Handheld rapid laughing gas detector

Abstract

The invention relates to the technical field of detectors, in particular to a handheld rapid laughing gas detector which comprises a hardware assembly and an algorithm processing system, the hardware assembly comprises a control circuit board, and the algorithm processing system operates in the control circuit board; the hardware assembly further comprises a pretreatment module, an internal filter, a sensor array, an air cavity and an air pump, and under the control of the control circuit board, the air pump drives external air to be sucked in through the pretreatment module; when the concentration of particulate matters is high, filtering pores of the front filtering cotton sleeve are reduced, and the service life of an internal filter is prolonged; when the particulate matter concentration of outside air is low, the size of a filtering hole is increased, the air inlet resistance is reduced, the air pump load is reduced, and the detection endurance time is prolonged.

Description

Technical Field

[0001] This invention relates to the field of detection instrument technology, specifically a handheld nitrous oxide rapid detection instrument. Background Technology

[0002] Nitrous oxide, also known as nitrous oxide, is of great significance in many fields. For example, in the medical field, it is necessary to accurately detect and control the concentration of nitrous oxide used in surgery to ensure the effectiveness of anesthesia and patient safety. In the field of environmental monitoring, it is necessary to detect the content of nitrous oxide in the atmosphere to assess its impact on environmental issues such as the greenhouse effect. In the field of law enforcement, it is necessary to conduct rapid testing on suspected nitrous oxide users to combat the illegal abuse of nitrous oxide.

[0003] However, existing nitrous oxide detection technologies and equipment have many shortcomings: Regarding laboratory testing methods, classic methods such as high-performance liquid chromatography (HPLC) and gas chromatography, while accurate and precise, involve cumbersome procedures requiring specialized technicians to strictly follow complex steps for sample preparation and injection. From sample collection to obtaining the final result, it often takes several days, failing to meet the needs of rapid on-site testing. Existing portable detection devices typically perform two filtrations on the gas in the target environment: a pretreatment module and an internal filter. The pretreatment module handles coarse filtration, removing dust particles and large-diameter droplets, while the internal filter further dehumidifies and filters the gas. The filtration efficiency of the pretreatment module directly affects the lifespan of the internal filter. Increasing the filtration density and reducing the pore size of the pretreatment module can significantly extend the lifespan of the internal filter at higher particulate concentrations, but this increases airflow resistance and pump power consumption. Portable detection devices are usually battery-powered, which significantly reduces the device's runtime given a fixed battery capacity. Summary of the Invention

[0004] The purpose of this invention is to provide a handheld nitrous oxide rapid detection device to solve the problem of the conflict between the lifespan of the internal filter and the power supply of portable detection devices mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides the following technical solution: a handheld nitrous oxide rapid detection device, comprising hardware components and an algorithm processing system. The hardware components include a control circuit board, and the algorithm processing system runs on the control circuit board. The hardware components also include a pre-processing module, an internal filter, a sensor array, a gas chamber, and an air pump. Under the control of the control circuit board, the air pump drives external gas to be drawn in through the pre-processing module, processed by the internal filter, and then enters the sensor array and the gas chamber for detection. The pre-processing module includes a module shell and a pre-filter cotton sleeve disposed inside the module shell. External gas enters the pre-processing module and undergoes coarse filtration through the pre-filter cotton sleeve. An annular air gap is provided inside the pre-filter cotton sleeve, and a axial pressure plate is provided at the end of the pre-filter cotton sleeve. The axial pressure plate compresses the pre-filter cotton sleeve to different degrees through axial movement. Under the first stage of compression, the annular air gap closes and disappears; under the second stage of compression, the porosity of the pre-filter cotton sleeve decreases.

[0006] The annular air gaps are provided in several groups and are evenly distributed along the axial direction of the front filter cotton sleeve. The cross-section of the annular air gaps is elliptical and disappears when the front filter cotton sleeve is subjected to axial compression.

[0007] The external part of the shaft pressure plate is fixedly provided with a sliding protrusion, and the inner wall of the module shell is provided with a wall sliding groove. The sliding protrusion slides along the wall sliding groove, so that the shaft pressure plate can only move axially.

[0008] A lead screw sleeve is fixedly installed on one side of the shaft pressure plate, and a lead screw shaft is screwed into the lead screw sleeve. When the lead screw shaft rotates, it can drive the shaft pressure plate to move axially. A rotor frame is fixedly installed outside the lead screw shaft, and a rotor part is fixedly installed outside the rotor frame. A stator part is fixedly installed on the inner wall of the module shell, and the stator part can drive the rotor part to rotate.

[0009] An inner partition ring is fixedly installed on the inner wall of the module shell. A counterweight ring is installed between the inner partition ring and the end wall of the module shell. An annular outer bladder is installed on both sides and the outer surface of the counterweight ring. A cross-shaped air passage is opened inside the counterweight ring, and the annular outer bladders are interconnected through the cross-shaped air passage. A flexible connecting tube is installed outside the counterweight ring, and one end of the flexible connecting tube is connected to the cross-shaped air passage.

