochemical sentinel channel transmitted in step S12, and use the ratio of signal changes between channels as the first dimension component of the two-dimensional time-series fingerprint vector; calculate the time difference between channels based on the time of occurrence of the event where the signal deviation change of each electrochemical sentinel channel first exceeds the noise floor threshold transmitted in step S12, and use this as the second dimension component of the two-dimensional time-series fingerprint vector; combine the first dimension component and the second dimension component of the two-dimensional time-series fingerprint vector to form the two-dimensional time-series fingerprint vector.

[0071] Specifically, step S13 constructs a two-dimensional temporal fingerprint vector for cross-interference identification based on the timing of the first occurrence of the signal change and signal deviation change events of each electrochemical sentinel channel exceeding the noise floor threshold, as transmitted in step S12. The system receives the signal change amounts of the CO channel and H2S channel, as well as the timing of the first occurrence ...

[0072] Before calculating the first dimension component of the two-dimensional time-series fingerprint vector, a validity determination of the H2S channel signal change is performed. The method for determining the validity of the H2S channel signal change is as follows: It is determined whether the absolute value of the H2S channel signal change is less than the current noise floor value of the H2S channel. If the absolute value of the H2S channel signal change is less than the current noise floor value, the signal change of the H2S channel is close to zero, and a valid amplitude ratio cannot be calculated. The H2S channel signal change is invalid, and the system determines that the current multi-channel linkage change is not caused by H2S cross-interference. This is because the premise of H2S cross-interference is that the H2S channel, as the interference source channel, has a clear signal change. The system directly marks this multi-channel linkage change as an event to be further observed and returns to step S12 to continue collecting subsequent data. This validity determination avoids the situation where the amplitude ratio calculation result tends to infinity or produces numerical instability when the divisor is close to zero, ensuring the numerical reliability of the first dimension component of the two-dimensional time-series fingerprint vector. If the absolute value of the H2S channel signal change is not less than the current noise floor value of the H2S channel, the construction of the two-dimensional time-series fingerprint vector continues.

[0073] The first dimension component of the two-dimensional time-series fingerprint vector is calculated as follows: the ratio of the signal change in the CO channel to the signal change in the H2S channel is taken as the first dimension component of the two-dimensional time-series fingerprint vector. The physical meaning of this first dimension component is the ratio of the CO sensor's response intensity to the gas source causing the signal change to the H2S sensor's response intensity to the same gas source. If the signal change is indeed caused by H2S cross-interference, this ratio should be approximately equal to the CO sensor's cross-sensitivity to H2S. Cross-sensitivity refers to the ratio between the equivalent target gas indication value generated by the electrochemical sensor in a calibrated concentration of non-target gas and the actual concentration of the non-target gas. The equivalent target gas indication value is the target gas concentration value obtained by converting the signal output by the electrochemical sensor in an environment only exposed to a non-target gas into the target gas concentration conversion formula of the sensor itself. This value does not represent the actual target gas concentration in the environment, but rather reflects the false target gas reading generated by the electrochemical sensor's response to the non-target gas. The actual non-target gas concentration refers to the true concentration value of the non-target gas applied to the electrochemical sensor, pre-calibrated using a standard gas mixing device. The equivalent target gas indication value and the actual concentration of the non-target gas are obtained as follows: During the factory calibration stage, a non-target standard gas of a preset concentration is applied to the electrochemical sensor as the actual concentration of the non-target gas. Simultaneously, the concentration value output by the electrochemical sensor through its internal target gas concentration conversion formula is read as the equivalent target gas indication value. The ratio of the two is the cross-sensitivity of the electrochemical sensor to that non-target gas. Since cross-sensitivity is an inherent characteristic of electrochemical sensors and has a relatively stable numerical range under the same environmental conditions, whether the first dimension component of the two-dimensional time-series fingerprint vector falls within the numerical range corresponding to the cross-sensitivity can be used as one of the criteria for determining cross-interference events.

[0074] The second dimension of the two-dimensional temporal fingerprint vector is calculated as follows: the difference between the time when the signal deviation change of the CO channel first exceeds the noise floor threshold and the time when the signal deviation change of the H2S channel first exceeds the noise floor threshold is taken as the second dimension of the two-dimensional temporal fingerprint vector. The physical meaning of this second dimension is the time lag of the CO channel response relative to the H2S channel response. If the signal change is indeed caused by H2S cross-interference, this time difference should be positive, meaning the CO channel response lags behind the H2S channel response. This is because, in the physical mechanism of cross-interference, H2S gas molecules must first reach the H2S sensor and generate a response, and then diffuse to the electrolyte surface of the CO sensor to cause the cross-interference response of the CO channel. This diffusion process requires a certain amount of time. The magnitude of the second dimension of the two-dimensional temporal fingerprint vector depends on the time required for H2S molecules to diffuse from the electrolyte surface of the H2S sensor to the electrolyte surface of the CO sensor. This diffusion time is affected by the physical distance between the sensors and the ambient temperature, but is usually within a predictable range. If the second dimension component of the two-dimensional time-series fingerprint vector is negative, meaning that the CO channel response appears before the H2S channel response, it indicates that the CO channel response cannot be caused by H2S cross-interference, because the cross-interference response cannot appear before the original response of the interference source channel.

[0075] The system combines the first and second dimensions of the two-dimensional time-series fingerprint vector into a single two-dimensional time-series fingerprint vector. This construction condenses the dispersed numerical information of signal changes in each electrochemical sentinel channel into a compact two-dimensional feature representation. This allows subsequent step S15 to complete the cross-interference identification task, which originally required multi-dimensional independent comparisons, on a two-dimensional plane through simple region attribution determination. If only the first or second dimension of the two-dimensional time-series fingerprint vector is used for determination, there is insufficient discriminative power: when judging solely based on the first dimension, the signal change in the CO channel caused by H2S cross-interference may be on the same order of magnitude as the signal change in the CO channel caused by independent low-concentration CO leakage, making it impossible to distinguish between the two scenarios; when judging solely based on the second dimension, the situation where independent CO leakage occurs precisely after the H2S channel signal change due to other reasons may have a similar temporal relationship to cross-interference. Combining the two dimensions into a two-dimensional vector results in cross-interference events and independent gas events exhibiting different distribution regions on the two-dimensional plane, significantly enhancing the discriminative power.

[0076] Furthermore, if the multi-channel linkage change transmitted in step S12 involves the O2 channel, i.e., the O2 channel also experiences a signal deviation change exceeding the noise floor threshold for the first time within the associated time window, the system additionally constructs a two-dimensional time-series fingerprint vector between the O2 channel and the H2S channel, and a two-dimensional time-series fingerprint vector between the O2 channel and the CO channel. The two-dimensional time-series fingerprint vectors between each pair of electrochemical sentinel channels are calculated in the same way, i.e., the ratio of signal changes and the difference in the time of occurrence of the signal deviation change exceeding the noise floor threshold are calculated respectively. The rationale for constructing a two-dimensional time-series fingerprint vector between the O2 channel and other electrochemical sentinel channels is that: in a real leak scenario, methane replacement of O2 will cause a decrease in the O2 channel signal, but there is no cross-sensitivity relationship between this O2 decrease signal and the cross-interference signal of the CO channel or H2S channel. Therefore, the two-dimensional time-series fingerprint vector between the O2 channel and other electrochemical sentinel channels will inevitably deviate from the cross-interference feature domain and can be used as an auxiliary discrimination criterion. If the two-dimensional temporal fingerprint vector construction in step S13 is missing, step S15 will lack the object to be determined for attribution determination, and the attribution determination operation cannot be performed, and the cross-interference identification function will not be realized.

[0077] Step S14: Load the cross-interference feature domain reference parameters pre-stored during the factory calibration stage. The cross-interference feature domain reference parameters include a first-dimensional reference interval and a second-dimensional reference interval. Read the current ambient temperature value and the current ambient humidity value. Perform dynamic boundary correction on the first-dimensional reference interval and the second-dimensional reference interval in the cross-interference feature domain reference parameters according to the current ambient temperature value and the current ambient humidity value to obtain the corrected cross-interference feature domain composed of the first-dimensional correction interval and the second-dimensional correction interval.

