Multi-mode adaptive air quality detection system

Abstract

The invention relates to the technical field of intelligent environment monitoring, and discloses a multi-mode adaptive air quality detection system, which comprises an air quality acquisition module, a dynamic closed-loop detection module, a multi-source data access module, a fusion and output module, an installation module, a wireless docking station module and an interaction and linkage module, by dynamically responding to environmental changes and integrating multi-dimensional data, air quality detection is upgraded from static parameter presentation to dynamic accurate management, detection deviation can be automatically corrected to guarantee data reliability, pollution sources can be positioned by combining actual scenes and user behaviors, detection is not limited to numerical feedback any more, and the detection efficiency is improved. The system can provide a clear direction for pollution treatment, flexibly adapt to different installation scenes, break through the limitation of traditional fixed installation, meet diversified use requirements, and allow air quality information to be perceived more easily through visual interaction and multi-device cooperative linkage, so that a user can clearly grasp pollution distribution and diffusion conditions.

Description

Technical Field

[0001] This invention relates to the field of intelligent environmental monitoring technology, and more specifically discloses a multimodal adaptive air quality detection system. Background Technology

[0002] Air quality testing is the process of detecting whether the air quality is good or bad. It mainly monitors the concentration of pollutants in the atmosphere or indoor environment and evaluates the environmental quality according to standards.

[0003] The prior art patent document with authorization announcement number CN219913379U discloses "an air detection box", which includes: a front shell with a quick-access button; a rear shell with pins for inserting into a matching socket; the front shell and the rear shell combined to form a receiving cavity, which contains: a PM2.5 sensor and a CO2 sensor, which are electrically connected to a wireless serial port module, respectively; the wireless serial port module is fixed on the rear shell and is used to transmit the collected data to the controller of an air purification device; and an AC-DC module, which is electrically connected to the PM2.5 sensor, the CO2 sensor and the wireless serial port module, respectively, and provides them with driving power.

[0004] Patent document CN119618311A discloses "An intelligent environmental monitoring device integrating NB-IoT and an embedded system," comprising a sensor module including temperature, humidity, and air quality sensor units; an embedded processing module including data receiving, processing, and fusion units; an NB-IoT communication module; and a power supply module. This device collects environmental data, processes and fuses the data, and transmits it to a remote monitoring terminal via NB-IoT. This invention enables real-time acquisition, processing, and remote transmission of environmental data, improving the accuracy and efficiency of environmental monitoring.

[0005] While existing technologies can achieve basic air quality parameter detection, data transmission, and simple purification linkage functions, providing some support for environmental monitoring, they lack sensor calibration capabilities, relying solely on factory calibration. Long-term use can easily lead to data drift, resulting in decreased detection accuracy. Furthermore, the data fusion dimension is limited, integrating only the sensor's own data without linking to smart home device status and user behavior logs, making it impossible to trace pollution sources. Moreover, most of these technologies are fixed wall-mounted or plug-in designs, which cannot meet the installation needs of various scenarios such as wall embedding, desktop placement, and ceiling hanging in old house renovations and rental settings. They also cannot connect to external sensors for special pollutants such as radiation and ozone. Finally, they lack intuitive visualization methods, and the linkage is limited to a single purification device, failing to form a closed loop for environmental management with multiple devices working together. Summary of the Invention

[0006] The main technical problem solved by this invention is to provide a multimodal adaptive air quality detection system that can solve the problems mentioned in the background art.

[0007] To address the aforementioned technical problems, according to one aspect of the present invention, more specifically, a multimodal adaptive air quality detection system includes: an air quality acquisition module that collects multi-dimensional air quality parameters through fixed sensors, performs outlier removal and standardization processing after data preprocessing, and outputs clean real-time monitoring data; a dynamic closed-loop detection module that uses nano-standard gas capsules as its core, combines temperature and humidity control to adjust the calibration environment in real time, and then triggers standard gas release and sensor calibration actions, while simultaneously monitoring the sealing and effectiveness of the standard gas capsules; a multi-source data access module that obtains the operating status of smart home devices such as doors, windows, air conditioners, and stoves through smart home integration, performs data cleaning to remove redundant and abnormal data, and records user movement trajectories and device operation duration behavior information; and a fusion and output module that extracts air quality parameters. The system incorporates temporal change rate, spatial distribution characteristics, and user behavior correlation features. Based on multi-source features, it predicts pollution trends and traces pollution sources. A pre-built pollution pattern library is then used to match pollution types, generating detection reports and visualized data. The installation module enables rapid device disassembly and sensor replacement, supports multiple length adjustments to adapt to different installation scenarios, and provides adapters and bases for wall embedding, desktop placement, and ceiling hanging. It also optimizes airflow paths for device heat dissipation. The wireless expansion dock module enables high-speed data transmission with external sensors, supports radiation and ozone extension sensor access, and allocates data transmission bandwidth according to the importance of monitoring parameters. The interaction and linkage module generates a 3D pollution distribution cloud map and provides touch and voice operation interfaces, triggering commands to activate the fresh air system and start the range hood based on pollution detection results.