[0010] A control panel is provided on one side of the pressure plate. The front filter cotton sleeve is disposed between the control panel and the inner partition ring. A closed awning is fixedly provided on the outside of the pressure plate. The control panel slides and seals against the closed awning. A tension spring is provided on one side of the control panel. The tension spring applies tension to the control panel, causing the control panel to tend to move away from the pressure plate.

[0011] A central tube is fixedly installed at the center of the axial pressure plate. A locking groove is opened in the sub-control status plate. A locking spring is installed in the central tube. The end of the locking spring is inserted into the locking groove, so that the positions of the axial pressure plate and the sub-control status plate are locked relative to each other. A piston body is installed inside the central tube. A pushing member is fixedly installed on the piston body. The end of the central tube is connected to the flexible connecting tube. When the annular outer bladder is squeezed, the piston body is driven by the medium to move, so that the pushing member squeezes the locking spring, and the end of the locking spring is pulled out from the locking groove.

[0012] The central tube is equipped with a state locking spring, and the state locking spring is fixedly provided with a triangular protrusion and an expansion portion; the end of the lead screw shaft is fixedly provided with a tapered portion.

[0013] When the filter cotton sleeve is in a compressed state, the triangular protrusion is stuck at the end of the pusher, so that the pusher will not compress the locking spring. When the filter cotton sleeve is in a non-compressed state, the tapered part pushes the expansion part to expand and move. At this time, the locking spring is elastically bent, and the triangular protrusion moves away from the end of the pusher.

[0014] The algorithm processing system includes, in sequence, a signal preprocessing module, a multi-dimensional feature extraction module, a deep model analysis module, and a result output and feedback module; taking the raw signals collected by the sensor array and the gas chamber as input, the algorithm processing system processes them to achieve nitrous oxide detection.

[0015] Compared with the prior art, the beneficial effects of the present invention are: 1. The handheld nitrous oxide rapid detector of the present invention enables portable and rapid on-site detection. Through the pre-treatment module, it can automatically adjust the pore size of the front filter cotton sleeve according to the particulate matter concentration of the outside air. When the particulate matter concentration is high, the pore size of the front filter cotton sleeve is reduced to further improve the coarse filtration effect of the gas and extend the life of the internal filter. When the particulate matter concentration of the outside air is low, the pore size is automatically increased to reduce the air intake resistance, reduce the air pump load, and extend the detection endurance time.

[0016] 2. By using the annular air gap in conjunction with the shaft pressure plate, the first and second stages can be controlled separately, thereby significantly improving the adjustment range of the front filter cotton sleeve.

[0017] 3. The present invention, through the combination of a set of sub-control status disk, central tube and status locking spring, automatically closes the annular air gap when the pretreatment module is impacted, only when the current filter cotton sleeve is in an uncompressed state, thus prioritizing the life of the internal filter; avoiding the risk of particulate matter concentration increasing and failing to protect the internal filter in time after the axial adjustment failure of the shaft pressure plate caused by the impact force. Attached Figure Description

[0018] Figure 1 This is a schematic diagram of the overall structure of the present invention.

[0019] Figure 2 This is a schematic diagram of the preprocessing module of this invention.

[0020] Figure 3 This is another schematic diagram of the preprocessing module.

[0021] Figure 4 This is a three-dimensional half-sectional schematic diagram of the preprocessing module of the present invention.

[0022] Figure 5 for Figure 4 Enlarged schematic diagram of region A in the middle.

[0023] Figure 6 This is a three-dimensional half-section front view of the preprocessing module of the present invention.

[0024] Figure 7 This is a schematic diagram of the components at the control status panel of the present invention.

[0025] Figure 8 This is a schematic diagram of the parts at the lead screw shaft of the present invention.

[0026] In the diagram: 1. Pre-treatment module; 2. Internal filter; 3. Sensor array and air chamber; 4. Air pump; 5. Control circuit board; 6. Module housing; 7. Pre-filter sleeve; 8. Annular air gap; 9. Shaft pressure plate; 901. Sliding protrusion; 902. Wall groove; 903. Lead screw sleeve; 904. Lead screw shaft; 905. Rotor frame; 906. Rotor section; 907. Stator section; 601. Inner diaphragm ring; 602. Counterweight ring; 603. Annular outer bladder; 604. Cross-shaped air passage; 60 5. Flexible connecting tube; 606. Individual control status panel; 607. Enclosed awning; 608. Tension spring; 609. Center tube; 610. Locking groove; 611. Locking spring; 612. Piston body; 613. Pushing component; 614. Status locking spring; 615. Triangular protrusion; 616. Expansion part; 617. Conical part; 501. Display screen; 502. Button; 503. Top cover of outer casing; 504. Bottom cover of outer casing; 505. Battery; 618. Connector; 619. Air intake protection mesh. Detailed Implementation

[0027] 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.