[0078] Specifically, the first dimension of the reference interval is defined by the nominal cross-sensitivity center value of the CO sensor to H2S, and the upper and lower deviations of this nominal cross-sensitivity center value under standard temperature and humidity conditions. The nominal cross-sensitivity center value is the arithmetic mean of the ratio of the change in CO channel signal to the change in H2S channel signal obtained from repeated tests of the CO sensor with the same known concentration of H2S standard gas under standard temperature and humidity conditions. It reflects the statistical center of the CO sensor's cross-response to H2S gas under standard temperature and humidity conditions. The second dimension of the reference interval is defined by the nominal time lag center value of the CO channel response relative to the H2S channel response, and the upper and lower deviations of this nominal time lag center value under standard temperature and humidity conditions. The nominal time lag center value refers to the arithmetic mean of the differences between the times when the CO channel signal deviation changes and the H2S channel signal deviation changes occur in repeated tests of the CO and H2S sensors under standard temperature and humidity conditions. This value reflects the statistical center of the time required for H2S gas molecules to diffuse from the surface of the H2S sensor electrolyte to the surface of the CO sensor electrolyte. The first and second dimension reference intervals together define a rectangular area of ​​the cross-interference characteristic domain under standard conditions on a two-dimensional plane. The method for obtaining the cross-interference characteristic domain reference parameters during the factory calibration stage is as follows: Under standard temperature and humidity conditions, a known concentration of H2S standard gas is applied to the sensor, and the signal responses of the H2S and CO channels are recorded simultaneously. Through repeated experiments, the statistical distribution of the ratio of the CO channel signal change to the H2S channel signal change, as well as the statistical distribution of the difference between the CO channel response time and the H2S channel response time, are obtained. The center value, upper deviation, and lower deviation are extracted from the statistical distributions and pre-stored in the device memory. Standard temperature and humidity conditions refer to the ambient temperature and humidity reference values ​​used during factory calibration. They are usually set as a fixed temperature value and a fixed relative humidity value under room temperature conditions.

[0079] The system reads the current ambient temperature value from the device's built-in temperature sensor and the current ambient humidity value from the built-in humidity sensor. Based on these values, dynamic boundary corrections are performed on the first and second dimension reference intervals. The physical basis for this dynamic boundary correction is that changes in temperature and humidity cause drift in the cross-sensitivity and response delay of the electrochemical sensors. Increased temperature accelerates the electrochemical reaction rate, potentially increasing cross-sensitivity and shortening response delay; decreased temperature slows the electrochemical reaction rate, potentially decreasing cross-sensitivity and lengthening response delay. Increased humidity alters the electrolyte's water content, affecting ionic conductivity and consequently the cross-sensitivity value.

[0080] For the dynamic boundary correction of the first-dimensional reference interval, when the current ambient temperature is higher than the standard temperature during factory calibration, the upper boundary of the first-dimensional reference interval is expanded in an increasing direction. The expansion amount is proportional to the difference between the current ambient temperature and the standard temperature. This correction law is based on the Arrhenius temperature dependence of the electrochemical reaction rate. For each preset temperature step value increase, the upper deviation of the cross-sensitivity increases by the preset sensitivity step value. The temperature step value refers to the temperature change corresponding to one expansion of the first-dimensional reference interval boundary. It is set as follows: based on the cross-sensitivity drift curves measured at different temperature points during the factory calibration phase of the electrochemical sensor, the drift curve is approximately fitted into a linear segment, and the temperature step value is taken as the temperature independent variable step size of this linear segment. The sensitivity step value refers to the numerical amount by which the boundary of the first-dimensional reference interval expands when the temperature changes by one temperature step value, and its value is equal to the cross-sensitivity drift corresponding to the temperature step value in the aforementioned linear segment. For example, the temperature step value can be set to a fixed value within the range of 5℃ to 10℃, and the corresponding sensitivity step value can be set to a fixed value within the range of 1% to 3% of the nominal cross-sensitivity center value. The specific values ​​are determined based on the temperature sensitivity characteristics of the electrochemical sensor and the factory calibration data. When the current ambient temperature is lower than the standard temperature, the lower boundary of the first-dimensional reference interval is expanded in the decreasing direction, and the expansion amount is determined according to the same rule. For the dynamic boundary correction of the second-dimensional reference interval, when the current ambient temperature increases, molecular diffusion accelerates, resulting in a shorter response delay. Therefore, the lower boundary of the second-dimensional reference interval is expanded towards a smaller value. When the current ambient temperature decreases, molecular diffusion slows down, resulting in a longer response delay. Therefore, the upper boundary of the second-dimensional reference interval is expanded towards a larger value. After dynamic boundary correction, the first-dimensional correction interval and the second-dimensional correction interval are obtained. The first-dimensional correction interval and the second-dimensional correction interval together constitute the corrected cross-interference feature domain. The corrected cross-interference feature domain remains a rectangular area on the two-dimensional plane, but its boundaries have been adaptively adjusted according to the current ambient temperature and humidity, reflecting the reasonable fluctuation range of the sensor's cross-interference response under the current actual environmental conditions. The system then passes this corrected cross-interference feature domain to step S15.

[0081] Step S14 employs dynamic boundary correction to change the boundary of the cross-interference feature domain from a fixed factory nominal value to an adaptive value that dynamically adjusts with ambient temperature and humidity. If the boundary of the cross-interference feature domain is not corrected for temperature and humidity, under high temperature and humidity conditions, the two-dimensional temporal fingerprint vector of the actual cross-interference signal may drift out of the fixed boundary of the cross-interference feature domain, leading to misjudgment as an independent gas event and false triggering. Under low temperature and humidity conditions, the two-dimensional temporal fingerprint vector of an independent gas event may fall into the fixed boundary of the cross-interference feature domain, being misjudged as cross-interference and missing the actual gas event. Dynamic boundary correction ensures that the accuracy of cross-interference identification remains at an acceptable level under different temperature and humidity conditions. For example, the temperature and humidity difference between the high temperature and humidity of municipal underground operations in summer and the low temperature and dry conditions of petrochemical pipeline inspections in winter can reach tens of degrees Celsius and tens of percentage points of relative humidity. The fixed boundary of the cross-interference feature domain cannot simultaneously adapt to both extreme environments, while dynamic boundary correction allows the cross-interference feature domain to adaptively adjust to track changes in environmental conditions. If the dynamic boundary correction in step S14 is missing, the attribution determination in step S15 will be performed based on the cross-interference feature domain of the fixed boundary. When the ambient temperature and humidity deviate from the standard conditions, the accuracy of the determination will decrease, which may lead to an increase in false triggers or an increase in the omission of real gas events.

[0082] Step S15: The first and second dimension components of the two-dimensional time-series fingerprint vector are respectively assigned to the first and second dimension correction intervals of the corrected cross-interference feature domain. Based on the assignment determination results, cross-interference events and independent gas events are distinguished. When an event is determined to be an independent gas event, a trigger request signal is generated and output to step S20.

[0083] Specifically, step S15 reads the two-dimensional time-series fingerprint vector constructed in step S13 and the corrected cross-interference feature domain generated in step S14, and performs attribution determination to distinguish between cross-interference events and independent gas events. The attribution determination method is as follows: it is determined whether the first dimension component of the two-dimensional time-series fingerprint vector falls within the first dimension correction interval of the corrected cross-interference feature domain, and simultaneously it is determined whether the second dimension component of the two-dimensional time-series fingerprint vector falls within the second dimension correction interval of the corrected cross-interference feature domain. If both the first dimension component and the second dimension component of the two-dimensional time-series fingerprint vector fall within their respective correction intervals, the system determines that the current multi-channel linkage signal change is a cross-interference event. The handling logic for cross-interference events includes: not generating trigger request signals, keeping high-power infrared sensors in sleep mode to avoid unnecessary wake-ups due to cross-interference and thus consuming battery power; recording the two-dimensional time-series fingerprint vector of this cross-interference event, the corresponding current ambient temperature and humidity values, and the signal changes of each electrochemical sentinel channel to the device's event log storage area for subsequent charging self-test phases to update the cross-interference feature domain baseline parameters, thereby achieving long-term self-learning optimization of the cross-interference feature domain; clearing the change time register of each electrochemical sentinel channel in step S12, restoring each electrochemical sentinel channel to silent monitoring state, and returning to step S11 to continue performing the dynamic calibration of the current noise baseline value for the next sentinel sampling cycle.

[0084] If at least one dimension component falls outside the corresponding correction interval, the system further distinguishes between two scenarios and executes different processing logic. Scenario 1 involves a negative second dimension component of the two-dimensional time-series fingerprint vector, meaning the event where the CO channel's signal deviation first exceeds the noise floor threshold occurs earlier than the event where the H2S channel's signal deviation first exceeds the noise floor threshold. This implies that the CO channel's response precedes the H2S channel's. In the cross-interference physical mechanism, the CO channel's cross-interference response cannot precede the H2S channel's original response because the cross-interference response is generated only after H2S gas molecules diffuse to the surface of the CO sensor's electrolyte. This diffusion process takes time, therefore the cross-interference response must lag behind the original response of the interference source channel in time. A negative second dimension component of the two-dimensional time-series fingerprint vector serves as strong evidence of an independent gas event. The system determines that the current multi-channel linkage signal change is an independent gas event, immediately generates a trigger request signal, and transmits this trigger request signal to step S20.