[0008] Furthermore, the air quality acquisition module includes: a fixed sensor module and a data preprocessing module;

[0009] Fixed sensor module: Includes PM2.5 sensor, PM10 sensor, TVOC sensor, formaldehyde sensor, CO sensor, CO2 sensor, temperature and humidity sensor, and oxygen sensor, arranged in a ring;

[0010] Data preprocessing module: The module uses the 3σ criterion to remove abnormal values ​​of sensor pulses, standardizes the monitoring data using Z-score, converts parameters of different dimensions into unified standard values, and uses the KNN algorithm to complete missing data.

[0011] Furthermore, the dynamic closed-loop detection module includes: a nano-standard gas capsule module, a temperature and humidity control module, a standard execution module, and a capsule status detection module;

[0012] Nanoscale standard gas capsule module: It is a sealed nanocavity with a diameter of 5mm and a length of 10mm, containing 50ppb±5% formaldehyde standard gas or 500ppm±5% CO2 standard gas. One end of the cavity is equipped with a PDMS slow-release membrane with a thickness of 0.1mm±0.02mm.

[0013] Temperature and humidity control module: Built-in temperature and humidity sensors to collect ambient temperature and humidity data in real time;

[0014] Standard execution module: After receiving the trigger signal from the temperature and humidity control module, it controls the electromagnetic release valve to open, so that the standard gas is directionally delivered to the sensor gas chamber along the L-shaped guide groove, with a calibration cycle of ≤2 minutes;

[0015] Capsule status detection module: The micro-pressure sensor detects the pressure inside the capsule. When the pressure deviates from the standard atmospheric pressure by more than 3%, the capsule is determined to be faulty and a replacement reminder signal is output.

[0016] Furthermore, the multi-source data access module includes: a smart home integration module, a data cleaning module, and a user behavior log collection module;

[0017] Smart home integration module: Connects to smart home system API via Wi-Fi or Bluetooth 5.0 to obtain data on door and window open / close status, air conditioner operating mode, stove open / close status, range hood operating status, and fresh air system operating status, with a response latency of ≤1s;

[0018] Data cleaning module: Uses a sliding window to filter pulse interference in smart home data, completes missing device status data by associating device data within the same time period, and removes redundant and duplicate data;

[0019] User behavior log collection module: Acquires user data through device touch panel and mobile APP, records user movement trajectory and dwell time in kitchen, bedroom and living room areas, and marks user cooking and ventilation behaviors.

[0020] Furthermore, the fusion and output module includes: a feature extraction module, a fusion inference module, a contamination matching module, and a result output module;

[0021] Feature extraction module: Extracts the time-domain features of the rate of change and cumulative values ​​of air quality parameters;

[0022] Fusion Inference Module: It uses an LSTM time series analysis model to predict pollution trends, combines a Bayesian-particle filter algorithm to correct the uncertainty of multi-source data, and outputs pollution trend prediction results and pollution source confidence.

[0023] Pollution Matching Module: A pre-built pollution pattern library is used to compare the features output by the fusion reasoning module with the feature parameter combinations in the pattern library. The pattern with the highest matching degree is used as the preliminary result of the pollution type determination.

[0024] Results output module: Generates a test report that includes pollution source type, occurrence area, occurrence time, confidence level and treatment recommendations. It also outputs 3D pollution cloud map data to the interaction and linkage module, supporting multi-dimensional data query and historical record backtracking.

[0025] Furthermore, the installation module includes: a magnetic quick-release panel module, a three-section telescopic bracket module, a multi-scenario adaptable accessory module, and an air duct heat dissipation module;

[0026] Magnetic quick-release panel module: The main body of the panel is made of anodized aluminum alloy with a diameter of 86×86×5mm. It has four N52 neodymium iron boron magnetic rings with a diameter of 8mm×1.5mm built in, four 24-pin gold-plated contacts on the inside, and four 0.5mm pop-up plastic clips on the edge.

[0027] Three-section telescopic support module: Made of T700 grade carbon fiber tube, it is designed in three sections with diameters of Φ12mm, Φ8mm and Φ5mm respectively. The inter-section fit tolerance is H7 / g6. The compressed length is 60mm±0.2mm and the unfolded length is 300mm. Each section connection is equipped with a Φ2mm spring pin and a damping washer.

[0028] Multi-scenario adaptable accessory modules: including a dedicated adapter board, a desktop silicone base, and a ceiling-mounted metal hanging ring;

[0029] Airflow cooling module: The back of the panel is equipped with a honeycomb array of holes with a diameter of 2mm and a center distance of 3mm, forming a vortex airflow channel of "bottom air intake → flow through sensor array → heat dissipation along PCB board → air exhaust on both sides", with a wind speed ≥0.5m / s.

[0030] Furthermore, the wireless expansion dock module includes: a UWB wireless communication module, a sensor interface module, and a parameter priority scheduling module;

[0031] UWB wireless communication module: adopts IEEE802.15.4z protocol, transmission distance ≥10m, transmission rate 1Mbps;

[0032] Sensor interface module: Equipped with a sensor interface, supporting external radiation sensors and ozone sensors;

[0033] Parameter priority scheduling module: PM2.5, formaldehyde, and CO2 are set as high priority parameters, while radiation and ozone are set as low priority parameters. High priority data occupies the transmission bandwidth first, and low priority data is transmitted in the gaps. The transmission delay is ≤100ms.