[0028] Please seeFigures 1 to 8 This invention provides a technical solution: a handheld nitrous oxide rapid detection device, comprising hardware components and an algorithm processing system. The hardware components include a control circuit board 5, and the algorithm processing system runs on the control circuit board 5. Figure 1 As shown, a display screen 501 and a button 502 are installed on the control circuit board 5. The display screen 501 is used to display information, and the button 502 is used for input control.

[0029] The hardware components also include a pre-processing module 1, an internal filter 2, a sensor array, an air chamber 3, and an air pump 4. Under the control of the control circuit board 5, the air pump 4 drives external gas to be drawn in through the pre-processing module 1. After being processed by the internal filter 2, the gas enters the sensor array and air chamber 3 for detection. The pre-processing module 1 acts as a coarse filter, removing large-diameter water droplets and dust particles from the gas. The internal filter 2, installed inside the detector, further absorbs water and filters the gas. The internal filter 2 is made of a highly absorbent resin filter element and is installed inside the detector. The internal filter 2 has a higher replacement cost. By using the pre-processing module 1 for coarse filtration, the service life of the internal filter 2 can be effectively extended, reducing operating costs. Through filtration and dehumidification by the pre-processing module 1 and the internal filter 2, the sample humidity can be stabilized within the optimal range for sensor adaptation, reducing the algorithm compensation pressure on the subsequent algorithm processing system.

[0030] The handheld nitrous oxide rapid detector of the present invention also includes an upper cover 503 and a lower cover 504. The upper cover 503 and the lower cover 504 are mated together to form the outer shell of the detector. The control circuit board 5, the internal filter 2, the sensor array, the gas chamber 3 and the gas pump 4 are all protected inside the outer shell.

[0031] The sensor array and gas cavity 3 include a nanomechanical film-type surface stress gas sensor, which consists of a circular silicon film suspended by four cantilever beams. A piezoresistor is embedded in the cantilever beam as the core sensing element. The stress change is generated by the interaction between gas molecules and the receptor material on the silicon film surface. The stress change is converted into a detectable electrical signal through a circuit. Specifically, nitrous oxide molecules combine with the specific receptor material on the silicon film. The combination process causes the silicon film to produce a small stress deformation, which drives the resistance change of the piezoresistor in the cantilever beam. The piezoresistor forms a Huygens bridge, which converts the resistance change into a voltage signal, which serves as the raw data for subsequent algorithm analysis.

[0032] like Figure 1 As shown, a battery 505 is also fixedly installed inside the housing of the detector. The battery 505 powers the control circuit board 5, air pump 4, sensor array, and air chamber 3. The air pump 4 is the single device with the highest power in the detector, which includes a small motor and cylinder structure. Therefore, the load change of the air pump 4 usually has the greatest impact on the detector's battery life.

[0033] The pretreatment module 1 includes a module housing 6 and a pre-filter cotton sleeve 7 disposed inside the module housing 6. The pre-filter cotton sleeve 7 has a short tubular structure and is made of materials such as sponge. It can change the size of the filter pores by compression and is elastic. External gas enters the pretreatment module 1 and undergoes coarse filtration through the pre-filter cotton sleeve 7. Figure 3 and Figure 4 As shown, a connector 618 is connected to the end of the module housing 6, and an air intake protection net 619 is embedded in the outer wall of the module housing 6. External gas enters the interior of the module housing 6 through the air intake protection net 619, is filtered by the front filter cotton sleeve 7, and is discharged through the connector 618. The connector 618 is connected to the internal filter 2 through a gas pipe, and the other end of the internal filter 2 is connected to the sensor array and the air chamber 3 through a gas pipe.

[0034] The front filter cotton sleeve 7 has an annular air gap 8 inside, and a shaft pressure plate 9 is provided at the end of the front filter cotton sleeve 7. The shaft pressure plate 9 squeezes the front filter cotton sleeve 7 to different degrees by axial movement. Under the first stage of compression, the annular air gap 8 closes and disappears. Under the second stage of compression, the porosity of the front filter cotton sleeve 7 decreases.

[0035] Several sets of annular air gaps 8 are provided and are evenly distributed along the axial direction of the front filter cotton sleeve 7. The cross-section of the annular air gaps 8 is elliptical and disappears when the front filter cotton sleeve 7 is subjected to axial compression.

[0036] The external part of the axial pressure plate 9 is fixedly provided with a sliding protrusion 901, and the inner wall of the module shell 6 is provided with a wall sliding groove 902. The sliding protrusion 901 slides along the wall sliding groove 902, so that the axial pressure plate 9 can only move axially.

[0037] A lead screw sleeve 903 is fixedly installed on one side of the shaft pressure plate 9. A lead screw shaft 904 is screwed in the lead screw sleeve 903. When the lead screw shaft 904 rotates, it can drive the shaft pressure plate 9 to move axially. A rotor frame 905 is fixedly mounted on the outside of the lead screw shaft 904, and a rotor part 906 is fixedly mounted on the outside of the rotor frame 905. A stator part 907 is fixedly mounted on the inner wall of the module housing 6, and the stator part 907 can drive the rotor part 906 to rotate.

