Intelligent safe turn-off and health diagnosis device for household heat supply system and control method of intelligent safe turn-off and health diagnosis device

Abstract

The invention relates to an intelligent safe turn-off and health diagnosis device for a household heat supply system and a control method thereof, and belongs to the technical field of intelligent home furnishing and heat supply systems. The device comprises a water inlet and outlet pipeline assembly cluster which is provided with a first electric shut-off valve and a second electric shut-off valve; the multi-sensor fusion module comprises a water inlet pressure sensor, a water inlet temperature sensor, a water inlet ultrasonic sensor and a corresponding water outlet side sensor and is used for collecting pressure, temperature and flow data in real time; and the edge calculation control unit is electrically connected with the sensor and the shut-off valve. The system has the advantages that quick leakage response, system health diagnosis, predictive maintenance and remote intelligent management are realized, and the safety, reliability and intelligent level of a household heat supply system are improved.

Description

Technical Field

[0001] This invention relates to an intelligent safety shutdown and health diagnosis device and its control method for a home heating system, belonging to the technical field of smart home and heating system. Background Technology

[0002] With the increasing popularity of centralized heating, the safety and reliability of household heating systems have received growing attention. Traditional household heating systems mainly consist of heating pipes, radiators, and manual or simple valves, and their operation depends entirely on the user's subjective observation and manual management.

[0003] However, existing technical solutions have the following drawbacks: 1. Lack of real-time monitoring and proactive protection: Traditional heating systems lack any intelligent monitoring capabilities. When pipe leaks occur (such as radiator corrosion or loose joints), users often fail to detect them in time. Only after the leaks cause serious consequences such as water damage to floors, mold growth on walls, or property damage to downstairs neighbors do they take reactive remedial measures, posing a significant risk to their household property.

[0004] 2. Delayed fault detection and difficulty in troubleshooting: For non-leakage faults, such as air blockages in the system causing insufficient heating or pipe blockages leading to poor circulation, users usually only notice the problem after experiencing a significant drop in room temperature. Troubleshooting these faults often relies on professional repair personnel coming to the site, which is time-consuming and labor-intensive, and is a typical "post-incident repair," affecting living comfort.

[0005] 3. Undetectable Equipment Health Status: Critical actuators in the system (such as shut-off valves) are in a standby state for extended periods, making it difficult for users to detect problems such as mechanical wear, corrosion, jamming, or poor sealing. If the valve fails in a critical moment when shutdown is truly needed (such as in the event of a leak), the protective function becomes ineffective, leading to even greater losses.

[0006] 4. Limited management methods and inability to intervene remotely: When users are away on vacation or own multiple properties, they cannot monitor the status of their home heating system in real time, nor can they take remote emergency measures when they receive abnormal alarms, posing a significant safety hazard.

[0007] Therefore, how to overcome the shortcomings of existing technologies and provide a home heating system safety protection solution that can achieve rapid leak response, system health diagnosis, predictive equipment maintenance, and remote intelligent management is a technical problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0008] To overcome the shortcomings of existing technologies, this invention provides an intelligent safety shutdown and health diagnosis device for a household heating system and its control method. The technical solution of this invention is as follows: A smart safety and health diagnostic device for a home heating system includes: an inlet and outlet water pipe assembly cluster, comprising an inlet water pipe and an outlet water pipe, wherein a first electrically operated shut-off valve (V1) is installed on the inlet water pipe and a second electrically operated shut-off valve (V2) is installed on the outlet water pipe; a multi-sensor fusion module, comprising an inlet water pressure sensor (P1), an inlet water temperature sensor (T1), and an inlet water ultrasonic sensor (U1) installed on the inlet water pipe for real-time acquisition of pressure, temperature, and flow data of the inlet water pipe of the heating system; and an outlet water pressure sensor (P2), an outlet water temperature sensor (T2), and an outlet water ultrasonic sensor (U2) installed on the outlet water pipe for real-time acquisition of pressure, temperature, and flow data of the outlet water pipe of the heating system. An edge computing control unit, electrically connected to the multi-sensor fusion module, the first electrically operated shut-off valve (V1), and the second electrically operated shut-off valve (V2), is used to perform localized AI decisions. The edge computing control unit is used to generate a leak detection confidence score based on the pressure and flow data using a multi-evidence fusion algorithm; when the confidence score exceeds a first preset threshold, an early warning mechanism is triggered; when it exceeds a second preset threshold higher than the first preset threshold, a shutdown command is generated within ≤3 seconds and the first electric shut-off valve (V1) and the second electric shut-off valve (V2) are simultaneously controlled to perform a shutdown action.

[0009] The edge computing control unit is also used to perform slow leak detection and fast leak detection steps. The slow leak detection steps are as follows: during a preset static pressure test time, the second electric shut-off valve (V2) and the first electric shut-off valve (V1) are sequentially controlled to close, so that the pipeline system is in a pressure-holding state; during the pressure holding period, the pressure decay rate is monitored by the outlet water pressure sensor (P2); the pressure decay rate is compensated and corrected according to the temperature data collected by the inlet water temperature sensor (T1) and the outlet water temperature sensor (T2) to distinguish between pressure drop caused by cooling and pressure drop caused by leakage; if the corrected pressure decay rate exceeds the micro-leakage threshold, it is identified as a micro-leakage and an alarm message is generated. The edge computing control unit performs the following process for rapid leak detection and automatic shutdown: real-time monitoring of sudden pressure drops; when the inlet pressure sensor (P1) and outlet pressure sensor (P2) simultaneously detect a drop and the first electric shut-off valve (V1) is closed, the pressure difference between the inlet pressure sensor P1 and the outlet pressure sensor (P2) is compared; if the pressure of the inlet pressure sensor P1 increases and the outlet ultrasonic sensor (U2) shows backflow characteristics, an indoor leak is identified and the second electric shut-off valve (V2) is shut off; simultaneously, flow data is collected through the inlet ultrasonic sensor (U1) and the outlet ultrasonic sensor (U2), and the difference between the inlet and outlet flow rates is graded and scored; when the graded score exceeds a preset threshold, a slow leak detection mode is activated for secondary confirmation or an alarm is triggered.