[0085] Scenario 2 involves a positive second-dimensional component of the two-dimensional time-series fingerprint vector, but the first-dimensional component deviates from the first-dimensional correction range. This means the temporal sequence matches the cross-interference characteristic, but the amplitude ratio deviates from it. This scenario may be caused by two reasons: first, sensor aging causes a drift in cross-sensitivity beyond the temperature and humidity correction range; in this case, the signal change still represents cross-interference, but the amplitude ratio has deviated from the factory calibration range; second, there is indeed a mixed leak of CO and H2S, with both gases existing independently but occurring close in time. In this case, the system does not immediately classify it as an independent gas event, but instead postpones the high-power infrared sensor trigger decision by one confirmation observation period. The confirmation observation period refers to the time period during which the system continuously monitors the signal change trends of the CO and H2S channels over the next several (e.g., 5 to 20) sentinel sampling periods. The confirmation and determination logic within the observation period is as follows: If the signal change of the CO channel and the signal change of the H2S channel maintain an approximately constant ratio and the signals of the two electrochemical sentinel channels rise or fall synchronously, it is considered a cross-interference event, because it is highly unlikely that two independent gas sources will remain synchronized in the time profile of concentration changes. If the change trends of the two electrochemical sentinel channel signals diverge, i.e., one electrochemical sentinel channel rises while the other falls, or the change rates of the two electrochemical sentinel channels show a significant difference, it is determined to be an independent gas event, a trigger request signal is generated, and it is transmitted to step S20. If, at the end of the confirmation and observation period, neither the conditions for confirming a cross-interference event nor the conditions for determining an independent gas event are explicitly met, the system determines the current multi-channel linkage signal change as an independent gas event according to the safety priority principle, generates a trigger request signal, and transmits it to step S20 to ensure that no real gas events are missed due to ambiguity in the determination conditions. For example, the confirmation and observation period can be set to 5 to 20 sentinel sampling periods, and the specific value is determined according to the shortest observation time required for the electrochemical sensor signal change trend to present distinguishable characteristics.

[0086] If the two-dimensional time-series fingerprint vectors transmitted in step S13 are a set, i.e., multiple two-dimensional time-series fingerprint vectors involving the O2 channel and other electrochemical sentinel channels, then the system performs the above-mentioned attribution determination on each two-dimensional time-series fingerprint vector. A cross-interference event is determined only when all two-dimensional time-series fingerprint vectors fall within the corrected cross-interference feature domain; if any two-dimensional time-series fingerprint vector falls outside the corrected cross-interference feature domain, then the determination is performed according to the logic of Case 1 or Case 2 above. The rationale for this approach is that cross-interference is a deterministic physical process. If the signal change is indeed caused by cross-interference, then the two-dimensional time-series fingerprint vectors between all electrochemical sentinel channel pairs should satisfy the cross-interference feature. The non-compliance of any electrochemical sentinel channel pair implies the existence of at least one independent gas source.

[0087] Step S15 distinguishes between cross-interference events and independent gas events by comparing the two-dimensional temporal fingerprint vector with the corrected cross-interference feature domain, using a simple rectangular region inclusion determination on the two-dimensional plane. The hierarchical determination logic for scenarios one and two achieves a balance between security and power consumption constraints: it immediately responds to strong evidence (i.e., the second dimension component of the two-dimensional temporal fingerprint vector is negative), ensuring that independent gas events are not overlooked; for weak evidence (i.e., only the first dimension component of the two-dimensional temporal fingerprint vector deviates from the first dimension correction interval), confirmation is delayed to avoid hasty determination and prevent misjudgments due to sensor aging or other reasons. The output of step S15 is a trigger request signal, which is generated only when determined to be an independent gas event and transmitted to step S20 as the start condition for waking up the high-power infrared sensor. The trigger request signal output in step S15 is the activation condition for waking up the high-power infrared sensor in step S20. The accuracy of the attribution determination in step S15 directly affects the frequency of activation in step S20. If the attribution determination is accurate, false triggers are suppressed and step S20 is only activated during independent gas events. If the attribution determination is inaccurate, false triggers are not suppressed, and step S20 is frequently activated, leading to increased battery consumption. Without the attribution determination in step S15, the system will be unable to distinguish between cross-interference events and independent gas events. All multi-channel linkage changes may lead to the generation of the trigger request signal, significantly increasing the wake-up frequency of the high-power infrared sensor. Battery life will drastically decrease due to unnecessary wake-ups caused by false triggers.

[0088] Step S10 eliminates false triggers caused by cross-interference at the signal source attribute level. The trigger request signals output after filtering in step S10 all correspond to real independent gas events. This filtering ensures that subsequent step S20 is only activated when a real gas event exists. The wake-up frequency of the high-power infrared sensor is reduced from the high-frequency state affected by cross-interference in the original scheme to a reasonable frequency corresponding only to independent gas events. This reduction in the wake-up frequency of the high-power infrared sensor directly reduces startup surge energy consumption and preheating energy consumption, allowing battery energy to be primarily used for the detection response of independent gas events rather than invalid wake-ups caused by cross-interference. The two-dimensional time-series fingerprint vector matching method in step S10 utilizes the deterministic physical mechanism of cross-interference. That is, the amplitude ratio and time difference of cross-interference must fall within a specific region determined by the sensor's physical characteristics, while independent gas events are not subject to this constraint. Therefore, the distribution area of ​​the two-dimensional time-series fingerprint vector on the two-dimensional plane exhibits distinguishable characteristics. This distinguishing ability stems from the understanding and utilization of the physical mechanism of cross-interference, rather than relying on the setting of empirical thresholds or the training of machine learning models. Therefore, it exhibits stable distinguishing performance across different batches of sensors and different application scenarios. The source filtering in step S10 reduces the activation frequency of the surge immunity window in step S20, reduces the computational load of surge immunity compensation, slows down the accumulation speed of the trigger time record in the trigger time ring buffer, and slows down the growth of the current trigger frequency in the sliding time window. This delays the timing of the working mode switching in step S24, allowing the high-power infrared sensor to maintain the on-demand trigger-and-sleep mode for a longer time, further enhancing the battery life.

[0089] Step S20: Receive a trigger request signal, activate the surge immunity window, acquire the original sampled values ​​of each electrochemical sentinel channel during the surge immunity window, perform dynamic compensation on the original sampled values ​​of each electrochemical sentinel channel during the immunity window, and after the surge immunity window is closed, count the current trigger frequency within the sliding time window. Based on the comparison between the current trigger frequency and the dynamic upper limit of the trigger frequency determined by the current battery power, perform adaptive switching on the working mode of the high-power infrared sensor.

[0090] Specifically, step S20 receives the trigger request signal output in step S15, performs a wake-up operation on the high-power infrared sensor, and dynamically compensates the sampled values ​​of each electrochemical sentinel channel through a surge immunity window during the wake-up process. Simultaneously, based on the comparison between the current trigger frequency within the sliding time window and the dynamic upper limit of the trigger frequency determined by the remaining battery power, the operating mode of the high-power infrared sensor is adaptively switched. The high-power infrared sensor is a sensor that uses the non-dispersive infrared absorption principle to detect methane concentration. During operation, it requires driving an infrared light source to emit infrared light of a specific wavelength and detecting the change in light intensity after absorption by an infrared detector. Both the luminous power of the infrared light source and the signal conditioning circuit of the infrared detector require high operating current; therefore, the power consumption of the high-power infrared sensor is much higher than that of the electrochemical sentinel sensor. The surge immunity window refers to the time period from the moment the high-power infrared sensor wake-up command is issued until the moment the electrochemical sensor baseline returns to a steady state. During this time period, the system performs dynamic compensation on the sampled values ​​of each electrochemical sentinel channel to remove the baseline offset component caused by surges. The dynamic upper limit of trigger frequency refers to the maximum number of times a high-power infrared sensor is allowed to be triggered within a sliding time window under specific battery power conditions. This upper limit decreases as the remaining battery power decreases to extend the device's usability.

[0091] Step S20 is executed after step S10 completes cross-interference identification and outputs a trigger request signal. This ensures that the wake-up process of the high-power infrared sensor does not interfere with the determination logic of step S10. When the current trigger frequency within the sliding time window exceeds the dynamic upper limit of the trigger frequency determined by the current remaining battery power, the accumulated power consumption is controlled by switching the working mode. Step S10 eliminates false triggers caused by cross-interference at the signal source attribute level. However, the surge current at the moment of startup of the high-power infrared sensor will cause a transient drop in the output voltage of the shared battery. This transient drop causes fluctuations in the bias voltage of the electrochemical sentinel sensor, which in turn causes a step shift in the baseline of each electrochemical sentinel channel. If this step shift is not processed, the determination logic of step S10 may misidentify it as a new round of independent gas events and generate a new trigger request signal, forming a chain of repeated triggers. Step S20 blocks the propagation path of the above-mentioned chain-repeated triggering through dynamic compensation during the surge immunity window. At the same time, it controls the cumulative power consumption of legal triggering through the coordinated control of dynamic upper limit of trigger frequency and adaptive switching of working mode. This makes steps S10 and S20 form a two-level protection architecture that combines source filtering and propagation blocking.

[0092] Further, step S20 includes:

[0093] Step S21: Receive the trigger request signal, read the current baseline estimate of each electrochemical sentinel channel and latch it into the corresponding baseline latch register, send a wake-up command to the power supply control switch of the high-power infrared sensor and record the time when the wake-up command is sent.