[0038] An inner partition ring 601 is fixedly installed on the inner wall of the module shell 6. A counterweight ring 602 is installed between the inner partition ring 601 and the end wall of the module shell 6. Annular outer bladders 603 are respectively installed on the two side surfaces and the outer surface of the counterweight ring 602. A cross-shaped air passage 604 is opened inside the counterweight ring 602, allowing the annular outer bladders 603 to communicate with each other. Figure 4As shown, the annular outer bladder 603 is disposed on both sides and the outer side of the counterweight ring 602, so that at least one set of annular outer bladders 603 can be squeezed when the counterweight ring 602 is subjected to radial or axial impact force. In its natural state, the annular outer bladders 603 are in a fully inflated state. When one set of annular outer bladders 603 is squeezed, the other annular outer bladders 603 will not inflate further to contain gas.

[0039] The counterweight ring 602 is provided with a flexible connecting tube 605 on its outside, and one end of the flexible connecting tube 605 is connected to the cross-shaped air passage 604.

[0040] A control panel 606 is provided on one side of the pressure plate 9. The front filter cotton sleeve 7 is located between the control panel 606 and the inner partition ring 601. A closed awning 607 is fixedly provided on the outside of the pressure plate 9. The control panel 606 and the closed awning 607 are in sliding sealing contact. A tension spring 608 is provided on one side of the control panel 606. The tension spring 608 applies tension to the control panel 606, so that the control panel 606 has a tendency to move away from the pressure plate 9.

[0041] A central tube 609 is fixedly installed at the center of the shaft pressure plate 9. A locking groove 610 is opened in the sub-control status plate 606. A locking spring 611 is installed in the central tube 609. The end of the locking spring 611 is inserted into the locking groove 610, so that the positions of the shaft pressure plate 9 and the sub-control status plate 606 are locked relative to each other. A piston body 612 is provided inside the central tube 609, and a pusher 613 is fixedly provided on the piston body 612. The end of the central tube 609 is connected to the flexible connecting tube 605. When the annular outer bladder 603 is squeezed, the piston body 612 is driven to move by the medium, so that the pusher 613 squeezes the locking spring 611, and the end of the locking spring 611 is pulled out from the locking groove 610.

[0042] The center tube 609 is provided with a status locking spring 614, and a triangular protrusion 615 and an expansion portion 616 are fixedly provided on the status locking spring 614; a tapered portion 617 is fixedly provided at the end of the lead screw shaft 904.

[0043] When the filter cotton sleeve 7 is in a compressed state, the triangular protrusion 615 is stuck at the end of the pusher 613, so that the pusher 613 will not compress the locking spring 611; when the filter cotton sleeve 7 is in a non-compressed state, the expansion part 616 is pushed to expand by the tapered part 617, at which time the locking spring 614 is elastically bent, and the triangular protrusion 615 is removed from the end of the pusher 613.

[0044] The algorithm processing system includes, in sequence, a signal preprocessing module, a multidimensional feature extraction module, a deep model analysis module, and a result output and feedback module.

[0045] Signal preprocessing module: The core objective of preprocessing is to purify the raw signal, eliminating irrelevant factors such as noise, baseline drift, and environmental interference, thus providing high-quality data for subsequent feature extraction. This mainly includes the following steps: Signal Acquisition and Synchronization: The piezoresistors corresponding to the four cantilever beams of the MSS sensor array in air chamber 3 output analog signals through a Huygens bridge. These analog signals are then converted into digital signals by a 16-bit ADC converter, with the sampling frequency set to 100Hz to ensure the capture of dynamic details of the sensor response. Simultaneously, a synchronization triggering mechanism synchronizes the operating status of air pump 4, sensor sampling, and environmental parameter (temperature, humidity) acquisition, providing timing-matched data for subsequent environmental compensation.

[0046] Noise Removal: A combined filtering algorithm of "adaptive median filtering + wavelet threshold filtering" is employed. First, adaptive median filtering targets impulse noise in the signal (such as spike signals caused by electromagnetic interference) by dynamically adjusting the filter window size, preserving the signal's abrupt change characteristics (such as the response peak at the initial stage of gas adsorption) while eliminating isolated noise points. Subsequently, wavelet threshold filtering selects the db4 wavelet basis and performs a three-level decomposition of the signal. An improved soft threshold function is used for high-frequency noise components (avoiding signal smoothing distortion caused by traditional soft threshold filtering) to reconstruct a low-noise sensor response signal.

[0047] Baseline calibration: A dynamic baseline model is established based on clean air signals collected during the equipment zeroing phase. A sliding window averaging method (with a window size of 50 sampling points) is used to update the baseline value in real time, eliminating baseline shifts caused by sensor drift and temperature changes. For any given sensor signal, the calibrated effective response signal is obtained by subtracting the real-time baseline value from the current signal value, ensuring signal comparability under different environments.