[0010] The edge computing control unit is also used to perform the blockage diagnosis step, specifically: acquiring the temperature data of the inlet water temperature sensor (T1) and the outlet water temperature sensor (T2), as well as the flow data of the inlet water ultrasonic sensor (U1) and the outlet water ultrasonic sensor (U2); calculating the inlet and outlet water temperature difference and comparing it with the pre-established historical reference temperature difference; when the inlet and outlet water temperature difference is higher than the historical reference temperature difference, and the flow rate change amplitude of the inlet water ultrasonic sensor (U1) and the outlet water ultrasonic sensor (U2) is lower than a preset threshold, it is determined that the system is airlocked, and a diagnostic warning is generated.

[0011] The edge computing control unit is also used to perform valve health self-checks and predictive maintenance: During the self-check cycle, it sequentially sends action commands to the first electric shut-off valve (V1) and the second electric shut-off valve (V2), and monitors the motor drive current in real time to generate a current curve including the starting peak current, operating current, and arrival current; it records the action time required for the valve to complete the action from receiving the command; it calculates the valve health score according to the initial health score model, which is: Valve Health Score = Base Score × (Time Coefficient × α + Current Coefficient × β + Sealing Coefficient × γ); where the time coefficient is generated based on the deviation between the action time and the reference time, the current coefficient is generated based on the matching degree between the current curve and the abnormal mode characteristics, the sealing coefficient is generated based on the sealing level of the subsequent pressure holding test, and α, β, and γ are preset weighting coefficients; based on the valve health score and its corresponding score level, it outputs the valve health status and predictive maintenance suggestions.

[0012] The edge computing control unit is also used to perform intelligent learning and adaptive scene mode switching. Specifically, during the preset learning period after installation, it collects and records historical data of the multi-sensor fusion module to establish a personalized benchmark library reflecting pressure range and flow range; it monitors the current ambient temperature in real time and dynamically adjusts the sensitivity threshold of leak detection according to the current ambient temperature; when the flow data of the inlet ultrasonic sensor (U1) and the outlet ultrasonic sensor (U2), as well as the temperature data of the inlet temperature sensor (T1) and the outlet temperature sensor (T2) are detected to increase synchronously during the winter heating season, the sensitivity of leak detection is increased; conversely, the sensitivity is reduced or the monitoring function is put into a dormant state during the non-heating season.

[0013] It also includes a communication module connected to the edge computing control unit, the edge computing The control unit is used to achieve unified management of multiple properties through the communication module: Establish a hierarchical account system with a main account and at least one sub-account. Each account is associated with one or more devices in different locations. The main account has global management permissions, while the sub-accounts have viewing and control permissions within a preset range. A unified dashboard interface is generated through a cloud-based intelligent platform. This interface aggregates and displays the real-time status, key health indicators, and current alarm information of all associated properties in the form of property cards on the same display screen. The key health indicators include leakage confidence score, valve health score, blockage diagnosis results, and total system health score. Receive quick operation instructions from users on any property card on the unified dashboard, generate and send remote control commands to the corresponding remote device, the remote control commands include emergency shutdown, valve self-test, sensitivity adjustment and mode switching; The cloud-based intelligent platform connects to the edge computing control units of each remote property through an encrypted communication link, supports the aggregation, storage, and synchronization of status data from multiple devices, and employs a breakpoint resume mechanism to retransmit historical data when the network recovers.

[0014] It also includes a dual power supply system, which includes: The main power module is used to connect to the external AC power grid and convert it to DC power supply; the backup lithium battery module uses lithium iron phosphate cells to seamlessly switch power supply when the main power is interrupted; the edge computing control unit is also used to monitor the main power status and backup battery power in real time, record the interruption event and control the device to enter low power mode when the main power is interrupted, and generate a low power warning when the battery power is lower than a preset threshold.

[0015] The edge computing control unit is pre-configured with a priority-based conflict resolution strategy. Its multi-layer decision architecture includes: a data acquisition layer, which is used to acquire raw data from the multi-sensor fusion module at a period of 10ms and perform data quality checks and caching; and a feature extraction layer, which is used to perform filtering calibration, difference / rate of change calculation and spectrum analysis on the preprocessed data to extract multi-dimensional features. The model inference layer runs at least several AI models, including a leak detection model, a blockage diagnosis model, and a valve health model, and performs independent inference based on the multi-dimensional features, outputting their respective confidence scores or diagnostic results. The decision arbitration layer receives the inference results from each model, user instructions, and scheduled task requests. The decision is made based on a preset priority matrix, where leakage events have a higher priority than user commands and periodic self-check tasks, in order to determine the final system action to be executed.