[0034] Furthermore, the interaction and linkage module includes: an AR pollution cloud map module, an interactive control module, and a smart home linkage module;

[0035] AR pollution cloud map module: Connects directly to AR glasses via Wi-Fi, outputs a 1080P resolution 3D pollution cloud map, updates at 2fps, marks pollution concentration with color, and marks pollution diffusion paths;

[0036] Interactive control module: Touch buttons for "Start Range Hood" and "Turn on Fresh Air" pop up on the AR interface, supporting voice command input with a command response delay of ≤2s;

[0037] Smart home linkage module: Receives instructions from the interactive control module or the fusion and output module, sends control signals to the corresponding devices through the smart home API, and receives feedback information after the device executes the command and displays it on the AR interface.

[0038] The beneficial effects of the multimodal adaptive air quality detection system of the present invention are as follows:

[0039] By dynamically responding to environmental changes and integrating multi-dimensional data, air quality monitoring is upgraded from static parameter presentation to dynamic and precise management. It can automatically correct detection deviations to ensure data reliability, and combine actual scenarios and user behavior to pinpoint the root causes of pollution. This allows monitoring to go beyond numerical feedback and provide clear direction for pollution control. It is also flexible enough to adapt to different installation scenarios, breaking the limitations of traditional fixed installations and meeting diverse usage needs. Through intuitive visual interaction and multi-device collaborative linkage, air quality information is more easily perceived, and users can clearly understand the distribution and diffusion of pollution. The system can also automatically trigger appropriate environmental improvement actions, forming a complete closed loop of "detection-analysis-treatment". This not only improves the user experience but also helps build an intelligent and integrated environmental management system, effectively filling the gaps in existing technologies in terms of functional integration and practical application adaptability. Attached Figure Description

[0040] The present invention will now be described in further detail with reference to the accompanying drawings and specific implementation methods.

[0041] Figure 1 This is a schematic diagram of the system principle. Detailed Implementation

[0042] The present invention will be described in detail below with reference to the accompanying drawings and embodiments. It should be noted that, unless otherwise specified, the embodiments and features described in the present application can be combined with each other.

[0043] According to one aspect of the invention, such as Figure 1 As shown, a multimodal adaptive air quality detection system is provided, comprising:

[0044] The air quality acquisition module collects multi-dimensional air quality parameters through fixed sensors, performs outlier removal and standardization after data preprocessing, and outputs clean, real-time monitoring data. This module includes:

[0045] Fixed sensor module: Includes PM2.5 sensor, PM10 sensor, TVOC sensor, formaldehyde sensor, CO sensor, CO2 sensor, temperature and humidity sensor, and oxygen sensor, arranged in a ring;

[0046] The preset detection ranges for the sensors are: PM2.5 (0-1000 μg / m³). 3 PM10 (0-2000μg / m 3 TVOC (0-10mg / m³) 3 Formaldehyde (0-1 mg / m³) 3 ), CO (0-100ppm), CO2 (0-5000ppm), temperature (-10℃~60℃), humidity (0-100%RH), oxygen (19.5%-25%VOL);

[0047] Meanwhile, the sensor adopts a synchronous sampling mechanism with a sampling frequency of 1Hz, generating one average data point every 5 minutes and marking the collection timestamp to ensure data temporal consistency;

[0048] Finally, a protective mesh cover (e.g., 0.5mm aperture, made of 304 stainless steel) is provided on the outside of the sensor to prevent dust and foreign objects from directly contacting the sensor probe and extend the sensor's service life.

[0049] Data preprocessing module: The 3σ criterion is used to remove abnormal values ​​of sensor pulses, and the monitoring data is standardized by Z-score to convert parameters of different dimensions into unified standard values. At the same time, the KNN algorithm is used to complete the missing data.

[0050] The process of removing outliers involves first calculating the mean and standard deviation of 10 consecutive sensor samples, then removing values ​​that deviate from the mean by more than 3 times the standard deviation. If the number of valid samples is less than 5, it is determined to be an "abnormal sampling" and a resampling command is triggered.

[0051] The standardization process is Z-score standardization, and its formula is:

[0052]

[0053] In the formula, X is the original monitoring data, μ is the mean of the historical data of the parameter in the past 24 hours, σ is the corresponding standard deviation, and the standardized data are distributed in the interval [-3,3] to eliminate the difference in the dimensions of different parameters;

[0054] Finally, missing data completion involves selecting three adjacent normal data points (K=3) from the parameter within the past hour as neighbor samples for the KNN algorithm when a certain parameter is missing. The missing value is then filled in by calculating the weighted average of the Euclidean distance between the samples, ensuring data continuity.

[0055] The dynamic closed-loop detection module, centered on a nano-standard gas capsule, combines temperature and humidity control to adjust the calibration environment in real time, then triggers the release of standard gas and sensor calibration, while simultaneously monitoring the capsule's seal and effectiveness. This module includes:

[0056] Nanoscale standard gas capsule module: It is a sealed nanocavity with a diameter of 5mm and a length of 10mm, containing 50ppb±5% formaldehyde standard gas or 500ppm±5% CO2 standard gas. One end of the cavity is equipped with a PDMS slow-release membrane with a thickness of 0.1mm±0.02mm.