[0048] Environmental compensation: Introducing temperature (-10℃~60℃) and humidity A compensation model (10% RH ~ 90% RH) was established. Standard nitrous oxide concentration response data were collected experimentally under different temperature and humidity conditions, and a multiple linear regression compensation equation was constructed. in, The compensated signal value, The signal value after baseline calibration. , Here, represents the standard environmental parameters, and a, b, and c are compensation coefficients calibrated experimentally. This equation can effectively offset the effects of temperature and humidity changes on the sensor response, improving signal stability.

[0049] Multidimensional feature extraction module: Based on the preprocessed signal, key features strongly correlated with nitrous oxide concentration and gas type are extracted from both static and dynamic dimensions to form a high-dimensional feature vector, as follows: Static feature extraction: Peak / valley value and difference calculation: Traverse the sensor response curve (detection period of 0~10 seconds) to identify the maximum response value (peak value). ()) and minimum response value (valley value) ), calculate the difference between the peak value and the baseline value ( )), The value reflects the sensor's response intensity to nitrous oxide and is positively correlated with the concentration.

[0050] Integration / Response Area Calculation: The trapezoidal integration method is used to calculate the area between the response curve and the baseline. (t is the detection period, taken as 10 seconds), quantifying the overall response energy of the sensor to compensate for the defect that the single peak characteristic is susceptible to instantaneous interference.

[0051] Stable value and variance calculation: The average signal value after 3 seconds of the detection period is taken as the stable value. This reflects the response state of the sensor after it reaches adsorption equilibrium; the variance of the signal is calculated over the entire detection period. (N is the number of sampling points) to evaluate the noise level of the signal. The smaller the variance, the higher the reliability of the signal.

[0052] Dynamic feature extraction: Slope / gradient calculation: The instantaneous slope of the response curve is calculated using the first-order difference method. , With a sampling interval of 0.01 seconds, the maximum rising slope is extracted. (Gas adsorption stage) and maximum descent slope (Gas desorption stage) reflects the adsorption / desorption rate of nitrous oxide molecules on the sensor receptor material, and is used to distinguish the kinetic differences between nitrous oxide and other gases.

[0053] Feature normalization: The extracted 10-dimensional features (peak, valley, difference, integral area, stable value, variance, maximum rise slope, maximum fall slope, peak curvature, response time) are normalized using Z-score to eliminate the influence of differences in the dimensions of different features. The normalization formula is as follows: in, The mean of the features, The standard deviation of the features is used to ensure that all features have equal weight in subsequent model training. Deep model analysis module: A fusion architecture combining a one-dimensional convolutional neural network (1D-CNN), a single-channel long short-term memory network (LSTM), and a channel attention mechanism is employed to perform deep processing on high-dimensional feature vectors, enabling accurate identification and concentration quantification of nitrous oxide. 1D-CNN Feature Extraction Layer: This layer serves as the basic feature extractor, taking into account a normalized temporal feature sequence (1000 sampling points, corresponding to a 10-second detection period). The network contains three convolutional layers with kernel sizes of 3, 5, and 7, each with a stride of 1. Each convolutional layer is followed by a batch normalization (BatchNorm) layer and a ReLU activation function to suppress gradient vanishing and improve feature representation. The specific structure is as follows: Convolutional layer 1: 1 input channel, 32 output channels, kernel size 3, output feature map size (998, 32). Convolutional layer 2: 32 input channels, 64 output channels, kernel size 5, output feature map size (994, 64). Convolutional layer 3: 64 input channels, 128 output channels, kernel size 7, output feature map size (988, 128). Each convolutional layer captures local feature patterns (such as the peak shape of the response curve and abrupt slope changes) through a sliding window, learns the local "fingerprint" features corresponding to nitrous oxide, and has strong robustness to noise.

[0054] Single-channel LSTM temporal modeling layer: The 128-channel feature map output from the 1D-CNN is input into 128 independent LSTM units (one channel corresponds to one LSTM). Each LSTM unit contains 64 hidden layer nodes, employing an "input gate-forget gate-output gate" gating mechanism to model the long-term dependencies of the sensor response. For example, to address the issue that nitrous oxide and interfering gases (such as carbon dioxide) have similar response curve shapes but different adsorption / desorption sequences, the LSTM effectively distinguishes the dynamic differences between the two by memorizing long-term temporal information. The LSTM layer output is a 128-dimensional temporal feature vector containing complete sensor response evolution information.

[0055] Channel Attention Mechanism Layer: Based on the Squeeze-and-Excitation (SE) module, weights are assigned to the 128-dimensional feature vector output by the LSTM. First, global average pooling is used to compress the features of each channel into a single value, reflecting the importance of that channel. Then, a two-layer fully connected network (32 hidden layer nodes, ReLU activation function; 128 output layer nodes, Sigmoid activation function) adaptively learns the weight coefficients of each channel. Finally, the weight coefficients are multiplied by the original feature vector to enhance key feature channels and suppress interfering channels. For example, when other gas interference exists in the detection environment, the network automatically reduces the weights of interference-sensitive channels and increases attention to nitrous oxide-specific channels, significantly improving the model's recognition accuracy.