[0016] The multi-evidence fusion algorithm is specifically as follows: (1) Obtain scoring data from at least two types of evidence sources, including: a flow difference evidence score X generated based on flow data collected by an inlet ultrasonic sensor (U1) and an outlet ultrasonic sensor (U2), wherein the flow difference evidence score X is graded according to the flow difference ΔQ = |U1 flow - U2 flow| per unit time: when 0.5L / h≤ΔQ<2L / h, the score X is 1-30 points; when 2L / h≤ΔQ<5L / h, the score X is 30-60 points; when ΔQ≥ 5L / h, the score X is 60-100 points; and a pressure drop evidence score Y generated based on pressure data collected by an inlet pressure sensor (P1) and an outlet pressure sensor (P2), wherein the pressure drop evidence score Y is calculated based on the pressure change rate dP / dt within a preset time window and the inlet and outlet pressure difference ΔP = |P1 - P2|. (2) Calculate the leak detection confidence score S using a weighted scoring system: Where ω1 is the weighting coefficient for flow difference evidence, and ω2 is the weighting coefficient for pressure evidence, and ; (3) Perform multi-level decision-making based on the confidence score S: when S < 60 points, it is determined to be a normal state and routine monitoring continues; when 60 points ≤ S < 80 points, it is determined to be a suspected leak state, triggering the early warning mechanism and controlling the valve to enter the slow leak detection mode for verification; when S ≥ 80 points, it is determined to be a confirmed leak state, generating a shut-off command within ≤ 3 seconds and simultaneously controlling the first electric shut-off valve (V1) and the second electric shut-off valve (V2) to perform shut-off actions.

[0017] A control method based on the intelligent safety and health diagnostic device for the home heating system includes the following steps: Step S1: Real-time acquisition of multi-source data: Through the multi-sensor fusion module, the inlet water pressure (P1), outlet water pressure (P2), inlet water temperature (T1), outlet water temperature (T2), inlet water flow rate, and outlet water flow rate of the heating system are collected in real time at a preset sampling period, and the motor drive current and power status of the first electric shut-off valve (V1) and the second electric shut-off valve V2 are monitored simultaneously. Step S2, Edge Feature Extraction and Preprocessing: The edge computing control unit processes the acquired features... The raw data is filtered, calibrated, and compensated; pressure change rate, inlet and outlet flow rate difference, and temperature gradient characteristics are extracted; and the flow data collected by the ultrasonic sensor is subjected to spectrum analysis to extract the energy characteristics of the preset frequency band. Step S3, Multi-model Parallel Inference: Input the extracted feature data into the preset... Parallel inference in multiple models within an AI coprocessor, said multiple models including at least: A leak detection model generates a leak detection confidence score based on a multi-evidence fusion algorithm. A blockage diagnosis model identifies pipe siltation or air blockage based on flow trends and flow characteristics; A valve health model generates a valve health score based on current curve analysis and action time monitoring. Step S4: Priority-based decision arbitration: The decision arbitration layer receives the inference results of each model and makes a decision based on the preset priority matrix; wherein, when the leakage detection confidence score exceeds the first preset threshold, an early warning instruction is generated; when it exceeds the second preset threshold which is higher than the first preset threshold, it is decided as the highest priority event and an emergency shutdown instruction is generated immediately. Step S5, Precise Control of the Actuator: Execute the corresponding action according to the arbitration instruction. If it is an emergency shutdown instruction, the first electric shut-off valve (V1) and the second electric shut-off valve (V2) will be controlled to shut off within ≤3 seconds, and a local audible and visual alarm and a remote push notification will be triggered. If it is a warning or diagnostic instruction, the corresponding health status, warning information or maintenance suggestions will be output through the local display interface and the remote client. Step S6, Data Caching and Synchronization: All running data and event logs are encrypted and stored in local storage media; when the network is connected, the data is synchronized to the cloud server through the communication module. The synchronization strategy adopts the mechanism of resuming interrupted transmission and prioritizing the transmission of important events.

[0018] The advantages of this invention are: (1) The AI ​​coprocessor built into the edge computing control unit generates a leak detection confidence score based on a multi-evidence fusion algorithm of pressure, flow rate, and flow data. When the score exceeds a second preset threshold, the first and second electric shut-off valves are simultaneously controlled to perform shut-off actions within ≤3 seconds. Compared with the prior art, the present invention realizes localized decision-making throughout the entire process from leak identification to physical isolation, without relying on cloud communication, significantly shortening the emergency response time and effectively avoiding property losses caused by response delays.

[0019] (2) A multi-evidence fusion algorithm is adopted to weight and fuse the ultrasonic flow difference evidence score and the pressure change evidence score to form a comprehensive confidence score. First and second preset thresholds are set to achieve graded decision-making: monitoring is maintained below 60 points, an early warning is triggered and slow leak detection is entered for review between 60 and 80 points, and emergency shutdown is executed above 80 points. Compared with the existing technology that judges based on a single sensor threshold, this invention significantly reduces the false alarm rate and false negative rate through cross-validation of multi-source data, and realizes accurate identification and graded handling of leak events.

[0020] (3) When performing slow leak detection, the pressure decay rate is compensated and corrected by using temperature data collected by the inlet and outlet water temperature sensors, which can accurately distinguish between pressure drop caused by medium cooling and pressure drop caused by actual leakage. Compared with the pressure holding test method that does not consider the effect of temperature, the present invention improves the accuracy of micro-leak detection, and can detect hidden dangers and generate alarm information in the early stage of leakage.

