Gas detection wireless alarm method and system based on machine learning

Abstract

The invention relates to a gas detection wireless alarm method and system based on machine learning. The method comprises the following steps: collecting gas data in an environment; when the hazardous gas concentration exceeds the standard or the oxygen content is insufficient, an uplink alarm signal is generated and synchronously sent to the data processing end and the mobile monitoring end; a machine learning trend prediction model is constructed, the machine learning trend prediction model is adopted to analyze the gas data, and the predicted risk level and risk probability are output; the data processing end generates a corresponding downlink alarm signal and synchronously sends the downlink alarm signal to the detection end and the mobile monitoring end; and triggering an audible and visual alarm of the mobile monitoring end to give an alarm, and calculating the position information of the detection end in real time. According to the invention, information of a dangerous environment where an operator is located can be fed back to a guardian in real time, so that the guardian can quickly start rescue or intervention measures, and accident risks are reduced. Accurate position information is synchronized to the mobile monitoring terminal, and rescue workers are guided to quickly position risk points and trapped persons, so that the rescue response time is shortened.

Description

Technical Field

[0001] This invention relates to the field of gas detection technology, and in particular to a wireless alarm method and system for gas detection based on machine learning. Background Technology

[0002] Alarm gas detectors are widely used in industrial production and municipal security. The performance of an alarm gas detector is determined by its core hardware and intelligent software; the components work together to achieve a complete "detection-analysis-alarm" function. However, current alarm gas detectors still have certain limitations. First, the detector relies on the physical or chemical reaction between the gas and the sensor to monitor concentration. It requires waiting for the gas to be generated, diffuse into the detection range, and reach a concentration threshold before triggering an alarm. It cannot combine environmental parameters (temperature, humidity, pressure) and equipment operating status to predict the conditions and trends of gas generation. For example, in chemical reactions, it cannot provide early warning of toxic gases that may be generated due to reaction imbalance; it can only respond passively after the gas is generated. Second, alarm gas detectors lack precise location tracking for workers, leading to difficulties in quickly locating personnel in emergencies and delaying rescue opportunities. Third, there is a disconnect in information transmission. Existing portable gas detectors can only issue an audible alarm via buzzer when detecting excessive levels of hazardous gases or insufficient oxygen. Monitoring personnel cannot obtain real-time environmental risk information about workers in a timely manner, easily missing rescue opportunities. Field tests showed that when the distance between the monitoring personnel and the detector exceeded 5 meters, the alarm sound was difficult to detect due to environmental noise and complex spatial structures, failing to meet the "alarm alert within 30 meters" requirement proposed by the health and safety department. In existing technologies, simply increasing the alarm decibel level has limited effectiveness under complex working conditions. Summary of the Invention

[0003] The purpose of this invention is to provide a gas detection wireless alarm method and system based on machine learning, thereby solving the aforementioned problems in the prior art.

[0004] The technical solution of the present invention to solve the above-mentioned technical problems is as follows: A machine learning-based wireless gas detection alarm method and system includes the following steps: Step S01: Collect gas data in the environment in real time through the detection terminal. The gas data includes the concentration of hazardous gases and the oxygen content. Step S02: When the detection terminal detects that the concentration of hazardous gas exceeds the standard or the oxygen content is insufficient, an uplink alarm signal is generated, triggering the audible and visual alarm of the detection terminal, and the gas data and the uplink alarm signal are simultaneously sent to the data processing terminal and the mobile monitoring terminal. Step S03: Deploy a machine learning trend prediction model based on a CNN model at the data processing end. When the data processing end receives an uplink alarm signal, it uses the machine learning trend prediction model to analyze the gas data and output the predicted risk level and risk probability. Step S04: Based on the risk level, the data processing terminal generates a corresponding downlink alarm signal and sends it synchronously to the detection terminal and the mobile monitoring terminal; Step S05: After receiving the downlink alarm signal, the mobile monitoring terminal triggers the audible and visual alarm and calculates the location information of the detection terminal in real time.