[0057] The sealed nanocavity is made of medical-grade 316L stainless steel with a wall thickness of 0.3mm, ensuring no leakage during long-term storage of the standard gas. The PDMS sustained-release membrane has a stable permeability coefficient of 1.2×10⁻⁶. -9 m 2 / s (at 25℃), the standard gas can be controlled to be released into the sensor gas chamber at a constant rate of 0.0025ml / s;

[0058] In addition, the standard gases inside the capsule are balanced with nitrogen (formaldehyde standard gas) and dry air (CO2 standard gas) to avoid interference from impurity gases with calibration accuracy.

[0059] Temperature and humidity control module: Built-in temperature and humidity sensors to collect ambient temperature and humidity data in real time;

[0060] The temperature and humidity sensors have detection accuracies of ±0.3℃ for temperature and ±2%RH for humidity. The acquisition frequency is 1Hz, which is synchronized with the sensor module. The data is directly transmitted to the control chip (model STM32L431) built into the module.

[0061] In addition, when the ambient temperature is <15℃, the module will automatically control the 0.5W heating resistance wire (resistance value 50Ω±1%) wrapped around the outside of the capsule to be energized, raising the temperature around the capsule to 25℃±1℃, compensating for the problem of decreased air permeability of PDMS membrane at low temperatures.

[0062] Standard execution module: After receiving the trigger signal from the temperature and humidity control module, it controls the electromagnetic release valve to open, so that the standard gas is directionally delivered to the sensor gas chamber along the L-shaped guide groove, with a calibration cycle of ≤2 minutes;

[0063] The electromagnetic release valve operates at a voltage of 5VDC, has a coil resistance of 22Ω±5%, an opening response time of ≤100ms, and an L-shaped guide groove with a width of 1.2mm and a depth of 0.8mm. The inner wall is polished (roughness Ra≤0.8μm) to reduce the flow resistance of the standard gas.

[0064] In addition, after the standard gas is released, the module will control the sensor gas chamber to remain sealed for 1 minute to ensure that the standard gas is in full contact with the sensor probe before starting data acquisition and calibration calculation. The calibration cycle from the generation of the trigger signal to the completion of the coefficient update does not exceed 2 minutes.

[0065] Capsule status detection module: Detects the pressure inside the capsule using a micro-pressure sensor. When the pressure deviates from the standard atmospheric pressure by more than 3%, the capsule is deemed to be faulty and a replacement reminder signal is output.

[0066] The micro-pressure sensor has a detection accuracy of ±1Pa and a sampling frequency of 0.5Hz. It monitors the pressure changes inside the capsule in real time, with the standard atmospheric pressure set at 101.325kPa as the reference value.

[0067] In addition, if a pressure deviation of >3% is detected three times consecutively, or a pressure drop of >10kPa is detected in a single instance (indicating capsule rupture), the module will display "Standard gas capsule failure" on the device's touch panel and simultaneously push a replacement reminder to the mobile APP to avoid invalid calibration;

[0068] For capsules that have not expired, the module will record pressure data once every 24 hours and generate a capsule life curve to help users predict when to replace them.

[0069] The multi-source data access module acquires the operating status of smart home devices such as doors, windows, air conditioners, and stoves through smart home integration. After data cleaning to remove redundant and abnormal data, it records user movement trajectories and device operation durations. This module includes:

[0070] Smart home integration module: Connects to smart home system API via Wi-Fi or Bluetooth 5.0 to obtain data on door and window open / close status, air conditioner operating mode, stove open / close status, range hood operating status, and fresh air system operating status, with a response latency of ≤1s;

[0071] Among them, Wi-Fi adopts the 802.11b / g / n protocol, Bluetooth 5.0 supports BLE low power mode, and OAuth2.0 authentication mechanism is used to ensure data transmission security during docking. It can uniformly convert data of different formats such as JSON and XML into structured data that the system can recognize.

[0072] The obtained door and window opening and closing status is recorded as a Boolean value (True / False), the air conditioner operation mode is marked by an enumeration value (e.g., cooling / heating / ventilation / automatic), and the operation status of the stove, range hood, and fresh air system includes "on / off" and corresponding gear information (e.g., range hood strong wind / weak wind).

[0073] In addition, the module will send a data request to the smart home system every 30 seconds. If there is no response after 3 consecutive requests, the communication will be interrupted, and the user will be reminded to check the network connection by flashing the device indicator light (red light once per second).

[0074] Data cleaning module: Uses a sliding window to filter pulse interference in smart home data, completes missing device status data by associating device data within the same time period, and removes redundant and duplicate data;

[0075] The sliding window is set to a window size of 5 minutes. By calculating the variance of the data within the window, pulse interference data with variances exceeding a preset threshold (e.g., variance of door and window opening / closing status > 0.5) is filtered out, such as invalid signals of instantaneous door and window opening / closing (lasting < 10 seconds).