[0021] (4) By acquiring the flow rate data and flow rate spectrum of the inlet and outlet ultrasonic sensors during the same historical period and the current period, the flow rate change trend and high-frequency bubble sound characteristics related to air blockage are extracted for analysis and judgment. Compared with the existing technology that relies on manual experience for troubleshooting, the present invention realizes the automated identification and early warning of air blockage or siltation blockage, reducing the time and maintenance costs for users to troubleshoot faults.

[0022] (5) During the self-inspection cycle, an action command is sent to the electric shut-off valve, the motor drive current is monitored in real time and the action time is recorded, the valve health score is calculated according to the health rating model, and the health status and maintenance suggestions are output. Compared with the prior art that only repairs after a failure, the present invention realizes the status monitoring and life prediction of key actuators, ensuring the reliable operation of the valve in emergency situations and improving the overall reliability of the system.

[0023] (6) During the preset learning period after installation, historical data is collected to establish a personalized benchmark library, and the detection sensitivity is dynamically adjusted according to the ambient temperature and heating season: during the winter heating season, the sensitivity is automatically increased when monitoring flow and temperature data, and the sensitivity is reduced or the monitoring function is put into a dormant state during the non-heating season. Compared with the existing technology with fixed thresholds, the present invention can adapt to different user habits and environmental changes, and reduce false alarms and power consumption while ensuring monitoring effect.

[0024] (7) A master account and sub-account system is established through the communication module. Each account can be associated with devices in multiple remote properties. Real-time status, health indicators and alarm information are aggregated and displayed on the same display interface in the form of property cards, and remote control commands are supported. Compared with the existing technology of independent management of single devices, the present invention solves the pain point of users with multiple properties who cannot centrally monitor, and improves management convenience.

[0025] (8) A dual-power system is constructed using a main power module and a lithium iron phosphate backup battery module, which seamlessly switches power supply when the main power supply is interrupted and supports low-power mode operation; at the same time, it has 8GB of built-in local storage to cache data when the network is interrupted and resume transmission after the interruption is restored. Compared with the existing technology that relies on a single power supply and real-time online operation, the present invention can still maintain key functions under extreme conditions such as power outages and network outages, which significantly improves the environmental adaptability and operational reliability of the system.

[0026] (9) A priority-based conflict resolution strategy is preset within the edge computing control unit, constructing a multi-layer architecture consisting of a data acquisition layer, a feature extraction layer, a model inference layer, and a decision arbitration layer. Leakage events have a higher priority than user instructions and periodic self-check tasks. Compared to existing technologies that prioritize tasks equally, this invention ensures that the highest security level events receive priority processing under any circumstances, avoiding decision conflicts and execution delays during multi-task concurrency. Attached Figure Description

[0027] Figure 1 This is a schematic diagram of the main structure of the present invention.

[0028] Figure 2 yes Figure 1 A schematic diagram of the structure of the inlet and outlet water pipeline component cluster.

[0029] Figure 3 This is a flowchart of the control method of the present invention. Detailed Implementation

[0030] The present invention will be further described below with reference to specific embodiments, and the advantages and features of the present invention will become clearer as a result. However, these embodiments are merely exemplary and do not constitute any limitation on the scope of the present invention. Those skilled in the art should understand that modifications or substitutions can be made to the details and form of the technical solutions of the present invention without departing from the spirit and scope of the present invention, but all such modifications and substitutions fall within the protection scope of the present invention.

[0031] See Figures 1 to 3 This invention relates to an intelligent safety and health diagnostic device for a household heating system, comprising: an inlet and outlet water pipe assembly cluster 1, including an inlet water pipe and an outlet water pipe, wherein a first electric shut-off valve V1 is installed on the inlet water pipe and a second electric shut-off valve V2 is installed on the outlet water pipe; The multi-sensor fusion module 2 includes an inlet water pressure sensor P1, an inlet water temperature sensor T1, and an inlet water ultrasonic sensor U1 installed on the inlet water pipe, for real-time acquisition of pressure, temperature, and flow data of the inlet water pipe of the heating system; and an outlet water pressure sensor P2, an outlet water temperature sensor T2, and an outlet water ultrasonic sensor U2 installed on the outlet water pipe, for real-time acquisition of pressure, temperature, and flow data of the outlet water pipe of the heating system. The edge computing control unit 3 is electrically connected to the multi-sensor fusion module, the first electric shut-off valve V1, and the second electric shut-off valve V2, and is used to perform localized AI decision-making. The edge computing control unit is used to: generate a leak detection confidence score based on the pressure and flow data through a multi-evidence fusion algorithm; trigger an early warning mechanism when the confidence score exceeds a first preset threshold; and generate a shut-off command within ≤3 seconds and simultaneously control the first electric shut-off valve V1 and the second electric shut-off valve V2 to perform shut-off actions when the score exceeds a second preset threshold higher than the first preset threshold.

[0032] By symmetrically arranging pressure sensors, temperature sensors, and ultrasonic sensors in the inlet and outlet water pipes, the pressure, temperature, and flow rate at the inlet and outlet of the heating system are simultaneously collected, providing a complete data foundation for accurate diagnosis.

[0033] The edge computing control unit is electrically connected to all sensors and actuators, forming a complete "perception-computation-execution" closed loop. All AI inference and decision-making are completed locally, without relying on the cloud or external networks, ensuring rapid response and system reliability in emergency situations.