[0005] The beneficial effects of this invention are as follows: As a key node connecting operators and supervisors, the mobile monitoring terminal can provide real-time feedback on the hazardous environment where the operators are located to the supervisors, facilitating their rapid initiation of rescue or intervention measures and reducing the risk of accidents. When the detection terminal detects that the gas concentration is about to exceed the standard, it immediately triggers an audible and visual alarm, and simultaneously synchronizes the precise location information to the mobile monitoring terminal held by the supervisor, guiding rescuers to quickly locate the risk point and the trapped personnel. Compared with traditional devices that only alarm but do not provide location information, this reduces the rescue response time by more than 80%.

[0006] Based on the above technical solution, the present invention can be further improved as follows.

[0007] Furthermore, step S03 specifically includes the following steps: Step S31: Compensate the gas data sent by the detection end using a multivariate linear compensation model; Step S32: Construct a machine learning trend prediction model based on a CNN model; Step S321: Input Feature Construction Let the current time be The model input is a fixed length Data within the time window: ; In the formula, They are respectively Concentration, temperature, and humidity at any given time; It is ( A matrix of (+1)×3 is used as the input to the CNN model; Step S322: One-dimensional convolution operation The convolutional layers of a CNN model pass through multiple filters. In the input Slide up to extract local temporal features; Each filter is located at The output is: ; In the formula, The size of the convolution kernel; This is the convolution operator; Step S323: Risk Level Probability Output The final layer of the CNN model uses the Softmax function to transform the output into a probability distribution for three risk levels: ; In the formula, To give input Under the conditions, the risk level is The predicted probability, =0, 1, 2 correspond to three risk levels: normal state, potential risk, and emergency alarm. These are the raw scores output by the fully connected layers of the model; Step S34: According to The function determines the final risk level and sets the warning confidence level based on the probability value.

[0008] The further beneficial effects of adopting the above are: based on the comprehensive analysis of the two types of data, the system achieves an upgrade from the traditional "passive detection alarm" to "active prediction and early warning", which can predict the trend of gas concentration changes in advance and detect potential risks earlier.

[0009] Furthermore, the multivariate linear compensation in step S31 adopts the following formula: ; In the formula, The accurate gas concentration value after compensation; The raw gas concentration value is read directly from the sensor; These are the real-time measured values ​​of temperature, humidity, and air pressure, respectively. These are the reference temperature, humidity, and air pressure values ​​for calibrating the sensor in a standard laboratory environment. These are the compensation coefficients for temperature, humidity, and air pressure, respectively.

[0010] The further beneficial effects of the above are: the present invention uses multivariate linear compensation, which combines the compensation coefficients of temperature, humidity and air pressure, to eliminate the influence of environmental cross-sensitivity on gas sensor readings and improve data quality.

[0011] Furthermore, the calculation of the location information of the detection end in step S05 specifically includes the following steps: Step S51: UWB-based positioning The detection terminal sends a UWB pulse signal to the mobile monitoring terminal. Assuming the coordinates of the detection terminal are... Base station The coordinates are Base station 1 serves as the reference base station; the measurement of UWB pulse signals reaching the base station is used for reference. Time difference with base station 1 The distance difference can be obtained: ; In the formula, Speed ​​of light m / s; For tag to base station The distance difference to base station 1; construct at least two different base stations. Time difference with base station 1 Solve the system of equations to obtain the precise coordinates of the detection end. .

[0012] The further beneficial effects of adopting the above method are: lower power consumption than the traditional TOF method, making it more suitable for battery-powered portable devices. Attached Figure Description

[0013] Figure 1 This is a flowchart of the method of the present invention. Detailed Implementation

[0014] The principles and features of the present invention are described below with reference to the accompanying drawings. The examples given are only for explaining the present invention and are not intended to limit the scope of the present invention.