[0094] Specifically, after receiving the trigger request signal from step S15, step S21 reads the current baseline estimate stored in the baseline register of each electrochemical sentinel channel, and copies the current baseline estimate of each electrochemical sentinel channel to the baseline latch register of the corresponding electrochemical sentinel channel to complete the baseline latching operation. The baseline latch register is an internal data storage unit of the system, used to save the baseline state of each electrochemical sentinel channel before the wake-up command is issued. Its storage capacity only needs to hold one value, and it is overwritten by a new baseline latch value each time a trigger request signal arrives. The current baseline estimate is a value dynamically updated by step S11 in each sentinel sampling cycle, representing the signal reference level of each electrochemical sentinel channel under the current environmental conditions. The baseline latching operation is completed within one sentinel sampling cycle, ensuring that the baseline latch values ​​of all electrochemical sentinel channels correspond to the baseline state at the same time, avoiding time reference deviations in subsequent dynamic compensation due to asynchronous baseline latching times of each electrochemical sentinel channel.

[0095] The physical significance of the baseline latching operation is as follows: the baseline offset caused by the surge during the activation of the high-power infrared sensor only begins after the wake-up command is issued. The current baseline estimate corresponding to the last sentinel sampling cycle before the wake-up command is issued represents the true baseline level of each electrochemical sentinel channel in the unaffected state. This baseline latch value will serve as the reference benchmark for surge immune compensation in step S22. The goal of compensation is to restore the sampling values ​​of each electrochemical sentinel channel during the surge to the unaffected state represented by this baseline latch value. If baseline latching is not performed before the wake-up command is issued, surge immune compensation will lose its reference benchmark and cannot be implemented. The baseline offset cannot be accurately estimated and removed, and the determination logic in step S10 will operate on the distorted data affected by the surge, which may generate false trigger request signals.

[0096] After baseline latching of all electrochemical sentinel channels is completed, a wake-up command is sent to the power supply control switch of the high-power infrared sensor, switching the sensor from sleep mode to power-on state. The power supply control switch, a power switching device connecting the battery and the high-power infrared sensor, is typically implemented using a metal-oxide-semiconductor field-effect transistor (MOSFET). Its on / off state is controlled by the digital output port of the system controller. The wake-up command turns on the power supply control switch, the battery begins supplying power to the high-power infrared sensor, and the infrared light source and signal conditioning circuit of the high-power infrared sensor enter the power-on process. Simultaneously, the system records the precise time of the wake-up command issuance into the wake-up time register. The wake-up time register is an internal data storage unit used to store the timestamp of the wake-up command issuance, and its storage accuracy must meet the time resolution requirements of subsequent dynamic compensation. The wake-up command issuance time will serve as the time reference for the dynamic compensation item in step S22, ensuring that the time reference of the compensation item is strictly aligned with the actual surge occurrence time. After the wake-up command is issued, the system immediately proceeds to step S22 without waiting for the high-power infrared sensor to complete startup, because surge immune compensation needs to take effect immediately upon the occurrence of a surge. The startup process of a high-power infrared sensor includes two stages: infrared light source preheating and signal conditioning circuit stabilization. The entire startup process requires 5 to 20 sentinel sampling cycles to complete. For example, if the sentinel sampling cycle is set to 500 milliseconds, the startup process usually takes 2.5 to 10 seconds. If the system waits for the high-power infrared sensor to complete startup before entering step S22, the sampled values ​​of each electrochemical sentinel channel during the surge will directly participate in the sliding data buffer update in step S11 and the signal change extraction in step S12 without compensation, which may lead to the detection of false signal deviation changes exceeding the noise floor threshold for the first time.

[0097] Step S21 accurately captures the undisturbed baseline state of each electrochemical sentinel channel at the last moment before the high-power infrared sensor wake-up command is issued, providing a temporal anchor and numerical reference for the surge immune compensation in step S22. The synchronous recording of the baseline latch value and the wake-up command issuance time ensures that the time reference of the compensation item is strictly aligned with the actual surge occurrence time, enabling step S22 to accurately calculate the baseline offset estimate of each electrochemical sentinel channel at the current sampling time based on the time difference between the wake-up command issuance time and the current sampling time. Without the baseline latching operation in step S21, step S22 would be unable to obtain the baseline reference value of each electrochemical sentinel channel before the surge, the calculation of the baseline offset estimate would lose its reference, dynamic compensation would be impossible, and the baseline offset caused by the surge would be misidentified as a signal change by the judgment logic in step S10, potentially generating a false trigger request signal and triggering a chain of repeated triggers.

[0098] Step S22: Start activating the surge immunity window from the moment the wake-up command is issued. During the surge immunity window, continue to collect the original sampled values ​​of each electrochemical sentinel channel according to the sentinel sampling cycle. Calculate the baseline offset estimate of each electrochemical sentinel channel at each sampling moment based on the moment the wake-up command is issued. Subtract the corresponding baseline offset estimate from the original sampled value of each electrochemical sentinel channel during the immunity window to obtain the compensated sampled value during the immunity window. Send the compensated sampled value back to the sliding data buffer in step S11 to replace the normal sampled value, so that all sentinel monitoring logic from steps S11 to S15 continues to run based on the compensated sampled value during the surge immunity window.

[0099] Specifically, step S22 activates the surge immunity window immediately after the wake-up command in step S21 is issued. The surge immunity window starts timing from the wake-up command issuance time stored in the wake-up time register. The duration of the surge immunity window is set to the sum of the maximum duration of the high-power infrared sensor startup surge and the maximum value of the electrochemical double-layer rebalancing time of the electrochemical sensors corresponding to each electrochemical sentinel channel. The high-power infrared sensor startup surge refers to the transient large current generated at the moment of power-on of the high-power infrared sensor due to the charging demand of the capacitors in the infrared light source and signal conditioning circuit. This surge current is usually several times the steady-state operating current, and its duration is on the order of milliseconds. The maximum duration of the high-power infrared sensor startup surge is determined by actual measurement during the factory calibration stage and pre-stored in the device memory. The measurement method is as follows: the high-power infrared sensor startup operation is performed multiple times under standard environmental conditions, the power supply current waveform is recorded synchronously, the time for the surge current to decay from the peak to the steady-state current is extracted from the current waveform, and the maximum value among the multiple measurement results is taken as the maximum duration of the high-power infrared sensor startup surge.

[0100] The electrochemical double-layer rebalancing time refers to the time required for the double-layer capacitance of an electrochemical sensor to return to charge equilibrium after a step change in bias voltage. A double-layer capacitance exists at the interface between the working electrode and the electrolyte of the electrochemical sensor. This capacitance is in a stable charged state during normal sensor operation. When the bias voltage fluctuates transiently due to power surges, the double-layer capacitance needs to recharge or discharge to adapt to the new bias voltage level. This rebalancing process takes time, during which the sensor's output signal deviates from its true gas concentration response value. The true gas concentration response value refers to the steady-state output signal value of the electrochemical sensor, directly caused by the target gas concentration in the current environment, when the bias voltage is maintained at its rated operating value and the double-layer capacitance is in charge equilibrium. This value only reflects the true change in target gas concentration and does not include the non-gas response component introduced by transient bias voltage fluctuations. The electrochemical double-layer rebalancing time of the electrochemical sensor corresponding to each electrochemical sentinel channel is updated during the self-test phase after each charge is completed in the following way: During the self-test phase, the high-power infrared sensor is controlled to perform a startup surge, or briefly discharged to a preset load resistor to simulate an equivalent surge current waveform. The preset load resistor refers to a power resistor whose resistance value is preset and used to consume battery energy during the self-test phase to equivalently simulate the startup surge current waveform of the high-power infrared sensor. The method for setting its resistance value is as follows: the peak value of the steady-state startup surge current obtained by the high-power infrared sensor during the factory calibration phase is used as the target current. The quotient obtained by dividing the output voltage of the battery in the current fully charged state by the target current is used as the resistance value of the preset load resistor. The peak value of the transient discharge current generated by the load resistor when the battery is connected is consistent with the peak value of the actual startup surge current of the high-power infrared sensor. The complete process of the baseline of each electrochemical sentinel channel recovering from the offset state to the steady state is recorded synchronously. The initial offset amplitude and the time required to recover to the steady state of each electrochemical sentinel channel are extracted from the recovery process. The initial offset amplitude refers to the offset of each electrochemical sentinel channel baseline relative to the baseline before the surge. The time required to recover to a steady state refers to the time required for the baseline offset to decay to below a preset proportion of the initial offset amplitude. For example, the preset proportion can be set to a value within the range of 5% to 10%, and the specific value is determined according to the baseline stability requirements of the subsequent step S10 judgment logic.The duration of the surge immunity window is set by summing the maximum duration of the surge initiated by the high-power infrared sensor and the maximum value of the electrochemical double-layer rebalancing time in each electrochemical sensor. The physical basis is that the baseline offset of the electrochemical sensor does not disappear immediately after the surge current ends, but needs to go through the double-layer rebalancing process to gradually recover to a steady state. Therefore, the duration of the surge immunity window needs to cover both the surge current duration and the double-layer rebalancing period to ensure that the entire baseline offset process is within the coverage of the surge immunity window. Dynamic compensation can completely remove the baseline offset component caused by the surge.