[0034] Electric shut-off valves are installed on both the inlet and outlet water pipes, forming a dual isolation mechanism. When a leak is detected and confirmed, both valves shut off simultaneously, ensuring that the water supply is completely cut off regardless of the location of the leak point inside the house, providing higher safety redundancy than traditional solutions.

[0035] The edge computing control unit performs weighted scoring based on two independent evidence sources: pressure and flow rate. By setting first and second preset thresholds, it achieves tiered early warning and tiered shutdown. This multi-evidence cross-verification mechanism effectively avoids false shutdowns caused by false alarms from a single sensor, while ensuring rapid response to actual leaks.

[0036] The integrated device consists of an inlet and outlet water pipe assembly cluster, a multi-sensor fusion module, and an edge computing control unit. During installation, it only needs to be connected in series to the inlet and outlet water pipes of the user's heating system. No complicated wiring or system modification is required, making it easy to promote and apply in new residential buildings and renovation of old communities.

[0037] The edge computing control unit is also used to perform slow leak detection and fast leak detection steps. The slow leak detection steps are as follows: during a preset static pressure test time, the second electric shut-off valve V2 and the first electric shut-off valve V1 are sequentially controlled to close, so that the pipeline system is in a pressure-holding state; during the pressure holding period, the pressure decay rate is monitored by the outlet water pressure sensor P2; the pressure decay rate is compensated and corrected according to the temperature data collected by the inlet water temperature sensor T1 and the outlet water temperature sensor T2 to distinguish between pressure drop caused by cooling and pressure drop caused by leakage; if the corrected pressure decay rate exceeds the micro-leakage threshold, it is identified as a micro-leakage and an alarm message is generated. A user's home has a very minor leak at the radiator joint, with a leakage rate of approximately 0.08 L / hour (equivalent to filling a 500 ml bottle of mineral water every 10 hours). The indoor temperature drops from 20°C during the day to 10°C at night during winter.

[0038] The drawbacks of traditional detection methods (without temperature compensation): During the overnight pressure test, the pressure sensor detected a pressure drop of 0.15 bar.

[0039] Because the cause of the pressure drop cannot be distinguished, the system may misjudge it as a leak and trigger an alarm; or, to avoid false alarms, the pressure drop threshold may be set too high, causing even a tiny leak of 0.08 L / hour to be ignored. The result is either frequent false alarms disturbing residents, or tiny leaks going undetected for a long time, eventually leading to corrosion and expansion at the joints, causing a flooding accident.

[0040] The working process and advantages of this invention for slow leakage detection: 1. Temperature Compensation Correction: Static pressure testing begins at 0:00 AM (time can be set by the user). Valves V2 and V1 close sequentially, and the pipeline enters a pressure-holding state. Inlet water temperature sensor T1 and outlet water temperature sensor T2 monitor the water temperature in real time as it drops from 20℃ to 15℃. The edge computing control unit uses the built-in temperature-pressure coupling physical model to calculate: a 5℃ drop in water temperature will cause a natural pressure drop of approximately 0.12 bar (thermal expansion and contraction effect).

[0041] The system subtracts the 0.12 bar caused by temperature from the measured 0.15 bar pressure decay, and the corrected actual leakage pressure decay is only 0.03 bar.

[0042] 2. Reliable identification of minute leaks The corrected pressure decay rate of 0.03 bar exceeded the preset micro-leakage threshold (e.g., 0.02 bar / 30 min). The system accurately determined that there was a micro-leakage of 0.08 L / hour and generated an alarm message to push to the user.

[0043] It distinguishes between "pressure drop due to cooling" and "pressure drop due to leakage," avoiding frequent false alarms caused by room temperature changes. It can identify extremely small leaks (≥0.05L / hour level) that are undetectable by traditional methods, alerting users to repair before the leak causes actual damage, truly achieving early detection and early handling of potential hazards. Quantitative compensation correction based on a physical model ensures that the detection results are unaffected by ambient temperature fluctuations, significantly improving detection accuracy and reliability.

[0044] The edge computing control unit executes a process package for rapid leak detection and automatic shutdown. Includes: Real-time monitoring of sudden pressure drops. When both inlet and outlet pressure sensors P1 and P2 simultaneously detect a drop and the first electric shut-off valve V1 is closed, the pressure difference between inlet and outlet pressure sensors P1 and P2 is compared. If the pressure of inlet pressure sensor P1 increases and the outlet ultrasonic sensor U2 shows backflow characteristics, an indoor leak is identified, and the second electric shut-off valve V2 is shut off. At the same time, flow data is collected through inlet and outlet ultrasonic sensors U1 and U2, and the difference between inlet and outlet flow is graded and scored. When the graded score exceeds a preset threshold, a slow leak detection mode is activated for secondary confirmation or an alarm is triggered.

[0045] This process does not rely solely on a single pressure signal to determine a leak, but rather constructs a dual verification mechanism of "pressure analysis and flow characteristics": By comparing the pressure changes of inlet pressure sensor P1 and outlet pressure sensor P2 under specific operating conditions (inlet pressure sensor P1 increases, outlet pressure sensor P2 decreases), the difference between "indoor leakage" and "external network pressure fluctuation" can be accurately distinguished.

[0046] The backflow characteristics were captured by the ultrasonic sensor U2 at the water outlet, and the abnormal water flow direction was verified from the sound level.

[0047] It effectively avoids misjudgments caused by fluctuations in municipal pipeline pressure or interference from neighbors' water usage, ensuring that only genuine indoor leaks will trigger the shutdown.