[0015] Example 1 like Figure 1 As shown, a machine learning-based wireless gas detection alarm method and system includes the following steps: Step S01: Collect gas data in the environment in real time through the detection terminal. The gas data includes the concentration of hazardous gases and the oxygen content. Step S02: When the detection terminal detects that the concentration of hazardous gas exceeds the standard or the oxygen content is insufficient, an uplink alarm signal is generated, triggering the audible and visual alarm on the detection terminal. The gas data and the uplink alarm signal are simultaneously sent to the data processing terminal and the mobile monitoring terminal. Before being sent, the raw gas concentration data is pre-processed by an embedded microprocessor (such as an ARM Cortex-M series). The built-in calibration algorithm, combined with real-time temperature and humidity data, compensates and corrects the gas readings to reduce environmental interference and improve data accuracy. This provides a higher quality data foundation for subsequent machine learning predictions. Step S03: Deploy a machine learning trend prediction model based on a CNN (convolutional neural network) model at the data processing end. When the data processing end receives an uplink alarm signal, it uses the machine learning trend prediction model to analyze the gas data and output the predicted risk level and risk probability; thus realizing the upgrade from passive alarm to active prediction. Step S04: Based on the risk level, the data processing terminal generates a corresponding downlink alarm signal and sends it synchronously to the detection terminal and the mobile monitoring terminal; Step S05: After receiving the downlink alarm signal, the mobile monitoring terminal triggers the audible and visual alarm and calculates the location information of the detection terminal in real time.

[0016] This invention employs a three-terminal architecture consisting of a detection terminal, a data processing terminal, and a mobile monitoring terminal. The detection terminal, worn by the operator, is used to collect real-time data on hazardous gas concentrations and oxygen content, and transmit alarm signals to the data processing terminal. The data processing terminal embeds a convolutional neural network model to intelligently model and predict gas concentration trends, achieving an upgrade from passive alarm to proactive early warning. The mobile monitoring terminal, carried by the monitoring personnel, is used to receive risk levels and locate the operator's position, significantly improving rescue response speed and operational safety.

[0017] As a crucial link between workers and supervisors, the mobile monitoring terminal can provide real-time feedback on the hazardous environment in which workers are located to supervisors, facilitating rapid initiation of rescue or intervention measures and reducing accident risks. When the detection terminal detects that the gas concentration is about to exceed the standard, it immediately triggers an audible and visual alarm, while simultaneously synchronizing precise location information to the mobile monitoring terminal held by the supervisor. This guides rescue personnel to quickly locate the risk point and trapped personnel, shortening the rescue response time by more than 80% compared to traditional devices that only alarm but do not provide location information.

[0018] Example 2 like Figure 1 As shown, this embodiment is a further improvement on embodiment 1, as detailed below: Step S03 specifically includes the following steps: Step S31: Use a multivariate linear compensation model to compensate the gas data sent by the detection end, eliminate the influence of environmental cross-sensitivity on the gas sensor readings, and improve data quality; Step S32: Construct a machine learning trend prediction model based on a CNN (Convolutional Neural Network) model; Step S321: Input Feature Construction (Temporal Sliding Window) Let the current time be The model input is a fixed length Data within the time window: ; In the formula, They are respectively Concentration, temperature, and humidity at any given time; It is ( A matrix of (+1)×3 is used as the input to the CNN model; The model input includes not only the gas concentration at the current moment. ,temperature ,humidity It also includes historical data from the past time window (e.g., the past 60 seconds, one point per second), forming a multidimensional time series. For example: .

[0019] Step S322: One-dimensional convolution operation (feature extraction) The convolutional layers of a CNN model use multiple filters (convolutional kernels). In the input Slide up to extract local temporal features; Each filter is located at The output is: ; In the formula, The size of the convolution kernel (time dimension); This is the convolution operator; Calculate and input the first derivative (rate of change), second derivative (acceleration of change), and moving average of the gas concentration within the time window. These features are key to identifying abnormal trends such as "sudden increases in concentration."

[0020] This operation automatically learns and captures key local time patterns such as "slow rise" and "violent fluctuation".

[0021] Step S323: Risk level probability output (Softmax function) The final layer of the CNN model uses the Softmax function to transform the output into a probability distribution for three risk levels: ; In the formula, To give input Under the conditions, the risk level is The predicted probability, =0, 1, 2 correspond to three risk levels: normal state, potential risk, and emergency alarm. These are the raw scores (logits) output by the fully connected layers of the model. Step S34: According to The function determines the final risk level and sets the warning confidence level based on the probability value.