[0101] During the surge immune window activation period, the system continues to perform analog-to-digital conversion acquisition on each electrochemical sentinel channel according to the normal sentinel sampling cycle, obtaining the raw sampled values ​​of each electrochemical sentinel channel during the immune window period. The raw sampled values ​​during the immune window period refer to the uncompensated sampled values ​​of each electrochemical sentinel channel acquired during the surge immune window activation period. These sampled values ​​are superimposed with the baseline offset component caused by the surge. If directly used for the sliding data buffer update in step S11 and the signal change extraction in step S12, it may lead to distortion in the calculation results of the current baseline estimate and the current noise floor value, thus affecting the detection accuracy of the event where the signal deviation change first exceeds the noise floor threshold. For each raw sampled value during the immune window period, the system calculates the time difference between the sampling time corresponding to the raw sampled value during the immune window period and the time when the wake-up command was issued, stored in the wake-up time register. This time difference represents the time interval between the current sampling time and the surge occurrence time. The system calculates the baseline offset estimate of each electrochemical sentinel channel corresponding to the current sampling time based on the baseline offset attenuation model. The baseline offset attenuation model is established based on the following: the baseline offset of the electrochemical sensor caused by power surge is an exponential decay process. Its decay law is jointly determined by the electrochemical double-layer capacitance of the sensor and the output impedance of the bias circuit. The product of the double-layer capacitance and the output impedance constitutes the time constant of this exponential decay process. In the baseline offset attenuation model, the baseline offset at any given time is equal to the initial offset amplitude of the electrochemical sentinel channel multiplied by an exponentially decaying term. The decay rate of this term is determined by the recovery time characteristics of the electrochemical sentinel channel measured during the self-test phase.

[0102] The baseline offset decay model is mathematically expressed as an exponential decay function. The baseline offset estimate at any sampling time is equal to the initial offset amplitude multiplied by the natural exponential function. The exponent of the natural exponential function is the negative value of the time difference between the sampling time and the wake-up command issuance time divided by the time constant. The initial offset amplitude is obtained from the recovery process data of each electrochemical sentinel channel recorded during the self-test phase. The time constant is calculated by back-calculating the time required to recover to steady state and the preset proportion of the recovery degree recorded during the self-test phase. For example, if the time required to recover to steady state is defined as the time required for the baseline offset to decay to 5% of the initial offset amplitude, then the time constant is equal to the time required to recover to steady state divided by the negative natural logarithm of 5%. This calculation method is based on the mathematical properties of the exponential decay function and is common knowledge in engineering calculations.

[0103] The system subtracts the corresponding baseline offset estimate from the original sampled values ​​during the immune window period of each electrochemical sentinel channel to obtain the immune window-compensated sampled values ​​for each channel. These immune window-compensated sampled values ​​represent the sentinel signal after removing the baseline offset component caused by surges, retaining only the actual gas concentration change component and normal background noise. The system sends the immune window-compensated sampled values ​​of each electrochemical sentinel channel back to the sliding data buffer in step S11, replacing the normal sampled values ​​in the dynamic calibration of the current noise baseline and the subsequent signal change extraction in step S12. Using the method of sending back the compensated sampled values ​​instead of shielding the electrochemical sentinel channel has the following advantages: While directly shielding the electrochemical sentinel channel can avoid false judgments caused by surges, it will create a safety detection blind spot during the surge and during the electrochemical double-layer rebalancing period. If a rapid gas concentration surge occurs during this period, the system will be completely unable to detect it, which may delay the critical alarm opportunity. By adopting a strategy of continuous acquisition but dynamic compensation, the baseline offset caused by the surge is removed from the sampled values ​​without interrupting sentinel monitoring. This ensures that all logical links of step S10 continue to operate normally during the surge immunity window, and the data it processes has been stripped of the impact of the surge, so it will not generate false trigger request signals due to the baseline offset caused by the surge.

[0104] Simultaneously, during the surge immunity window, the system adds an additional safety criterion: for the compensated sampled values ​​of each electrochemical sentinel channel during the immunity window, if the deviation of the compensated sampled value of any electrochemical sentinel channel from the baseline latched value of that channel exceeds the preset emergency alarm threshold, the system determines that an emergency gas event has occurred during the surge immunity window and immediately issues an independent alarm command to the detector's audible, visual, and vibration alarm module. The preset emergency alarm threshold is set to a value within the range of 3 to 5 times the preset alarm threshold. The basis for this setting is that during normal gas concentration changes, the concentration change within a single sentinel sampling cycle will not exceed the concentration increment corresponding to the preset emergency alarm threshold; only in extremely dangerous situations involving a sudden surge in gas concentration will a deviation exceeding the preset emergency alarm threshold occur. This additional safety criterion ensures that even during the special period of the surge immunity window, extremely dangerous sudden increases in gas concentration will not be missed, thus guaranteeing the system's safety constraints.

[0105] When the surge immunity window reaches its set duration, the system closes the surge immunity window, and the sampled values ​​of each electrochemical sentinel channel no longer undergo dynamic compensation. The sliding data buffer in step S11 resumes receiving normal uncompensated sampled values. The system then proceeds to step S23 to perform trigger frequency statistics. Step S22 ensures uninterrupted monitoring of each electrochemical sentinel channel during the high-power infrared sensor startup surge and the electrochemical double-layer rebalancing period, and its data has been accurately stripped of surge effects. This completely blocks the closed positive feedback loop of triggering causing voltage drops, which in turn causes baseline shift and triggers again, and eliminates the safety detection blind spot caused by simple shielding schemes. The baseline shift attenuation model parameters are updated periodically using measured data from the self-test phase after charging, which can track changes in surge characteristics caused by sensor and battery aging, ensuring that compensation accuracy remains reliable throughout the entire lifespan of the device. Step S22 sends the compensated sampled value back to the sliding data buffer of step S11, ensuring that all the decision logic in step S10 continues to run based on the compensated data during the surge immunity window, thus guaranteeing the continuity and accuracy of the cross-interference identification function during surges. Without the dynamic compensation in step S22, the decision logic in step S10 would run on distorted data from surge disturbances, reducing the accuracy of the two-dimensional temporal fingerprint vector extraction, degrading the reliability of cross-interference determination, and weakening the source filtering effect due to surge interference.

[0106] Step S23: After the surge immunity window is closed, the sliding data buffer is restored to receive the normal uncompensated sampled value, the current time is recorded in the trigger time ring buffer, the number of trigger times in the trigger time ring buffer that are within the preset sliding time window range is counted, and the number of trigger times is used as the current trigger frequency within the sliding time window.

[0107] Specifically, step S23 is executed after the surge immunity window is closed. It restores the sliding data buffer to receive normal, uncompensated sampled values, and simultaneously records the current time into the trigger time circular buffer and counts the current trigger frequency within the sliding time window. The trigger time circular buffer is a first-in, first-out data storage structure. Its length is equal to the upper limit of the maximum number of triggers allowed within the sliding time window. For example, this upper limit of the maximum number of triggers can be set to a value within the range of 10 to 50 times. The specific value is determined based on the preset length of the sliding time window and the system's desired trigger frequency control precision. The system's desired trigger frequency control precision refers to the minimum resolvable change in the number of triggers allowed by the system when counting the current trigger frequency within the sliding time window. It is set by dividing the preset length of the sliding time window by the length of the trigger time circular buffer, and the quotient is used as the time resolution corresponding to this precision. The larger the length of the trigger time circular buffer, the finer the time resolution and the higher the trigger frequency control precision; the smaller the length of the trigger time circular buffer, the coarser the time resolution and the lower the trigger frequency control precision. In specific settings, the length of the trigger time circular buffer is selected within the range of 10 to 50 trigger times, based on the balance between the detector's requirements for trigger frequency statistical sensitivity and storage resource constraints. The trigger time circular buffer stores the most recent 10 to 50 trigger times of the high-power infrared sensor. When the buffer is full and a new trigger time arrives, the system removes the oldest trigger time from the buffer and writes the new trigger time. The circular buffer data structure has the following advantages: the storage capacity of the circular buffer is fixed and will not increase indefinitely with runtime, meeting the storage resource constraints of the portable detector; the first-in, first-out (FIFO) characteristic of the circular buffer naturally eliminates expired trigger times without requiring additional periodic cleanup operations.

[0108] The sliding time window refers to the time range used to statistically analyze the trigger frequency of a high-power infrared sensor. Its starting point is the current moment minus the preset length of the sliding time window, and its ending point is the current moment. The preset length of the sliding time window is determined based on the following factors: if the length is too short, the statistical results of the trigger frequency within the sliding time window will fluctuate significantly, potentially leading to premature switching of the operating mode due to occasional triggers within a short period; if the length is too long, the statistical results of the trigger frequency within the sliding time window will respond slowly to environmental changes, potentially delaying the switching of the operating mode in intermittent leakage scenarios and causing excessive battery consumption. For example, the preset length of the sliding time window can be set to a fixed value within the range of several minutes to tens of minutes, with the specific value determined based on the detector's battery capacity and the startup power consumption of the high-power infrared sensor.