[0048] The process involves grading and scoring the difference between the influent and effluent (U1-U2), and then applying differentiated strategies based on the scores: When the score is low (e.g., 30-60 points): Do not shut down directly, but start the slow leak detection mode for secondary confirmation, giving the system another chance to verify.

[0049] For high scores (e.g., 60 points or above): trigger an alarm or shut down the system directly.

[0050] This "verify first, then act" mechanism ensures rapid response while providing a buffer for critical states, significantly reducing the risk of accidental shutdown due to momentary anomalies or sensor jitter.

[0051] The prerequisite is that "when the first electrically operated shut-off valve V1 is closed". Under this condition, compare the pressure difference between P1 and P2: If P1 increases, it indicates that the external water source is still squeezing into the house, but the pressure inside the house is decreasing (P2 decreases), and U2 shows backflow, which conclusively proves that the leak point is located inside the house (after V1 and before V2). By controlling the valve status to create specific test conditions, the pressure analysis has a clear causal relationship, the reasoning logic is rigorous, and the conclusion is reliable.

[0052] Simultaneously, four sensors, P1, P2, U1, and U2, were activated to work together: Pressure sensors are responsible for capturing transient pressure changes (response time in seconds); ultrasonic sensors are responsible for flow balance analysis and flow characteristic identification.

[0053] For example, if the connection between the radiator and the pipe in a user's bathroom suddenly cracks, causing a medium-sized leak, the leakage rate is approximately 4L / hour.

[0054] The actual operation process of the rapid leak detection procedure of this invention: Step 1: Pressure Drop Detection (Trigger) The inlet pressure sensor P1 and the outlet pressure sensor P2 simultaneously detected a rapid drop in pressure (e.g., a drop of 0.3 bar within 3 seconds).

[0055] The system detects a "sudden pressure drop event" and immediately initiates a rapid leak detection process. Step 2: Operating condition settings and pressure comparison (location). The system controls the first electric shut-off valve V1 to close, cutting off the external water supply. At this time, comparing the pressures of P1 and P2: P1 remains at a higher value due to the external network pressure, while P2 continues to decrease due to indoor leakage, resulting in a significant pressure difference.

[0056] The inlet pressure sensor P1 showed a pressure increase (relative to the moment of shutdown), while the outlet ultrasonic sensor U2 detected the echo characteristics of backflow (because the leak point is inside, water is flowing back from the radiator). Based on this, it was accurately determined that the leak point was located indoors (after V1 and before V2), rather than an external network problem.

[0057] Step 3: Traffic difference tiering and scoring (verification) Meanwhile, the inlet ultrasonic sensor U1 and the outlet ultrasonic sensor U2 continuously collect flow data. The calculated flow rate ΔQ = |U1 flow rate - U2 flow rate| = 4 L / h (within the range of 2-5 L / h).

[0058] According to preset rules, a flow difference evidence score X = 45 points (within the range of 30-60 points) is generated. When the flow data of the outlet ultrasonic sensor U2 and the inlet ultrasonic sensor U1 are found to be inconsistent, the system automatically enters the pressure holding detection mode. If there is no significant pressure change during the pressure holding period, it is determined to be a momentary sensor deviation or water flow disturbance, and the system automatically calibrates the flow matching degree of the inlet ultrasonic sensor U1 and the outlet ultrasonic sensor U2 to a unified flow benchmark.

[0059] Step 4: Intelligent Decision Making System assessment: Leakage location is clear (indoor leakage) + flow difference score of 45 (reaching medium confidence level).

[0060] According to the grading strategy: if the score is 30-60 points and the location is clear, you can choose to directly trigger an alarm or enter the slow leak detection review according to the user's preset.

[0061] Assuming the user sets the system to "directly shut off when the score is ≥40 points and the location is clear", the system will simultaneously control the first electric shut-off valve V1 and the second electric shut-off valve V2 to perform shut-off actions within ≤3 seconds, and push alarm information to the user's mobile phone and control panel.

[0062] The user received an alarm within minutes and returned home to find that the radiator was cracked, but the leak had been stopped in time, only wetting a local area of ​​the floor and avoiding major damage from flooding the entire house.

[0063] This rapid leak detection process, through its progressive design of "pressure drop triggering, valve condition creation, differential pressure comparison positioning, and flow rate classification verification," enables the rapid detection, accurate location, verification, and classified handling of indoor leaks. It achieves the best balance between response speed and false alarm prevention, providing an efficient and reliable safety barrier for household property.

[0064] The purpose of the temperature-pressure coupled physical model is to eliminate the interference of temperature changes on the measurement of pressure decay rate and to accurately identify minute leaks.

[0065] In a closed heating piping system, pressure changes can be caused by two reasons: 1. Actual leakage: The loss of medium leads to a drop in pressure; 2. Temperature change: The medium cools down, causing its volume to shrink and the pressure to drop naturally (thermal expansion and contraction effect).

[0066] This temperature-pressure coupled physical model removes the temperature-related components from the measured pressure decay, extracting the true pressure decay component caused only by leakage.

[0067] For water as the medium, within the commonly used heating temperature range (10℃-80℃), a linearized approximation model can be used: ΔP thermal =K·ΔT; where: ΔP thermal : The pressure change caused by temperature change; ΔT: The temperature change; K: The temperature-pressure coupling coefficient, which is related to fluid volume, expansion coefficient, and system stiffness.

[0068] In practical applications, the model determines the system-specific K value through a combination of experimental calibration and theoretical calculation, and stores it in the edge computing control unit.