[0076] In addition, when completing missing data, if the status of the kitchen stove is missing, it can be inferred by associating it with the status of the range hood during the same time period (if it is on, the stove is likely to be on) and the user's kitchen stay record (if there is a stay, the probability of the stove being on increases).

[0077] Finally, redundant and duplicate data removal adopts timestamp-based deduplication logic. If the same device sends the same status data continuously within 10 seconds, only the first data is retained, reducing data storage and processing pressure.

[0078] User behavior log collection module: Acquires user data through device touch panel and mobile APP, records user movement trajectory and dwell time in kitchen, bedroom and living room areas, and marks user cooking and ventilation behaviors;

[0079] The device's touch panel collects user operation records (such as manually starting the detection function and viewing the detection report). The mobile APP obtains location permissions through user authorization and records the user's movement trajectory in different areas by combining indoor positioning (accuracy ±0.3m). Only stays of ≥1 minute are recorded as valid stays.

[0080] Meanwhile, when tagging user behavior, "cooking behavior" must meet two conditions simultaneously: "stove turned on for ≥15 minutes" and "user stays in the kitchen for ≥10 minutes". "Ventilation behavior" must meet the conditions of "doors and windows opened for ≥30 minutes" and "outdoor wind force ≤3" (outdoor wind force data is obtained by connecting to the weather API) to ensure the accuracy of behavior tagging.

[0081] The fusion and output module extracts the temporal rate of change, spatial distribution characteristics, and user behavior correlation features of air quality parameters. Simultaneously, based on multi-source features, it enables pollution trend prediction and pollution source tracing. Then, it calls a pre-built pollution pattern library to complete pollution type matching and generate detection reports and visualized data. This module includes:

[0082] Feature extraction module: Extracts the time-domain features of the rate of change and cumulative values ​​of air quality parameters;

[0083] Specifically, the rate of change in the time-domain characteristics refers to the 10-minute rate of change of core pollution parameters (e.g., PM2.5, TVOC). When the rate of change is >10 μg / m³, the change rate is considered to be significant. 3 / min (PM2.5) or >0.2mg / m 3 The value of / min (TVOC) is marked as a "sudden increase" characteristic, and the cumulative value is the 24-hour average concentration of formaldehyde and CO2, which is used to identify long-term pollution problems;

[0084] Simultaneously, spatial distribution features will be extracted, and pollution occurrence areas (such as kitchens, bedrooms, and living rooms) will be marked based on the device positioning module. Each area will be assigned a risk weight (e.g., 0.6 for kitchens, 0.3 for bedrooms, and 0.1 for living rooms). Combined with the time overlap between user behavior and pollution events (e.g., the proportion of overlap between cooking time and the sudden increase in PM2.5), a multi-dimensional feature set will be formed.

[0085] Fusion Inference Module: It uses an LSTM time series analysis model to predict pollution trends, combines a Bayesian-particle filter algorithm to correct the uncertainty of multi-source data, and outputs pollution trend prediction results and pollution source confidence.

[0086] The LSTM time series analysis model adopts a 3-layer stacked structure (input layer + 2 hidden layers + output layer). Each hidden layer contains 64 neurons. It uses the multi-dimensional features of the past hour as input to predict the pollution concentration change trend in the next hour. The model training uses the Adam optimizer and MSE loss function.

[0087] In addition, the Bayesian-particle filter algorithm sets the number of particles to 1000. By updating the particle weights, it corrects the contradictions between multi-source data (such as conflicting data like "doors and windows are closed but users claim ventilation"). It integrates the LSTM prediction probability, the overlap between behavior and pollution, and regional risk weights to calculate the confidence of pollution sources, so that each factor can reasonably reflect its influence in the result judgment.

[0088] Finally, for example, when the confidence level is >85%, it is judged as a high-confidence result and output directly; when the confidence level is ≤85%, the data collection is extended by 30 minutes and the calculation is recalculated to ensure the accuracy of traceability.

[0089] Pollution Matching Module: A pre-built pollution pattern library is used to compare the features output by the fusion reasoning module with the feature parameter combinations in the pattern library. The pattern with the highest matching degree is used as the preliminary result of the pollution type determination.

[0090] The pre-set pollution model library contains typical pollution models, and each model corresponds to a specific combination of characteristic parameters (for example, the characteristic combination for the "renovation release" model is "TVOC > 1.2 mg / m³"). 3 +Formaldehyde > 0.1 mg / m³ 3 +Doors and windows closed for ≥2 hours+Bedroom area", "Cooking pollution" mode is "PM2.5 > 150 μg / m³" 3 +Stove on +Range hood off +Kitchen area);

[0091] The feature comparison uses the cosine similarity algorithm to calculate the similarity between the features output by the fusion reasoning module and the features of each pattern in the pattern library. The pattern with the highest similarity (e.g., ≥80%) is the preliminary judgment result. If the highest similarity is <60%, it is marked as "unknown contamination type" and manual review is triggered.

[0092] Finally, the pattern library supports monthly updates. By collecting feature parameters of newly added pollution cases, it optimizes existing pattern features or adds new pattern types to improve matching coverage.