[0109] After each surge immunity window closes, the system records the current time in a trigger time circular buffer. It then scans all trigger times stored in the buffer and subtracts the preset length of the sliding time window from the current time as the window's start time. The system counts the number of trigger times in the circular buffer that are after the window's start time and uses this count as the current trigger frequency within the sliding time window. Trigger times that are before the window's start time are no longer covered by the current sliding time window and are removed from the trigger time circular buffer. By using a trigger time circular buffer to store trigger times and filtering based on the window's start time, the system avoids traversing all historical trigger events, ensuring that the time complexity of the statistical operation remains constant and does not increase with runtime, thus meeting the computational resource constraints of the portable detector.

[0110] The system transmits the current trigger frequency within the sliding time window to step S24 for operating mode determination. Step S23 maintains the trigger frequency statistics of the high-power infrared sensor in real time with extremely low computational overhead, providing a quantitative decision basis for the adaptive switching of operating modes in step S24. The closing of the surge immunity window in step S22 signifies that the surge impact handling of this trigger wake-up has been completed, and the system has returned to normal sentinel monitoring status. At this time, the execution of trigger time recording and frequency statistics ensures that the statistical trigger count is consistent with the actual effective trigger count. If trigger time recording is performed before the surge immunity window closes, invalid triggers from chained repeated triggers may also be included in the statistics in the case of surge immunity compensation failure, resulting in distorted trigger frequency statistics. The current trigger frequency within the sliding time window provided in step S23 is the input data for operating mode determination in step S24, and the two together realize the adaptive switching function of operating modes based on trigger frequency. If the trigger frequency statistics in step S23 are missing, step S24 will be unable to obtain the current trigger frequency information, and the working mode switching will lose its quantitative basis, which may lead to improper timing of working mode switching and affect battery life or detection sensitivity.

[0111] Step S24: Read the current percentage of the remaining battery power, determine the dynamic upper limit of the trigger frequency based on the current percentage of the remaining battery power, compare the current trigger frequency within the sliding time window with the dynamic upper limit of the trigger frequency, maintain the on-demand triggering and sleep mode of the high-power infrared sensor when the current trigger frequency does not exceed the dynamic upper limit of the trigger frequency, and switch the high-power infrared sensor to the continuous low duty cycle cruise mode when the current trigger frequency exceeds the dynamic upper limit of the trigger frequency.

[0112] Specifically, step S24 reads the current percentage of remaining battery power from the device's battery management unit, determines the dynamic upper limit of the trigger frequency based on this percentage, compares the current trigger frequency within the sliding time window with the dynamic upper limit, and adaptively switches the operating mode of the high-power infrared sensor based on the comparison result. The battery management unit is a circuit module inside the portable detector used to monitor and manage battery status. Its functions include battery voltage monitoring, charge / discharge current monitoring, and remaining power estimation. The current percentage of remaining battery power refers to the ratio of the current remaining battery power to the battery's full charge. This ratio ranges from 0 to 1, where 0 indicates the battery is depleted and 1 indicates the battery is fully charged. The current percentage of remaining battery power is obtained through the battery management unit's coulomb counting method or voltage lookup table method, which are well-known technologies in the field of battery management.

[0113] The method for determining the dynamic upper limit of trigger frequency is as follows: The system pre-sets two threshold values: a sufficient battery power threshold and a critical battery power threshold, with the sufficient battery power threshold being higher than the critical battery power threshold. When the current percentage of remaining battery power is greater than or equal to the sufficient battery power threshold, the dynamic upper limit of trigger frequency is set to the maximum allowed value. For example, this maximum value can be set to a value within the range of 30 to 50 times within a sliding time window. Setting this maximum value allows the system to impose almost no restriction on the trigger frequency when the battery power is sufficient, maintaining a strategy that prioritizes detection sensitivity. When the current percentage of remaining battery power is less than or equal to the critical battery power threshold, the dynamic upper limit of trigger frequency is set to the minimum allowed value. For example, this minimum value can be set to a value within the range of 3 to 10 times within a sliding time window. Setting this minimum value allows the system to impose a strict limit on the trigger frequency when the battery power is critical, maintaining a strategy that prioritizes battery life. When the current percentage of remaining battery power is greater than the critical power threshold but less than the ample power threshold, the dynamic upper limit of the trigger frequency decreases linearly as the current percentage of remaining battery power decreases. The slope of this decrease is determined by the two endpoints: the maximum value at the ample power threshold and the minimum value at the critical power threshold.

[0114] The physical basis for setting the dynamic upper limit of trigger frequency based on the remaining battery power is as follows: When the remaining battery power is sufficient, the system has the capacity to withstand a relatively high frequency of cold starts of the high-power infrared sensor. The surge consumption of each cold start accounts for a small proportion of the remaining power, so a higher trigger frequency should be allowed to maintain detection sensitivity. When the remaining battery power is insufficient, the surge consumption of each cold start accounts for a larger proportion of the remaining power, and the system should switch to a lower power operating mode earlier to extend the device's usability. The dynamic correlation between the dynamic upper limit of trigger frequency and the remaining battery power enables an adaptive strategy of limiting speed earlier when the battery power is low. For example, the sufficient power threshold can be set to a value within the range of 60% to 80%, and the critical power threshold can be set to a value within the range of 20% to 30%. The specific values ​​are determined based on the detector's battery capacity, the startup power consumption of the high-power infrared sensor, and the target battery life.

[0115] The system compares the current trigger frequency within the sliding time window transmitted in step S23 with the dynamic upper limit of the trigger frequency. See Figure 5 This is a schematic diagram comparing the two working modes of the high-power infrared sensor provided in the embodiments of this application. Figure 5 The diagram illustrates the difference in power state changes over time between the on-demand trigger-to-sleep mode and the continuous low duty cycle cruise mode. In the on-demand trigger-to-sleep mode, the sensor is in a completely off, zero-power state during the sampling interval, only achieving high-power operation during a cold start at sampling time. In the continuous low duty cycle cruise mode, the sensor maintains a very low-power, dimly lit, hot-standby state during the sampling interval, only rapidly heating up to achieve high-power operation during sampling time. If the current trigger frequency within the sliding time window does not exceed the dynamic upper limit of the trigger frequency, the system maintains the current operating mode of the high-power infrared sensor. If the high-power infrared sensor is currently in the on-demand trigger-to-sleep mode, this mode will continue to be maintained. Figure 5 The timing diagram for the on-demand trigger-to-sleep mode shows that this mode means the high-power infrared sensor completes one methane concentration sampling after each trigger and then completely shuts down. The infrared light source and signal conditioning circuit both stop receiving power, waiting for the next trigger request signal to arrive before powering on again. On-demand trigger-to-sleep mode is the lowest power consumption operating mode for the high-power infrared sensor and is suitable for quiet environments with low gas event frequency. After completing this sampling, the high-power infrared sensor shuts down, and the system returns to step S11 to wait for the next sentinel monitoring cycle.

[0116] If the current trigger frequency within the sliding time window exceeds the dynamic upper limit of the trigger frequency, the system determines that the current environment exhibits intermittent leakage or fluctuating concentration. Continuing to maintain the on-demand trigger-and-sleep mode will lead to excessive battery consumption due to the surge power consumption from repeated cold starts. The system will then switch the high-power infrared sensor from the on-demand trigger-and-sleep mode to a continuous low duty cycle cruise mode. Figure 5 The power timing of the sustained low duty cycle cruise mode refers to a high-power infrared sensor no longer being completely shut off, but instead maintaining its infrared light source in a very low-power, dimly lit state. In this dimly lit state, the light source's emission power is only a very small proportion of its normal operating brightness, insufficient for effective methane concentration measurement, but sufficient to maintain the basic temperature balance inside the high-power infrared sensor's chamber. The dimly lit state of the infrared light source means that for each subsequent sampling, the light source only needs to be increased from a dimly lit state to the operating brightness, rather than from a completely off, cold state. Figure 5 The power change process of rapid heating does not require a complete power-on process of cold start, thus reducing startup energy consumption and startup delay.

[0117] In continuous low duty cycle cruise mode, the high-power infrared sensor performs periodic methane concentration sampling at a preset low duty cycle cruise sampling interval. The low duty cycle cruise sampling interval is much longer than the sampling interval in normal operating mode to control total power consumption. The setting of the low duty cycle cruise sampling interval needs to consider the following factors: if the interval is too short, the total power consumption of periodic sampling increases, potentially offsetting the energy savings from the low-brightness hot standby state; if the interval is too long, the system's response speed to changes in methane concentration decreases, potentially delaying alarm activation. For example, the low duty cycle cruise sampling interval can be set to a fixed value within the range of tens of seconds to several minutes, with the specific value determined based on the typical rate of methane concentration change and the detector's response time requirements.