[0056] Concentration Quantization and Classification Output Layer: The feature vector processed by the attention mechanism is input into a two-layer fully connected network. The first layer (64 hidden nodes, ReLU activation function) performs feature fusion. The second layer outputs two types of results: one is the probability of nitrous oxide presence using a sigmoid activation function (used to distinguish between nitrous oxide and non-nitrous oxide), and the other is the concentration value of nitrous oxide using a linear activation function (based on a standard concentration calibration curve, achieving quantization within the range of 0~5000ppm). Simultaneously, a confidence threshold (0.8) is set. When the probability of nitrous oxide presence is higher than the threshold, the concentration value is output; otherwise, it is determined that there is no nitrous oxide, avoiding false positives.

[0057] Results output and feedback module: Real-time result output: The nitrous oxide concentration value output by the model is compared with the preset threshold (slight exceedance: 100ppm, severe exceedance: 500ppm). The concentration value and the corresponding status color (green: ≤100ppm, yellow: 100~500ppm, red: >500ppm) are displayed on the screen, and an audible and visual alarm is triggered at the same time (when the limit is exceeded).

[0058] Data storage and backtracking: The pre-processed raw signal, extracted feature vectors, model output results, detection time, environmental parameters and other data are stored locally on the device and transmitted to the cloud server via Bluetooth / Wi-Fi.

[0059] Online model updates: The cloud server regularly iterates and trains the model based on historical data uploaded from multiple devices (using incremental learning algorithms to avoid the high cost of retraining), and pushes the optimized model parameters to handheld devices via wireless communication to continuously improve the accuracy of the algorithm.

[0060] The algorithm processing system in this invention effectively resists interference from temperature, humidity, and cross-gas through multi-layer protection of algorithm combination filtering, environmental compensation, and channel attention mechanism. It can still maintain high-precision detection in complex environments. The entire algorithm process (from signal acquisition to result output) takes ≤3 seconds, which is far superior to traditional laboratory methods (several hours to several days) and meets the needs of rapid on-site detection.

[0061] The algorithm of this invention is highly robust, integrates static and dynamic features, and combines the advantages of 1D-CNN and LSTM. It can capture local signal patterns and model long-term temporal relationships. It can stably identify nitrous oxide of different concentrations and under different environments. Furthermore, the algorithm architecture supports the expansion of sensor arrays (such as increasing the number of sensors) and automatically adapts to new sensor channels through the channel attention mechanism without requiring significant modifications to the model structure.

[0062] In the algorithm processing system of this invention: Static characteristics include peak / valley values, difference, integral / response area, variance, and stable values.

[0063] Peak / valley value and difference: Identify the maximum response intensity of the gas sensor and determine baseline drift.

[0064] Integral / Response Area: Quantifies the overall response energy of the sensor and is strongly correlated with the gas concentration.

[0065] Stability and variance: These assess the stability and noise level of the signal.

[0066] Dynamic characteristics: mainly slope / gradient and curvature.

[0067] Slope / gradient: Analyze the speed of sensor response and recovery to reveal the kinetics of gas adsorption / desorption.

[0068] 1D-CNN (One-dimensional Convolutional Neural Network): As a basic feature extractor, it slides on a single sensor signal, effectively capturing local response patterns (such as shape changes in a specific response phase), is robust to noise, and can learn local "fingerprint" features related to gas species.

[0069] Single-channel LSTM (Long Short-Term Memory Network): It processes the entire response time of each sensor separately, models its long-term dependencies and dynamic evolution process, and can understand the complete "response story" of the sensor. It is crucial for distinguishing gases with similar response curves but different temporal dynamics.

[0070] Channel attention mechanism: This mechanism automatically learns and assigns importance weights to different gas sensor channels, enabling adaptive feature fusion. The network dynamically focuses on the sensors most critical to the current identification task, suppressing redundant or interfering information and significantly improving model accuracy and robustness.

[0071] This invention achieves in-depth analysis of gas sensor data through a combination of multi-dimensional feature extraction and a dedicated deep learning architecture. It can accurately identify the type of target gas, the concentration of nitrous oxide, and the gas dynamics process, and has important application value in the fields of gas detection and environmental monitoring.

[0072] In terms of hardware, the preprocessing module 1 in this invention is installed on the outer casing of the detector during use, such as... Figure 1 As shown, when air pump 4 is working, it generates negative pressure to drive gas flow. The external gas flow path is as follows: first, it enters the pre-processing module 1, passes through the internal filter 2, then enters the sensor array and air chamber 3 for data acquisition and detection, and finally is drawn into air pump 4 and discharged through air pump 4. Figure 4 As shown, in the pretreatment module 1, external gas is drawn in through the air intake protection net 619, coarsely filtered by the pre-filter cotton sleeve 7 to remove large-diameter water droplets and particulate matter, and then discharged through the plug nozzle 618.