[0069] In a slow leak detection process, the specific steps are as follows: Step 1: Data Collection The second electric shut-off valve V2 and the first electric shut-off valve V1 are closed in sequence, and the system enters the pressure holding state. Record the initial pressure P0 and initial temperature T0 (take the average of T1 and T2); During the pressure holding period (e.g., 30 minutes), continuously monitor the pressure Pt and temperature Tt.

[0070] Step 2: Temperature Compensation Calculation Calculate the temperature change: ΔT = T t -T0; Calculate the pressure change caused by temperature: ΔP thermal =K·ΔT; Calculate the actual leakage pressure decay after compensation: ΔP leak =(P0-P t )-ΔP thermal ; Step 3: Leakage Assessment Calculate the pressure decay rate: R = ΔP leak / t; If R exceeds the preset micro-leakage threshold (e.g., 0.02 bar / 30 min), a micro-leakage is determined to exist, and an alarm message is generated.

[0071] For example, during a winter night, when the indoor temperature drops from 20℃ to 10℃, the water temperature drops by 5℃ accordingly. The measured pressure drop is 0.15 bar.

[0072] Model application process: 1. Temperature acquisition: Water temperature dropped by ΔT = -5℃ at T1 and T2. 2. Model Calculation: Based on the pre-embedded temperature-pressure coupling coefficient K = 0.024 bar / ℃ (obtained from system calibration), calculate: ΔP thermal =0.024 × 5 = 0.12 bar; 3. Compensation correction: ΔP leak =0.15-0.12=0.03bar; 4. Judgment and decision: A pressure decrease of 0.03 bar corresponds to a leakage of 0.08 L / hour, which exceeds the micro-leakage threshold. The system determines that there is a leak and issues an alarm.

[0073] Without this model: the system will misjudge all 0.15 bar as leakage, triggering a false alarm; or to avoid false alarms, the threshold will be increased, causing the real leakage of 0.03 bar to be ignored.

[0074] The edge computing control unit is also used to perform congestion diagnosis steps, specifically: The inlet ultrasonic sensor U1 and outlet ultrasonic sensor U2 were obtained in historical data. The system collects flow data and flow spectrum for the current period; extracts the trend of flow data changes and the high-frequency bubble sound features related to air blockage in the flow spectrum; analyzes the flow trend and the flow spectrum features, and if a continuous decrease in flow accompanied by high-frequency bubble sound features is identified, it is determined that there is air blockage or stagnation blockage, and a diagnostic warning is generated.

[0075] This step does not rely solely on a single indicator to determine congestion, but rather combines traffic trend analysis with traffic spectrum characteristics: Flow rate dimension: By analyzing the flow rate change trends of influent ultrasonic sensor U1 and effluent ultrasonic sensor U2 in the same historical period and the current period, it can be determined whether there is a continuous decline. Flow difference dimension: By extracting high-frequency bubble sound features related to air blockage in the flow spectrum, the blockage type is verified from the sound level.

[0076] Advantages: Flow trends reflect macroscopic physical changes, while flow characteristics reveal microscopic mechanisms. The two corroborate each other, effectively avoiding misjudgment based on a single indicator (such as misjudging that the valve is not fully open based solely on a decrease in flow).

[0077] The system stores and retrieves flow and volume data from the same period last year (such as the beginning of the heating season) as a baseline; it compares current data with historical data for the same period, rather than just with last week or yesterday. This effectively eliminates interference from normal operating conditions such as seasonal changes and heating water temperature adjustments on diagnostic results, ensuring that only genuine abnormal trends trigger warnings.

[0078] Extracting "high-frequency bubble acoustic features associated with air blockage": When air blockage occurs, the bursting of bubbles in the pipe will generate high-frequency flow at a specific frequency (usually in the range of 1kHz-5kHz); while siltation blockage (such as sediment deposition) will not produce such flow characteristics.

[0079] By analyzing flow characteristics, the system can accurately distinguish between air blockage and sludge blockage, providing users with differentiated treatment suggestions (air blockage requires venting, sludge requires cleaning), thus improving the precision of diagnosis.

[0080] The process uses a trend of "continuously decreasing traffic" for judgment, rather than relying solely on a single point exceeding the threshold: The system continuously monitors the long-term trend of traffic flow. Once a continuous decline is detected accompanied by characteristic traffic flow patterns, an early warning is immediately generated. This trend warning mechanism can detect early signs of congestion before it worsens (such as complete blockage leading to insufficient heating), allowing users time for repairs.

[0081] This invention automatically collects and analyzes data through inlet ultrasonic sensor U1 and outlet ultrasonic sensor U2 to generate diagnostic warnings, achieving standardization, automation, and intelligence in blockage diagnosis. This reduces the cost and time of manual on-site inspections and improves the user experience.

[0082] A user's radiators gradually developed a "hot top, cold bottom" phenomenon a month after the start of the heating season, with the room temperature 1-2℃ lower than the same period in previous years. The user initially thought it was a normal phenomenon and did not pay much attention to it.

[0083] The actual operation process of the blockage diagnosis procedure of this invention: Step 1: Data Acquisition and Trend Analysis The system retrieved the flow data of the inlet ultrasonic sensor U1 and the outlet ultrasonic sensor U2 in November of last year (the same period in history) as a benchmark: the average flow rate at that time was 800L / h.

[0084] Monitoring the current November traffic data reveals that over the past two weeks, the traffic volume has gradually decreased from 780L / h to 550L / h, showing a continuous downward trend.