[0118] The system continues to run all sentinel monitoring logic in step S10 while maintaining continuous low duty cycle cruise mode. When a new trigger request signal is output in step S15, the system no longer executes the cold start procedure for the high-power infrared sensor. Instead, it advances the timing of the next low duty cycle cruise sampling to immediate execution, meaning that the light source is directly switched from a dim state to operating brightness for sampling in the next sentinel sampling cycle. Since the light source is already in a dim, hot-standby state, the startup energy consumption of this operation is much lower than that of a cold start from a completely off state, and the startup delay is also reduced, ensuring the system's response speed to independent gas events. In continuous low duty cycle cruise mode, whenever a trigger request signal is output and immediate sampling is performed in step S15, the system records the current time in the trigger time circular buffer in step S23, and updates the current trigger frequency within the sliding time window according to the statistical method in step S23. This ensures that trigger events during the cruise mode are included in the frequency statistics, so that the exit condition determination can accurately reflect the actual triggering status of the current environment. This avoids mode oscillation caused by the system prematurely exiting the continuous low duty cycle cruise mode due to missed trigger frequency statistics, followed by repeated cold starts and re-entry into the continuous low duty cycle cruise mode. The exit condition for continuous low duty cycle cruise mode is: the current trigger frequency within the sliding time window falls below the dynamic upper limit of the trigger frequency, and the methane concentration readings of 3 to 10 consecutive low duty cycle cruise samples are all below the preset safe concentration threshold. For example, 5 consecutive low duty cycle cruise samples can be used as the sampling number for exit determination. The specific value is determined based on the stability of methane concentration measurement and the need to prevent frequent mode switching. The safe concentration threshold refers to the upper limit of methane concentration within a safe range, which is set below the alarm concentration threshold. A logic that determines if the methane concentration readings from several consecutive (e.g., 3 to 10) low-duty-cycle cruise samples are all below the preset safe concentration threshold confirms that the current environment has returned to a stable state, preventing premature exit from the continuous low-duty-cycle cruise mode due to accidental fluctuations in a single sampling result. When the exit condition is met, the system switches the high-power infrared sensor from the continuous low-duty-cycle cruise mode back to the on-demand trigger-i.e., sleep mode. First, the light source power is reduced from a dim state to zero and completely shut off. Then, the sentinel trigger mechanism, which relies entirely on step S10, is restored, returning to normal operation. Figure 5 The power timing states shown are alternating between complete shutdown and cold start.

[0119] Step S24 establishes an intermediate working mode—a continuous low duty cycle cruise mode—between the on-demand trigger (sleep mode) and the all-time continuous working mode. It adaptively adjusts the mode switching timing by correlating the dynamic upper limit of the trigger frequency with the ratio of the remaining battery power. The all-time continuous working mode means that the infrared light source and signal conditioning circuit of the high-power infrared sensor maintain their rated operating power throughout the entire working cycle, continuously sampling methane concentration at fixed normal operating intervals without entering sleep mode or switching to a low-brightness hot standby state. This is the highest power consumption and most timely detection mode among the three working modes of the high-power infrared sensor. In intermittent leakage scenarios, the continuous low duty cycle cruise mode avoids the cumulative energy consumption of repeated cold starts, while the low-brightness hot standby state of the light source eliminates the loss of detection timeliness caused by cold start delays. When the environment returns to stability, the system automatically exits the continuous low duty cycle cruise mode to restore maximum power saving, achieving full-cycle adaptive power management. In continuous low duty cycle cruise mode, the system continues to run all sentinel monitoring logic in step S10, ensuring the continuity of the cross-interference identification function in continuous low duty cycle cruise mode. At the same time, the trigger request signal output in step S15 can trigger immediate sampling, maintaining the system's responsiveness to independent gas events. Without the adaptive switching of the operating mode in step S24, the system will continuously adopt the on-demand trigger-i.e., sleep mode in intermittent leakage scenarios. The cumulative consumption of surge energy from repeated cold starts will lead to excessive battery consumption, and the equipment's available time will be shortened.

[0120] Step S10 eliminates false triggers caused by cross-interference at the signal source attribute level. The trigger request signals output after filtering in step S10 all correspond to genuine independent gas events. Step S20 blocks the chain-like false trigger propagation path that may arise after each legitimate trigger at the trigger execution stage. Even if the startup surge of the high-power infrared sensor does cause a baseline shift in the electrochemical sensor, this baseline shift is accurately canceled out by the baseline shift estimate and will not be misidentified as a new independent gas event by step S10. The trigger frequency statistics in step S23 and the adaptive mode switching in step S24 limit the cumulative energy consumption even when all triggers are legitimate from the perspective of total power consumption. When the current trigger frequency within the sliding time window is too high, it automatically switches to a continuous low duty cycle cruise mode. Without the source filtering in step S10, step S20 needs to handle a large number of false triggers. Frequent activation of the surge immunity window itself also increases the power consumption of the control logic. The current trigger frequency within the sliding time window is forced to exceed the dynamic upper limit of the trigger frequency too early, thus pushing the system into a continuous low duty cycle cruise mode and sacrificing detection sensitivity. Without the propagation blocking step S20, the judgment logic of step S10 operates on distorted data caused by surge disturbances, resulting in decreased extraction accuracy of the two-dimensional time-series fingerprint vector, deterioration of the reliability of cross-interference judgment, and weakening of the source filtering effect due to surge interference. Without the total rate limiting step S24, while steps S10 and S22 ensure the rationality of each trigger, they cannot prevent the high-frequency cumulative consumption of reasonable triggers in intermittent leakage scenarios. The three-level protection allows each level to handle only a small number of events missed by the previous level. The system can maintain extremely high sensitivity settings at each level without worrying about uncontrolled cumulative power consumption, achieving the effect of reduced total power consumption without reduced sensitivity. This enhances the battery life of the portable four-in-one detector in complex multi-gas coexistence conditions.

[0121] Example 2

[0122] This embodiment, based on Embodiment 1, provides an asynchronous cooperative sampling control system for a multimodal gas sensor, such as... Figure 6 As shown, it includes:

[0123] Trigger request generation module: used to acquire real-time raw sampled values ​​of each electrochemical sentinel channel, extract the occurrence time of the first time the signal change and signal deviation change of each electrochemical sentinel channel exceed the noise floor threshold, construct a two-dimensional time-series fingerprint vector based on the occurrence time of the first time the signal change and signal deviation change exceed the noise floor threshold; acquire the corrected cross-interference feature domain, determine the attribution of the two-dimensional time-series fingerprint vector and the corrected cross-interference feature domain, and generate a trigger request signal based on the attribution determination result;

[0124] Infrared mode switching module: Used to receive trigger request signals, activate surge immunity window, acquire raw sampled values ​​of each electrochemical sentinel channel during the surge immunity window, perform dynamic compensation on the raw sampled values ​​of each electrochemical sentinel channel during the immunity window, and count the current trigger frequency within the sliding time window after the surge immunity window is closed. Based on the comparison between the current trigger frequency and the dynamic upper limit of the trigger frequency determined by the current battery power, adaptively switch the working mode of the high-power infrared sensor.

[0125] Furthermore, in the trigger request generation module, the method for detecting the event where the signal deviation change first exceeds the noise floor threshold includes:

[0126] The analog-to-digital conversion of multiple electrochemical sentinel channels is performed synchronously at a preset sentinel sampling period to acquire the real-time raw sample values ​​of each electrochemical sentinel channel. The real-time raw sample values ​​are written into the sliding data buffer corresponding to each electrochemical sentinel channel. Statistical calculations are performed on the sample values ​​in the sliding data buffer to obtain the current baseline estimate and current noise floor value of each electrochemical sentinel channel.

[0127] The current signal deviation change of each electrochemical sentinel channel is calculated based on the current baseline estimate. The event in which the signal deviation change of the electrochemical sentinel channel first exceeds the noise floor threshold is detected based on the current signal deviation change, and the time of occurrence of the event in which the signal deviation change first exceeds the noise floor threshold is recorded.

[0128] The method for detecting the event where the deviation of the electrochemical sentinel channel signal first exceeds the noise floor threshold is as follows:

[0129] The current signal deviation change of each electrochemical sentinel channel is compared with the corresponding noise floor threshold. When the current signal deviation change of the electrochemical sentinel channel exceeds the noise floor threshold and the electrochemical sentinel channel was previously in a silent state, it is determined that the electrochemical sentinel channel has experienced a signal deviation change exceeding the noise floor threshold for the first time.

[0130] The noise floor threshold is obtained by multiplying the current noise floor value by a preset noise multiple threshold;

[0131] The silent state refers to the situation where, in the previous several consecutive sentinel sampling cycles, the deviation of the current signal of the electrochemical sentinel channel has not exceeded the noise floor threshold of the corresponding electrochemical sentinel channel.

[0132] The method for statistically determining the number of electrochemical sentinel channels that first exceed the noise floor threshold within a preset correlation time window includes:

[0133] Within a preset associated time window, determine whether at least two electrochemical sentinel channels have experienced an event where the signal deviation change amount first exceeds the noise floor threshold. If so, read the occurrence time of the event where the signal deviation change amount first exceeds the noise floor threshold and calculate the signal change amount of each electrochemical sentinel channel. Construct a two-dimensional time-series fingerprint vector based on the signal change amount and the occurrence time of the event where the signal deviation change amount first exceeds the noise floor threshold.

[0134] If only a single electrochemical sentinel channel experiences a signal deviation change that exceeds the noise floor threshold for the first time, the decision on whether to directly generate a trigger request signal is based on the comparison between the signal change of that electrochemical sentinel channel and the preset alarm threshold.