[0093] Results output module: Generates a test report containing pollution source type, occurrence area, occurrence time, confidence level and treatment recommendations, and simultaneously outputs 3D pollution cloud map data to the interaction and linkage module, supporting multi-dimensional data query and historical record backtracking;

[0094] The test report is automatically generated in PDF format and includes key time points of the pollution event (e.g., "PM2.5 surge start time: 2024-05-20 18:30"), related equipment status (e.g., "stove operation record: 18:25-19:00"), and user behavior (e.g., "user's time in the kitchen: 18:28-18:50"). The treatment recommendations are customized for different types of pollution, such as "renovation release" recommending "turn on the fresh air system + place activated carbon".

[0095] In addition, the output 3D pollution cloud map data includes pollution concentration gradient information, marked with colors of "green (compliant) - yellow (light pollution) - red (heavy pollution)", and also marks the pollution diffusion path (e.g., arrows indicating "from stove → range hood → window"). The data format is adapted to the rendering requirements of AR glasses.

[0096] The multi-dimensional data query supports filtering historical data by "time range (e.g., the last 24 hours), pollution type (e.g., cooking pollution), and region (e.g., kitchen)". Users can manually export data in Excel format for analysis.

[0097] The installation module enables quick device assembly and disassembly, as well as sensor replacement. It supports multiple length adjustments to adapt to different installation scenarios and provides adapters and bases for wall mounting, desktop placement, and ceiling suspension. It also optimizes airflow for efficient device cooling. This module includes:

[0098] Magnetic quick-release panel module: The main body of the panel is made of anodized aluminum alloy with a diameter of 86×86×5mm. It has four N52 neodymium iron boron magnetic rings with a diameter of 8mm×1.5mm built in, four 24-pin gold-plated contacts on the inside, and four 0.5mm pop-up plastic buckles on the edge.

[0099] Three-section telescopic support module: Made of T700 grade carbon fiber tube, it is designed in three sections with diameters of Φ12mm, Φ8mm and Φ5mm respectively. The inter-section fit tolerance is H7 / g6. The compressed length is 60mm±0.2mm and the unfolded length is 300mm. Each section connection is equipped with a Φ2mm spring pin and a damping washer.

[0100] Multi-scenario adaptable accessory modules: including a dedicated adapter board, a desktop silicone base, and a ceiling-mounted metal hanging ring.

[0101] Airflow cooling module: The back of the panel is equipped with a honeycomb hole array with a diameter of 2mm and a center distance of 3mm, forming a vortex airflow channel of "bottom air intake → flow through sensor array → heat dissipation along PCB board → air exhaust on both sides", with a wind speed ≥0.5m / s;

[0102] The honeycomb array is arranged in a 30×30 matrix with an opening rate of 38%. The air inlet is treated with a tapered angle (e.g., 60°) to reduce airflow resistance. A 3mm gap is reserved between the sensor array and the air duct to ensure that cold air can flow directly through the sensor chamber and carry away the heat generated by the sensor.

[0103] The main control chip (STM32H743VI) area on the PCB board has a heat dissipation copper foil (10×15mm in area and 0.03mm in thickness) in the corresponding airflow location. The copper foil is aligned with the airflow path of the airflow channel to enhance the chip's heat dissipation effect.

[0104] The wireless extension dock module enables high-speed data transmission with external sensors, supports the connection of radiation and ozone extension sensors, and allocates data transmission bandwidth according to the importance of the monitored parameters. This module includes:

[0105] UWB wireless communication module: adopts IEEE802.15.4z protocol, transmission distance ≥10m, transmission rate 1Mbps;

[0106] The communication process employs the AES-128 encryption algorithm to ensure data transmission security and has strong anti-interference capabilities.

[0107] Sensor interface module: Equipped with a sensor interface, supporting external radiation sensors and ozone sensors;

[0108] The interface type is a standard I2C / SPI interface, which is compatible with mainstream radiation sensors (detection range 0.01-10μSv / h, accuracy ±0.001μSv / h) and ozone sensors (detection range 0-1ppm, accuracy ±0.01ppm). The interface is equipped with an anti-reverse insertion structure (protruding positioning pin + groove) to avoid equipment damage caused by incorrect sensor wiring.

[0109] Parameter priority scheduling module: PM2.5, formaldehyde, and CO2 are set as high priority parameters, while radiation and ozone are set as low priority parameters. High priority data occupies the transmission bandwidth first, and low priority data is transmitted in the gaps. The transmission delay is ≤100ms.

[0110] Among them, the sampling frequency of high-priority parameters is fixed at 1Hz, and the data transmission adopts a "real-time push" mechanism to ensure that core data can be fed back in time when pollution occurs. The sampling frequency of low-priority parameters is set to 0.5Hz, and the data is transmitted in "batch packaging" (for example, every 2 data are packaged into 1 data packet) to reduce the number of transmissions and save bandwidth.

[0111] Meanwhile, parameter priorities can be manually adjusted via the device's touch panel or a mobile app, allowing users to temporarily increase the priority of specific low-priority parameters based on actual monitoring needs (such as radiation values ​​that require special attention in special environments).