[0022] The architecture of CNN in this invention is described as follows: Input layer: Receives the multidimensional time series data mentioned above.

[0023] One-dimensional convolutional layer (1D-CNN): Multiple convolutional kernels of different sizes (such as 3x1, 5x1) are used to slide in the time dimension to automatically extract local time patterns (such as small increases, smooth fluctuations, and violent jitters).

[0024] Pooling layer (MaxPooling1D): Downsamples the convolutional features, retaining the most salient features, while enhancing the model's robustness to small temporal fluctuations.

[0025] Fully connected layer: flattens the extracted features and performs non-linear combination.

[0026] Output layer: The Softmax activation function is used to output the probabilities of three nodes, which correspond to the three levels of "normal state", "potential risk" (early warning) and "emergency alarm".

[0027] Training and deployment process: Offline training: The CNN model is trained on a server using a large amount of historical normal operating conditions and leakage accident data from industrial sites to learn normal baseline patterns and abnormal evolution patterns.

[0028] Model lightweighting: Pruning and quantizing the trained model to convert it into TensorFlow Lite or ONNX format, which can run efficiently on embedded hardware (such as microprocessors at the data processing end).

[0029] Online inference and self-optimization: The deployed model performs real-time inference. The system periodically (e.g., every 24 hours) encrypts and uploads new field data to the cloud for incremental training or model fine-tuning, generating optimized model parameters before distributing them to the terminal to achieve continuous learning and performance evolution.

[0030] Based on the comprehensive analysis of the two types of data, the system upgrades from the traditional "passive detection and alarm" to "active prediction and early warning", which can predict the trend of gas concentration changes in advance and detect potential risks earlier.

[0031] In practice, the machine learning trend prediction model also has self-optimization capabilities. It can continuously learn from new leakage cases and environmental change data on site, and continuously correct the trend identification parameters, so that the early warning accuracy gradually improves with the use time, solving the problem of false alarms and missed alarms in traditional detection under complex working conditions.

[0032] Example 3 like Figure 1 As shown, this embodiment is a further improvement on embodiment 2, as detailed below: The multivariate linear compensation in step S31 uses the following formula: ; In the formula, The accurate gas concentration value after compensation; The raw gas concentration value is read directly from the sensor; These are the real-time measured values ​​of temperature, humidity, and air pressure, respectively. These are the reference temperature, humidity, and air pressure values ​​for calibrating the sensor in a standard laboratory environment. The compensation coefficients for temperature, humidity, and air pressure are determined through multi-dimensional calibration experiments before the sensor leaves the factory.

[0033] This invention employs multivariate linear compensation, combining compensation coefficients for temperature, humidity, and air pressure, to eliminate the influence of environmental cross-sensitivity on gas sensor readings and improve data quality.

[0034] Example 4 like Figure 1 As shown, this embodiment is a further improvement on embodiment 1, as detailed below: Step S04 specifically includes: Potential Risk (Level 1 / Blue Alert): Triggered when the machine learning trend prediction model identifies that the rate of change in gas concentration continuously exceeds twice the normal baseline standard deviation, but the absolute value is still below the safety threshold; at this time, the detection end only receives a text prompt and does not trigger the audible and visual alarm on the detection end to avoid unnecessary panic; Imminent Exceeding Limit (Level 2 / Yellow Alert): Triggered when the machine learning trend prediction model predicts that the concentration has a very high probability (>90%) of exceeding 50% of the safety threshold within the next 10-15 seconds; the sound and light alarm at the detection end will sound, the LED light will flash slowly, and the speaker will emit a gentle reminder sound; Emergency Response (Level 3 / Red Alert): Triggered when the measured concentration of harmful gas reaches the threshold, or when the machine learning trend prediction model predicts an immediate exceedance; both the audible and visual alarms at the detection end and the mobile monitoring end will sound, and the precise location information will be continuously broadcast. In practice, the broadcast is made using the speaker built into the mobile monitoring end.