[0073] By energizing the stator section 907, the stator section 907 can generate magnetic force to drive the rotor section 906 to rotate. The rotor section 906 then drives the rotor frame 905 and the lead screw shaft 904 to rotate. Figure 4 and Figure 5 As shown, when the lead screw shaft 904 rotates, the lead screw sleeve 903 can be driven to move axially relative to the lead screw shaft 904 for adjustment. The shaft pressure plate 9 is fixed integrally with the lead screw sleeve 903, so the shaft pressure plate 9 can move axially for adjustment.

[0074] This invention also incorporates a particle size sensor on the casing of the handheld nitrous oxide rapid detector. When the particle size sensor detects a low concentration of external gas particles, such as... Figure 4 As shown, the axial pressure plate 9 is positioned at the far right. At this time, the front filter cotton sleeve 7 is in a non-compressed state and is not subjected to any axial compression. The annular air gap 8 remains. When external gas enters the inside of the front filter cotton sleeve 7 from the outside, because the front filter cotton sleeve 7 is in a non-compressed state, the pore structure is larger, and the annular air gap 8 exists, the filtration resistance is greatly reduced when the gas passes through the annular air gap 8. This results in extremely low air intake resistance for the pretreatment module 1, thus reducing the load on the air pump 4. This makes the device more energy-efficient and extends the battery life under battery power conditions.

[0075] As the concentration of particulate matter in the external gas increases, the axial pressure plate 9 of this invention squeezes the front filter cotton sleeve 7 to the left to varying degrees. In the first stage, the axial pressure plate 9, by moving to the left and squeezing, causes the annular air gap 8 to close and disappear. At this time, the pore size of the front filter cotton sleeve 7 does not change significantly. The closure and disappearance of the annular air gap 8 increases the filtration path when the gas passes through the front filter cotton sleeve 7, improving the filtration effect. After the annular air gap 8 closes and disappears, the axial pressure plate 9 continues to move to the left and squeeze, entering the second stage. In the second stage, the pores of the front filter cotton sleeve 7 become flatter and smaller with the squeezing axis, gradually increasing the filtration capacity of the front filter cotton sleeve 7. According to the concentration of particulate matter in the external gas, the filtration pores of the front filter cotton sleeve 7 are automatically adjusted. When the particulate matter concentration is high, the filtration pores of the front filter cotton sleeve 7 are reduced to further improve the coarse filtration effect of the gas, thereby extending the life of the internal filter 2.

[0076] like Figure 4 As shown, in cases where the pretreatment module 1 is dropped and impacted, or when the lead screw shaft 904 is bent and deformed, the shaft pressure plate 9 may not be able to be driven and adjusted axially. If the shaft pressure plate 9 is stuck and cannot be driven to the left, the concentration of gas particles in the external environment will increase, and the coarse filtration effect of the pre-filter cotton sleeve 7 will be insufficient, which will lead to a rapid reduction in the lifespan of the internal filter 2 and its scrapping.

[0077] This invention, through the setting of the counterweight ring 602, when the module shell 6 is impacted, the counterweight ring 602 will compress the annular outer bladder 603 due to inertia. The gas in the annular outer bladder 603 enters the flexible connecting tube 605 through the cross-shaped air passage 604, such as... Figure 5 As shown, the drive piston 612 moves to the right.

[0078] The tapered portion 617 will only press against the expansion portion 616 when the axial pressure plate 9 is in the rightmost position, causing the triangular protrusion 615 to move upward and away from the end of the pusher 613. The piston body 612 can only drive the pusher 613 to press and push the locking spring 611 when the axial pressure plate 9 is in the rightmost position and the front filter sleeve 7 is in a non-pressed state. This causes the locking spring 611 to be pulled out of the locking groove 610, releasing the lock between the sub-control state plate 606 and the central tube 609. At this time, under the elastic tension of the tension spring 608, the sub-control state plate 606 moves away from the axial pressure plate 9. The sub-control state plate 606 then presses against the front filter sleeve 7, causing the annular air gap 8 in the front filter sleeve 7 to close and disappear, automatically entering a safer coarse filtration mode.

[0079] If the axial pressure plate 9 is already under pressure on the front filter sleeve 7 when the module housing 6 is impacted, there is no need to trigger the pressure adjustment protection of the front filter sleeve 7. At this time, because the axial pressure plate 9 and the lead screw sleeve 903 and other structures have moved a certain distance to the left relative to the lead screw shaft 904, such as... Figure 5As shown, the tapered portion 617 and the expansion portion 616 do not come into contact. Under the elastic force of the state locking spring 614, the triangular protrusion 615 will move down and lock onto the end of the pusher 613, so that the impact flow medium delivered by the flexible connecting tube 605 will not push the piston body 612 and the pusher 613 to the right. At this time, the locking spring 611 is always inserted in the locking groove 610, maintaining the relative locking between the sub-control state disk 606 and the central tube 609.

[0080] With the above structure, when the filter cotton sleeve 7 is not compressed, the annular air gap 8 is automatically closed when the pretreatment module 1 is impacted, prioritizing the life of the internal filter 2. This avoids the risk that the concentration of particulate matter in the external environment will increase and the internal filter 2 cannot be protected in time after the impact causes the lead screw shaft 904 to deform and fail. It can selectively carry out automatic prevention and control according to different situations.