[0112] The interaction and linkage module generates a 3D pollution distribution cloud map and provides touch and voice operation interfaces. Based on pollution detection results, it triggers commands to activate the fresh air system and start the range hood. This module includes:

[0113] AR pollution cloud map module: Connects directly to AR glasses via Wi-Fi, outputs a 1080P resolution 3D pollution cloud map, updates at 2fps, marks pollution concentration with color, and marks pollution diffusion paths;

[0114] Among them, the Wi-Fi connection adopts the 802.11ac protocol with a transmission rate of ≥300Mbps, ensuring real-time synchronization of 3D cloud map data without lag;

[0115] The color-coded pollution concentration levels are as follows:

[0116] Green corresponds to PM2.5 < 35 μg / m³ 3 TVOC < 0.6 mg / m³ 3 The yellow level corresponds to PM2.5 levels of 35-75 μg / m³. 3 TVOC 0.6-1.2 mg / m³ 3Light pollution, red indicates PM2.5 > 75 μg / m³ 3 TVOC > 1.2 mg / m³ 3 Severe pollution;

[0117] Finally, the pollution diffusion path is marked based on the pollutant concentration gradient changes located by UWB, and the direction of pollution diffusion is indicated by dynamic arrows (e.g., "stove → range hood → window"). The arrow movement speed is synchronized with the actual pollution diffusion rate (e.g., the arrow accelerates when the diffusion rate is >0.5m / min).

[0118] Interactive control module: Touch buttons for "Start Range Hood" and "Turn on Fresh Air" pop up on the AR interface, supporting voice command input with a command response delay of ≤2s;

[0119] The voice commands support Mandarin Chinese recognition, and the recognized keywords include "start the range hood", "turn on the fresh air", and "turn off the linkage".

[0120] Smart home linkage module: Receives instructions from the interactive control module or the fusion and output module, sends control signals to the corresponding devices through the smart home API, and receives feedback information after the device executes the command and displays it on the AR interface.

[0121] The API connection adopts a RESTful architecture, the data transmission format is JSON, and the waiting time for device response after the command is sent is ≤1.5s. If the timeout occurs, it is judged as "linkage failure" and a red warning icon is displayed on the AR interface.

[0122] In addition, the equipment execution feedback information includes "execution status (success / failure)" and "current operating parameters (such as range hood setting and fresh air volume)".

[0123] Of course, the above description is not a limitation of the present invention, and the present invention is not limited to the examples given above. Any changes, modifications, additions or substitutions made by those skilled in the art within the scope of the present invention are also within the protection scope of the present invention.

Claims

1. A multimodal adaptive air quality detection system, characterized in that, include: The air quality acquisition module collects multi-dimensional air quality parameters through fixed sensors, performs outlier removal and standardization after data preprocessing, and outputs clean real-time monitoring data. The dynamic closed-loop detection module uses nano-standard gas capsules as its core, combining temperature and humidity control to adjust the calibration environment in real time, then triggers the release of standard gas and sensor calibration, while simultaneously monitoring the sealing and effectiveness of the standard gas capsules. The multi-source data access module obtains the operating status of smart home devices such as doors, windows, air conditioners, and stoves through smart home integration, cleans and removes redundant and abnormal data, and records user movement trajectories and device operation durations. The fusion and output module extracts the temporal rate of change, spatial distribution characteristics, and user behavior correlation characteristics of air quality parameters, and simultaneously achieves pollution trend prediction and pollution source tracing based on multi-source features, then calls a pre-set pollution model. The system completes pollution type matching and generates detection reports and visualized data. The installation module enables quick device disassembly and sensor replacement, supports multiple length adjustments to adapt to different installation scenarios, and provides adapters and bases for wall embedding, desktop placement, and ceiling hanging. It also optimizes airflow paths for device heat dissipation. The wireless expansion dock module enables high-speed data transmission with external sensors, supports the access of radiation and ozone expansion sensors, and allocates data transmission bandwidth according to the importance of monitoring parameters. The interaction and linkage module generates a three-dimensional pollution distribution cloud map and provides touch and voice operation interfaces, triggering linkage commands to start fresh air and range hood based on pollution detection results.

2. The multimodal adaptive air quality detection system according to claim 1, characterized in that: The air quality acquisition module includes: a fixed sensor module and a data preprocessing module; Fixed sensor module: Includes PM2.5 sensor, PM10 sensor, TVOC sensor, formaldehyde sensor, CO sensor, CO2 sensor, temperature and humidity sensor, and oxygen sensor, arranged in a ring; Data preprocessing module: The module uses the 3σ criterion to remove abnormal values ​​of sensor pulses, standardizes the monitoring data using Z-score, converts parameters of different dimensions into unified standard values, and uses the KNN algorithm to complete missing data.