[0035] By modeling gas concentration trends using machine learning, abnormal conditions such as sudden increases in concentration can be identified in advance. When a flammable gas leak occurs, the system can detect subtle trends in concentration deviation from the normal baseline within 0.8 seconds, quickly triggering a three-level warning: "potential risk - impending exceedance - emergency response." This response speed is 5 times faster than traditional detectors (which rely on concentration reaching a threshold for alarm, with an average response time of 4-5 seconds), providing on-site personnel with more time to respond and significantly reducing the risk of leaks causing explosions and fires.

[0036] Example 5 like Figure 1As shown, this embodiment is a further improvement on embodiment 1, as detailed below: The calculation of the location information of the detection end in step S05 specifically includes the following steps: Step S51: UWB-based positioning The detection terminal sends a UWB pulse signal to the mobile monitoring terminal. Assuming the coordinates of the detection terminal are... Base station The coordinates are Base station 1 serves as the reference base station; the measurement of UWB pulse signals reaching the base station is used for reference. Time difference with base station 1 The distance difference can be obtained: ; In the formula, For the speed of light ( m / s); For tag to base station The distance difference to base station 1; construct at least two different base stations. Time difference with base station 1 Solve the system of equations to obtain the precise coordinates of the detection end. .

[0037] The precise coordinates of the label can be calculated by solving a system of equations consisting of at least two such hyperbolic equations. This is the core of achieving a positioning accuracy of 10-30 centimeters; The algorithm employs Time Difference of Arrival (TDOA). The operator's detection device (tag) sends a UWB pulse signal, which is received by at least three fixed base stations at known locations. By calculating the minute time difference between the arrival times of the signal at different base stations, the tag's precise two-dimensional coordinates (10-30 cm accuracy) are determined. Power consumption is lower than the traditional Time of Flight (TOF) method, making it more suitable for battery-powered portable devices.

[0038] Step S05, which involves calculating the location information of the detection end, also includes the following steps: Step S52: BLE-based positioning Multiple BLE beacons were deployed on-site, and the distance was estimated by measuring the signal strength (RSSI) using a logarithmic distance path loss model, as shown in the following formula: ; In the formula, In distance The signal strength measured at the location; To be at the reference distance The known signal strength at a location (typically 1 meter); It is a path loss index, which is closely related to the environment (e.g., about 2.0 in open spaces, and may reach 3.0-4.0 in complex factory environments). Let be a random variable that follows a normal distribution, representing the shadowing fading effect; Measure the RSSI values ​​of at least three beacons and use this model to estimate the distances, then estimate the approximate area of ​​the detection end using trilateration. Multiple BLE beacons are deployed on-site, and the detection device scans and reads the RSSI (Received Signal Strength Indication) values ​​of nearby beacons. Using trilateration or the simplest "nearest beacon" method, area positioning within 3-5 meters is achieved.

[0039] Step S53: When a detection node A is in danger and located in a UWB blind zone, it broadcasts an alarm message via BLE. Upon receiving the message, other nearby detection nodes B, if they are within the coverage area of ​​UWB or a stronger signal, will act as relay nodes and anonymously forward the alarm message and approximate location of detection node A to the mobile monitoring terminal, ensuring the reachability of the alarm message.

[0040] Example 6 like Figure 1 As shown, a gas detection wireless alarm system includes a detection terminal, a data processing terminal, and a mobile monitoring terminal connected by wireless communication; it uses the gas detection wireless alarm method of any one of the embodiments 1 to 5 for detection, alarm, and location.

[0041] This invention adopts a three-terminal architecture consisting of a detection terminal, a data processing terminal, and a mobile monitoring terminal, enabling monitoring personnel to obtain real-time environmental data and risk levels of workers, so as to quickly locate or rescue workers in emergency situations.