[0135] The method for constructing the two-dimensional temporal fingerprint vector includes:

[0136] The ratio of signal changes between channels is calculated based on the signal changes of each electrochemical sentinel channel, and this ratio is used as the first dimension component of the two-dimensional time-series fingerprint vector. The time difference between channels is calculated based on the time when the signal deviation change of each electrochemical sentinel channel first exceeds the noise floor threshold, and this time difference is used as the second dimension component of the two-dimensional time-series fingerprint vector. The first dimension component and the second dimension component of the two-dimensional time-series fingerprint vector are combined to form the two-dimensional time-series fingerprint vector.

[0137] The methods and systems of this application may be implemented in many ways. For example, they may be implemented by software, hardware, firmware, or any combination of software, hardware, and firmware. The above-described order of steps for the method is for illustrative purposes only, and the steps of the method of this application are not limited to the order specifically described above, unless otherwise specifically stated.

[0138] In addition, the parts of the technical solutions provided in the embodiments of this application that are consistent with the implementation principles of the corresponding technical solutions in the prior art have not been described in detail, so as to avoid excessive elaboration.

[0139] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the invention. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. An asynchronous cooperative sampling control method for a multimodal gas sensor, characterized in that, The method includes: The system acquires real-time raw sampled values ​​from each electrochemical sentinel channel, extracts the time of occurrence of the first occurrence of the signal change and signal deviation change exceeding the noise floor threshold for each channel, calculates the ratio of signal changes between channels based on the signal changes of each channel, and constructs a two-dimensional time-series fingerprint vector based on the time of occurrence of the first ... Upon receiving a trigger request signal, the surge immunity window is activated. During the surge immunity window, the original sampled values ​​of each electrochemical sentinel channel are acquired. Dynamic compensation is performed on the original sampled values ​​of each electrochemical sentinel channel during the immunity window. After the surge immunity window is closed, the current trigger frequency within the sliding time window is counted. Based on the comparison between the current trigger frequency and the dynamic upper limit of the trigger frequency determined by the current remaining battery power, the working mode of the high-power infrared sensor is adaptively switched. The method for detecting the event that the signal deviation change of the electrochemical sentinel channel exceeds the noise floor threshold for the first time is as follows: the current signal deviation change of each electrochemical sentinel channel is compared with the corresponding noise floor threshold. When the current signal deviation change of the electrochemical sentinel channel exceeds the noise floor threshold and the electrochemical sentinel channel was previously in a silent state, it is determined that the electrochemical sentinel channel has experienced the event that the signal deviation change exceeds the noise floor threshold for the first time. The noise floor threshold is obtained by multiplying the current noise floor value by a preset noise multiplier threshold; the silent state refers to the fact that the change in the current signal of the electrochemical sentinel channel has not exceeded the noise floor threshold of the corresponding electrochemical sentinel channel in the previous several consecutive sentinel sampling cycles.

2. The asynchronous cooperative sampling control method for a multimodal gas sensor according to claim 1, characterized in that, The method for detecting the event where the signal deviation change first exceeds the noise floor threshold includes: The analog-to-digital conversion of multiple electrochemical sentinel channels is performed synchronously at a preset sentinel sampling period to acquire the real-time raw sample values ​​of each electrochemical sentinel channel. The real-time raw sample values ​​are written into the sliding data buffer corresponding to each electrochemical sentinel channel. Statistical calculations are performed on the sample values ​​in the sliding data buffer to obtain the current baseline estimate and current noise floor value of each electrochemical sentinel channel. The current signal deviation change of each electrochemical sentinel channel is calculated based on the current baseline estimate. The event in which the signal deviation change of the electrochemical sentinel channel first exceeds the noise floor threshold is detected based on the current signal deviation change, and the time of occurrence of the event in which the signal deviation change first exceeds the noise floor threshold is recorded.

3. The asynchronous cooperative sampling control method for a multimodal gas sensor according to claim 2, characterized in that, Within a preset correlation time window, the number of electrochemical sentinel channels that experience an event where the signal deviation change first exceeds the noise floor threshold is statistically determined, and the signal change of each electrochemical sentinel channel is extracted.

4. The asynchronous cooperative sampling control method for a multimodal gas sensor according to claim 3, characterized in that, The method for statistically determining the number of electrochemical sentinel channels that first exceed the noise floor threshold within a preset correlation time window includes: Within a preset associated time window, determine whether at least two electrochemical sentinel channels have experienced an event where the signal deviation change amount first exceeds the noise floor threshold. If so, read the occurrence time of the event where the signal deviation change amount first exceeds the noise floor threshold and calculate the signal change amount of each electrochemical sentinel channel. Construct a two-dimensional time-series fingerprint vector based on the signal change amount and the occurrence time of the event where the signal deviation change amount first exceeds the noise floor threshold. If only a single electrochemical sentinel channel experiences a signal deviation change that exceeds the noise floor threshold for the first time, the decision on whether to directly generate a trigger request signal is based on the comparison between the signal change of that electrochemical sentinel channel and the preset alarm threshold.

5. The asynchronous cooperative sampling control method for a multimodal gas sensor according to claim 4, characterized in that, The method for constructing the two-dimensional temporal fingerprint vector includes: The ratio of signal changes between channels is used as the first dimension component of the two-dimensional time-series fingerprint vector; the time difference between channels is calculated based on the time of occurrence of the event when the signal deviation change of each electrochemical sentinel channel first exceeds the noise floor threshold, and is used as the second dimension component of the two-dimensional time-series fingerprint vector; the first dimension component and the second dimension component of the two-dimensional time-series fingerprint vector are combined to form the two-dimensional time-series fingerprint vector.

6. The asynchronous cooperative sampling control method for a multimodal gas sensor according to claim 1, characterized in that, The method for obtaining the corrected cross-interference feature domain includes: Load the cross-interference feature domain reference parameters, which include a first dimension reference interval and a second dimension reference interval; Read the current ambient temperature and humidity values, and perform dynamic boundary correction on the first and second dimension reference intervals in the cross-interference feature domain reference parameters based on the current ambient temperature and humidity values, respectively, to obtain the corrected cross-interference feature domain composed of the first and second dimension correction intervals.

7. The asynchronous cooperative sampling control method for a multimodal gas sensor according to claim 1, characterized in that, The method for determining the attribution of the two-dimensional temporal fingerprint vector and the corrected cross-interference feature domain includes: Determine whether the first dimension component of the two-dimensional temporal fingerprint vector falls within the first dimension correction interval of the corrected cross-interference feature domain, and simultaneously determine whether the second dimension component of the two-dimensional temporal fingerprint vector falls within the second dimension correction interval of the corrected cross-interference feature domain. Based on the attribution determination result, cross-interference events and independent gas events are distinguished, and a trigger request signal is generated when the event is determined to be an independent gas event.

8. The asynchronous cooperative sampling control method for a multimodal gas sensor according to claim 7, characterized in that, The activation time of the surge immune window is: A wake-up command is sent to the power supply control switch of the high-power infrared sensor and the time of the wake-up command is recorded. The surge immunity window is activated from the time the wake-up command is sent.

9. The asynchronous cooperative sampling control method for a multimodal gas sensor according to claim 8, characterized in that, The method for dynamically compensating the original sampled values ​​during the immune window of each electrochemical sentinel channel includes: Based on the wake-up command issuance time, the baseline offset estimate of each electrochemical sentinel channel at each sampling time is calculated. The original sampled value during the immune window period of each electrochemical sentinel channel is subtracted from the corresponding baseline offset estimate to obtain the sampled value after compensation during the immune window period.

10. The asynchronous cooperative sampling control method for a multimodal gas sensor according to claim 9, characterized in that, The method for adaptively switching the operating mode of the high-power infrared sensor includes: When the current trigger frequency does not exceed the dynamic upper limit of the trigger frequency, the high-power infrared sensor maintains the on-demand trigger-and-sleep mode. When the current trigger frequency exceeds the dynamic upper limit of the trigger frequency, the high-power infrared sensor is switched to the continuous low duty cycle cruise mode.

11. An asynchronous cooperative sampling control system for a multimodal gas sensor, used to implement the asynchronous cooperative sampling control method for a multimodal gas sensor according to any one of claims 1-10, characterized in that, The system includes: Trigger request generation module: used to acquire real-time raw sampled values ​​of each electrochemical sentinel channel, extract the occurrence time of the first time the signal change and signal deviation change of each electrochemical sentinel channel exceed the noise floor threshold, construct a two-dimensional time-series fingerprint vector based on the occurrence time of the first time the signal change and signal deviation change exceed the noise floor threshold; acquire the corrected cross-interference feature domain, determine the attribution of the two-dimensional time-series fingerprint vector and the corrected cross-interference feature domain, and generate a trigger request signal based on the attribution determination result; Infrared mode switching module: Used to receive trigger request signals, activate surge immunity window, acquire raw sampled values ​​of each electrochemical sentinel channel during the surge immunity window, perform dynamic compensation on the raw sampled values ​​of each electrochemical sentinel channel during the immunity window, and count the current trigger frequency within the sliding time window after the surge immunity window is closed. Based on the comparison between the current trigger frequency and the dynamic upper limit of the trigger frequency determined by the current battery power, adaptively switch the working mode of the high-power infrared sensor.

Images