3. The multimodal adaptive air quality detection system according to claim 1, characterized in that: The dynamic closed-loop detection module includes: a nano-standard gas capsule module, a temperature and humidity control module, a standard execution module, and a capsule status detection module; Nanoscale standard gas capsule module: It is a sealed nanocavity with a diameter of 5mm and a length of 10mm, containing 50ppb±5% formaldehyde standard gas or 500ppm±5% CO2 standard gas. One end of the cavity is equipped with a PDMS slow-release membrane with a thickness of 0.1mm±0.02mm. Temperature and humidity control module: Built-in temperature and humidity sensors to collect ambient temperature and humidity data in real time; Standard execution module: After receiving the trigger signal from the temperature and humidity control module, it controls the electromagnetic release valve to open, so that the standard gas is directionally delivered to the sensor gas chamber along the L-shaped guide groove, with a calibration cycle of ≤2 minutes; Capsule status detection module: The micro-pressure sensor detects the pressure inside the capsule. When the pressure deviates from the standard atmospheric pressure by more than 3%, the capsule is determined to be faulty and a replacement reminder signal is output.

4. The multimodal adaptive air quality detection system according to claim 1, characterized in that: The multi-source data access module includes: a smart home integration module, a data cleaning module, and a user behavior log collection module; Smart home integration module: Connects to smart home system API via Wi-Fi or Bluetooth 5.0 to obtain data on door and window open / close status, air conditioner operating mode, stove open / close status, range hood operating status, and fresh air system operating status, with a response latency of ≤1s; Data cleaning module: Uses a sliding window to filter pulse interference in smart home data, completes missing device status data by associating device data within the same time period, and removes redundant and duplicate data; User behavior log collection module: Acquires user data through device touch panel and mobile APP, records user movement trajectory and dwell time in kitchen, bedroom and living room areas, and marks user cooking and ventilation behaviors.

5. The multimodal adaptive air quality detection system according to claim 1, characterized in that: The fusion and output module includes: a feature extraction module, a fusion inference module, a contamination matching module, and a result output module; Feature extraction module: Extracts the time-domain features of the rate of change and cumulative values ​​of air quality parameters; Fusion Inference Module: It uses an LSTM time series analysis model to predict pollution trends, combines a Bayesian-particle filter algorithm to correct the uncertainty of multi-source data, and outputs pollution trend prediction results and pollution source confidence. Pollution Matching Module: A pre-built pollution pattern library is used to compare the features output by the fusion reasoning module with the feature parameter combinations in the pattern library. The pattern with the highest matching degree is used as the preliminary result of the pollution type determination. Results output module: Generates a test report that includes pollution source type, occurrence area, occurrence time, confidence level and treatment recommendations. It also outputs 3D pollution cloud map data to the interaction and linkage module, supporting multi-dimensional data query and historical record backtracking.

6. The multimodal adaptive air quality detection system according to claim 1, characterized in that: The installation module includes: a magnetic quick-release panel module, a three-section telescopic bracket module, a multi-scene adaptable accessory module, and an air duct heat dissipation module; Magnetic quick-release panel module: The main body of the panel is made of anodized aluminum alloy with a diameter of 86×86×5mm. It has four N52 neodymium iron boron magnetic rings with a diameter of 8mm×1.5mm built in, four 24-pin gold-plated contacts on the inside, and four 0.5mm pop-up plastic clips on the edge. Three-section telescopic support module: Made of T700 grade carbon fiber tube, it is designed in three sections with diameters of Φ12mm, Φ8mm and Φ5mm respectively. The inter-section fit tolerance is H7 / g6. The compressed length is 60mm±0.2mm and the unfolded length is 300mm. Each section connection is equipped with a Φ2mm spring pin and a damping washer. Multi-scenario adaptable accessory modules: including a dedicated adapter board, a desktop silicone base, and a ceiling-mounted metal hanging ring; Airflow cooling module: The back of the panel is equipped with a honeycomb array of holes with a diameter of 2mm and a center distance of 3mm, forming a vortex airflow channel of "bottom air intake → flow through sensor array → heat dissipation along PCB board → air exhaust on both sides", with a wind speed ≥0.5m / s.

7. The multimodal adaptive air quality detection system according to claim 1, characterized in that: The wireless expansion dock module includes: a UWB wireless communication module, a sensor interface module, and a parameter priority scheduling module; UWB wireless communication module: adopts IEEE802.15.4z protocol, transmission distance ≥10m, transmission rate 1Mbps; Sensor interface module: Equipped with a sensor interface, supporting external radiation sensors and ozone sensors; Parameter priority scheduling module: PM2.5, formaldehyde, and CO2 are set as high priority parameters, while radiation and ozone are set as low priority parameters. High priority data occupies the transmission bandwidth first, and low priority data is transmitted in the gaps. The transmission delay is ≤100ms.

8. The multimodal adaptive air quality detection system according to claim 1, characterized in that: The interaction and linkage module includes: an AR pollution cloud map module, an interactive control module, and a smart home linkage module; AR pollution cloud map module: Connects directly to AR glasses via Wi-Fi, outputs a 1080P resolution 3D pollution cloud map, updates at 2fps, marks pollution concentration with color, and marks pollution diffusion paths; Interactive control module: Touch buttons for "Start Range Hood" and "Turn on Fresh Air" pop up on the AR interface, supporting voice command input with a command response delay of ≤2s; Smart home linkage module: Receives instructions from the interactive control module or the fusion and output module, sends control signals to the corresponding devices through the smart home API, and receives feedback information after the device executes the command and displays it on the AR interface.

Images