[0042] The detection unit includes a gas detection module, a wireless transmission module, and an alarm module; The gas detection module is used to collect gas data in the environment in real time, including the concentration of hazardous gases and oxygen content, and generate alarm signals based on the gas detection results; it uses semiconductor gas sensors to detect the concentration of hydrogen sulfide (H2S), carbon monoxide (CO), and methane (CH4); it uses an oxygen electrode sensor to detect the oxygen concentration in the environment; it also includes temperature and humidity sensors and air pressure sensors as environmental compensation units. The alarm module is used to activate the local audible and visual alarm at the detection end based on the alarm signal; it integrates a buzzer and an LED warning light. The wireless transmission module is used to send detected gas data and alarm signals to the data processing terminal and the mobile monitoring terminal. The wireless signal transmission uses a 315MHz wireless communication module, including a 315MHz controller and modem, ensuring a transmission distance of no less than 30 meters in complex environments, and enabling point-to-point communication between the detection terminal, data processing terminal, and mobile monitoring terminal. Communication Protocol and Data Packet Structure: Data packets should include: a frame header, a unique device ID, a gas concentration data array, environmental parameters (temperature, humidity), device status (battery level, self-test status), a timestamp, and a CRC checksum. This structured data packet ensures the integrity and resolvability of data transmission.

[0043] "Transmit-Acknowledge-Retransmit" Mechanism: To ensure reliable delivery of alarm signals, the wireless transmitting module should adopt a "transmit-acknowledge-retransmit" mechanism. After the detection end sends an alarm signal, it will wait for an acknowledgment signal (ACK) from the receiving end. If no ACK is received within a set time (e.g., 100ms), the detection end will automatically retransmit the alarm data packet (up to 3 times), greatly reducing missed detections caused by momentary signal interference.

[0044] The mobile monitoring terminal includes a visualization display device; the visualization display device is used to display an electronic map and the specific location of the operator on the electronic map, and to display the optimal path planning from the mobile monitoring terminal to the specific location.

[0045] In practice, when an alarm is triggered, the corresponding person's icon will flash and display their name, ID, type of hazardous gas, real-time concentration, and risk level; it will also provide optimal route planning to guide monitoring personnel to avoid obstacles and reach the incident location as quickly as possible.

[0046] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A machine learning-based wireless gas detection alarm method, characterized in that, Includes the following steps: Step S01: Collect gas data in the environment in real time through the detection terminal. The gas data includes the concentration of hazardous gases and the oxygen content. Step S02: When the detection terminal detects that the concentration of hazardous gas exceeds the standard or the oxygen content is insufficient, an uplink alarm signal is generated, triggering the audible and visual alarm of the detection terminal, and the gas data and the uplink alarm signal are simultaneously sent to the data processing terminal and the mobile monitoring terminal. Step S03: Deploy a machine learning trend prediction model based on a CNN model at the data processing terminal. When the data processing terminal receives the uplink alarm signal, it uses the machine learning trend prediction model to analyze the gas data and outputs the predicted risk level and risk probability. Step S04: The data processing terminal generates a corresponding downlink alarm signal based on the risk level and sends it synchronously to the detection terminal and the mobile monitoring terminal; Step S05: After receiving the downlink alarm signal, the mobile monitoring terminal triggers the audible and visual alarm and calculates the location information of the detection terminal in real time.

2. The gas detection wireless alarm method according to claim 1, characterized in that, Step S03 specifically includes the following steps: Step S31: Compensate the gas data sent by the detection end using a multivariate linear compensation model; Step S32: Construct a machine learning trend prediction model based on a CNN model; Step S321: Input Feature Construction Let the current time be The model input is a fixed length Data within the time window: ； In the formula, They are respectively Concentration, temperature, and humidity at any given time; It is ( A matrix of (+1)×3 is used as the input to the CNN model; Step S322: One-dimensional convolution operation The convolutional layers of a CNN model pass through multiple filters. In the input Slide up to extract local temporal features; Each filter is located at The output is: ； In the formula, The size of the convolution kernel; This is the convolution operator; Step S323: Risk Level Probability Output The final layer of the CNN model uses the Softmax function to transform the output into a probability distribution for three risk levels: ； In the formula, To give input Under the conditions, the risk level is The predicted probability, =0, 1, 2 correspond to three risk levels: normal state, potential risk, and emergency alarm. These are the raw scores output by the fully connected layers of the model; Step S34: According to The function determines the final risk level and sets the warning confidence level based on the probability value.