[0081] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A handheld nitrous oxide rapid detection device, comprising hardware components and an algorithm processing system, characterized in that: The hardware components include a control circuit board, in which the algorithm processing system runs; The hardware components also include a pre-processing module, an internal filter, a sensor array, an air chamber, and an air pump. Under the control of the control circuit board, the air pump drives the external gas to be drawn in through the pre-processing module, processed by the internal filter, and then enters the sensor array and air chamber for detection. The pretreatment module includes a module shell and a pre-filter cotton sleeve disposed inside the module shell. External gas enters the pretreatment module and undergoes coarse filtration through the pre-filter cotton sleeve. An annular air gap is provided inside the pre-filter cotton sleeve, and a axial pressure plate is provided at the end of the pre-filter cotton sleeve. The axial pressure plate compresses the pre-filter cotton sleeve to different degrees by axial movement. Under the first stage of compression, the annular air gap closes and disappears. Under the second stage of compression, the porosity of the pre-filter cotton sleeve decreases.

2. The handheld nitrous oxide rapid detection device according to claim 1, characterized in that: The annular air gaps are provided in several groups and are evenly distributed along the axial direction of the front filter cotton sleeve. The cross-section of the annular air gaps is elliptical and disappears when the front filter cotton sleeve is subjected to axial compression.

3. The handheld nitrous oxide rapid detection device according to claim 1, characterized in that: The external part of the shaft pressure plate is fixedly provided with a sliding protrusion, and the inner wall of the module shell is provided with a wall sliding groove. The sliding protrusion slides along the wall sliding groove to limit the movement, so that the shaft pressure plate can only move axially.

4. The handheld nitrous oxide rapid detection device according to claim 1, characterized in that: A lead screw sleeve is fixedly installed on one side of the shaft pressure plate, and a lead screw shaft is screwed into the lead screw sleeve. When the lead screw shaft rotates, it can drive the shaft pressure plate to move axially. A rotor frame is fixedly mounted on the outside of the lead screw shaft, and a rotor part is fixedly mounted on the outside of the rotor frame. A stator part is fixedly mounted on the inner wall of the module shell, and the stator part can drive the rotor part to rotate.

5. The handheld nitrous oxide rapid detection device according to claim 4, characterized in that: An inner partition ring is fixedly installed on the inner wall of the module shell. A counterweight ring is installed between the inner partition ring and the end wall of the module shell. An annular outer bladder is installed on both sides and the outer surface of the counterweight ring. A cross-shaped air passage is opened inside the counterweight ring, and the annular outer bladders are interconnected through the cross-shaped air passage. A flexible connecting tube is installed outside the counterweight ring, and one end of the flexible connecting tube is connected to the cross-shaped air passage.

6. The handheld nitrous oxide rapid detection device according to claim 5, characterized in that: A control panel is provided on one side of the pressure plate. The front filter cotton sleeve is disposed between the control panel and the inner partition ring. A closed awning is fixedly provided on the outside of the pressure plate. The control panel slides and seals against the closed awning. A tension spring is provided on one side of the control panel. The tension spring applies tension to the control panel, causing the control panel to tend to move away from the pressure plate.

7. The handheld nitrous oxide rapid detection device according to claim 6, characterized in that: A central tube is fixedly installed at the center of the shaft pressure plate, a locking groove is opened in the sub-control status plate, a locking spring is installed in the central tube, and the end of the locking spring is inserted into the locking groove, so that the positions of the shaft pressure plate and the sub-control status plate are locked relative to each other. The central tube is equipped with a piston body, and a pusher is fixedly mounted on the piston body. The end of the central tube is connected to the flexible connecting tube. When the annular outer bladder is compressed, the piston body is driven to move by the medium, so that the pusher compresses the locking spring, and the end of the locking spring is pulled out from the locking groove.

8. The handheld nitrous oxide rapid detection device according to claim 7, characterized in that: The central tube is equipped with a state locking spring, and the state locking spring is fixedly provided with a triangular protrusion and an expansion portion; the end of the lead screw shaft is fixedly provided with a tapered portion.

9. The handheld nitrous oxide rapid detection device according to claim 8, characterized in that: When the filter cotton sleeve is under compression, the triangular protrusion is stuck at the end of the pusher, so that the pusher will not compress the locking spring. When the filter cotton sleeve is in a non-compressed state, the expansion part is pushed to expand and move by the conical part. At this time, the state locking spring is elastically bent, and the triangular protrusion moves away from the end of the pusher.

10. The handheld nitrous oxide rapid detection device according to claim 1, characterized in that: The algorithm processing system includes, in sequence, a signal preprocessing module, a multidimensional feature extraction module, a deep model analysis module, and a result output and feedback module; Using the raw signals collected by the sensor array and the gas chamber as input, the system processes the signals through an algorithm processing system to achieve nitrous oxide detection.

Images