3. The gas detection wireless alarm method according to claim 2, characterized in that, The multivariate linear compensation in step S31 adopts the following formula: ； In the formula, The accurate gas concentration value after compensation; The raw gas concentration value is read directly from the sensor; These are the real-time measured values ​​of temperature, humidity, and air pressure, respectively. These are the reference temperature, humidity, and air pressure values ​​for calibrating the sensor in a standard laboratory environment. These are the compensation coefficients for temperature, humidity, and air pressure, respectively.

4. The gas detection wireless alarm method according to claim 1, characterized in that, Step S04 specifically includes: Potential risk: When the machine learning trend prediction model identifies that the rate of change in gas concentration continuously exceeds twice the normal baseline standard deviation, but the absolute value is still below the safety threshold, it will trigger; at this time, the detection end will only receive a text prompt and will not trigger the audible and visual alarm of the detection end. Imminent Exceeding Limit: This is triggered when the machine learning trend prediction model predicts that the concentration has a very high probability of exceeding 50% of the safety threshold within the next 10-15 seconds; the sound and light alarm at the detection end will sound, the LED light will flash slowly, and the speaker will emit a gentle reminder sound. Emergency Response: When the measured concentration of harmful gas reaches the threshold, or when the machine learning trend prediction model predicts that the limit will be exceeded immediately, the alarm will be triggered; both the audible and visual alarms at the detection end and the audible and visual alarms at the mobile monitoring end will sound, and the precise location information will be continuously broadcast.

5. The gas detection wireless alarm method according to claim 1, characterized in that, The calculation of the location information of the detection end in step S05 specifically includes the following steps: Step S51: UWB-based positioning The detection terminal sends a UWB pulse signal to the mobile monitoring terminal. Assuming the coordinates of the detection terminal are... Base station The coordinates are Base station 1 is the reference base station; the arrival time of the UWB pulse signal at the base station is measured. Time difference with base station 1 The distance difference can be obtained: ； In the formula, Speed ​​of light m / s; For tag to base station The distance difference to base station 1; construct at least two different base stations. Time difference with base station 1 The equations are used to solve the system of equations, and the precise coordinates of the detection end are obtained. .

6. The gas detection wireless alarm method according to claim 5, characterized in that, The calculation of the location information of the detection end in step S05 further includes the following steps: Step S52: BLE-based positioning Multiple BLE beacons were deployed on-site, and the distance was estimated by measuring the signal strength using a logarithmic distance path loss model, as shown in the following formula: ； In the formula, In distance The signal strength measured at the location; To be at the reference distance The known signal strength at the location; It is a path loss index, which is closely related to the environment; Let be a random variable that follows a normal distribution, representing the shadowing fading effect; Measure the RSSI values ​​of at least three beacons and estimate the approximate area of ​​the detection end using the trilateration method; Step S53: When a detection node A is in danger and located in a UWB blind zone, it broadcasts an alarm message via BLE. Upon receiving the message, other nearby detection nodes B, if they are within the coverage area of ​​UWB or a stronger signal, will act as relay nodes and anonymously forward the alarm message and approximate location of detection node A to the mobile monitoring terminal.

7. A wireless alarm system for gas detection, characterized in that, It includes a detection terminal, a data processing terminal, and a mobile monitoring terminal with wireless communication connection; and uses the gas detection wireless alarm method as described in any one of claims 1 to 6 for detection, alarm, and location.

8. The gas detection wireless alarm system according to claim 7, characterized in that, The detection terminal includes a gas detection module, a wireless transmission module, and an alarm module; The gas detection module is used to collect gas data in the environment in real time, including hazardous gas concentration and oxygen content data, and generate alarm signals based on the gas detection results. The alarm module is used to activate the local audible and visual alarm at the detection end based on the alarm signal; The wireless transmission module is used to send the detected gas data and alarm signals to the data processing terminal and the mobile monitoring terminal.

9. The gas detection wireless alarm system according to claim 7, characterized in that, The mobile monitoring terminal includes a visualization display device; the visualization display device is used to display an electronic map and the specific location of the operator on the electronic map, and to display the optimal path plan from the mobile monitoring terminal to the specific location.

Images