[0085] Step 2: Flow spectrum feature extraction While monitoring the flow rate, the inlet ultrasonic sensor U1 and the outlet ultrasonic sensor U2 continuously collect flow data within the pipeline.

[0086] The system performed a spectrum analysis on the flow data and found intermittent high-energy pulse signals in the 1kHz-3kHz frequency band. These signals are the high-frequency sound characteristics of bubbles bursting in the pipe.

[0087] Step 3: Comprehensive Analysis and Judgment The edge computing control unit performs a correlation analysis between two pieces of evidence: "a continuous 30% decrease in traffic" and "high-frequency bubble sound characteristics." The decreasing flow rate indicates increased circulation resistance; the high-frequency bubbling sound indicates the presence of gas in the pipeline; the combination of these two factors rules out the possibility of a valve not being fully open or a pump malfunction, precisely pointing to an air blockage.

[0088] Step 4: Generate diagnostic alerts The system determines that there is "air blockage or stagnation" and generates a diagnostic warning message which is then pushed to the user's mobile app.

[0089] Example of a warning message: "Air blockage has been detected in the heating system. The flow rate is 30% lower than the same period last year, and bubbling sounds have been detected. We suggest you try bleeding the radiators or contact a repair technician for assistance." After receiving the warning, users should follow the instructions to bleed the air from their home's radiators (open the air vent valve to release the trapped air).

[0090] After the venting was completed, the system monitored that the flow rate had risen to 780L / h, the high-frequency bubbling sound disappeared, the radiators resumed even heat dissipation, and the room temperature rose.

[0091] Result: Users intervened proactively before the problem worsened, avoiding energy waste and decreased comfort caused by prolonged low-temperature operation.

[0092] The edge computing control unit is also used to perform valve health self-checks and predictive maintenance: During the self-test cycle, a signal is sent to either the first electrically operated shut-off valve (V1) or the second electrically operated shut-off valve (V2). The system generates an action command and monitors the motor drive current in real time, producing a current curve that includes the peak starting current, operating current, and arrival current. It records the action time required for the valve to complete the action from receiving the command. A valve health score is calculated based on a pre-embedded health scoring model: Valve Health Score = Base Score × (Time Coefficient × α + Current Coefficient × β + Sealing Coefficient × γ). The time coefficient is generated based on the deviation between the action time and the baseline time; the current coefficient is generated based on the matching degree between the current curve and the abnormal mode characteristics; the sealing coefficient is generated based on the sealing level of the subsequent pressure holding test; and α, β, and γ are preset weighting coefficients. Based on the valve health score and its corresponding rating level, the system outputs the valve health status and predictive maintenance suggestions.

[0093] This step does not rely solely on a single indicator to determine the valve's condition, but rather constructs a three-dimensional evaluation system combining current curves, action time, and sealing level. Current dimension: By analyzing the peak starting current, operating current, and final current, the mechanical resistance, motor status, and sealing contact pressure of the valve are fully reflected. Time dimension: By monitoring the motion time, the wear and jamming degree of the transmission mechanism is reflected; Sealing dimension: The sealing level is evaluated through pressure holding test to verify the actual sealing effect after the valve is closed.

[0094] The three-dimensional parameters complement each other and cross-validate each other, quantifying the abstract valve health status into specific health scores, making the status assessment more comprehensive, objective and accurate.

[0095] The steps employ a formulaic health scoring model: Valve health score = base score × (time coefficient × α + current coefficient × β + sealing coefficient × γ); where α, β, and γ are preset weighting coefficients (which can be dynamically adjusted according to valve type and usage environment).

[0096] Advantages: By integrating multi-dimensional features into a quantifiable health score through a mathematical model, and corresponding to a rating level (such as 90-100 points for excellent, 75-89 points for good, 60-74 points for caution, and <60 points for warning), the health status of valves is standardized and hierarchically managed, making it easy for users to understand intuitively.

[0097] Traditional valve maintenance follows a "break down, fix it" approach—the problem isn't discovered until the valve is jammed and unable to close, often resulting in damage. This step involves continuous monitoring and trend analysis: A gradual increase in operating time is observed (indicating mechanical jamming); an abnormal current curve is observed (indicating wear of the motor or transmission mechanism); a decrease in sealing grade is observed (indicating aging of the seals).

[0098] Early warnings and maintenance recommendations are issued before the valve completely fails, allowing users to schedule maintenance at a convenient time and avoid safety accidents caused by valve failure during emergency shutdown.

[0099] The current coefficient in the process not only reflects the magnitude of the current, but more importantly, it is generated based on the "matching degree between the current curve and the characteristics of the abnormal mode": Stuck mode: Abnormally high starting peak current, fluctuating operating current; Locked rotor mode: Current remains high and fails to reach the target position; Wear mode: Operating current gradually decreases, and operating time is prolonged; Poor sealing: Abnormal target current (too high or too low), sealing test fails. By matching the shape of the current curve with typical fault modes, we can not only determine whether the system is "healthy" or not, but also initially locate "where the problem is," providing directional guidance for maintenance.

[0100] The steps mention that the "self-inspection cycle" can be set flexibly (e.g., weekly self-inspection during the heating season, and every two months during the non-heating season).

[0101] Advantages: It ensures continuous monitoring of valve status while avoiding unnecessary mechanical wear caused by excessively frequent operations, achieving the best balance between monitoring effectiveness and valve life.

[0102] Data from each self-test (current curve, operating time, sealing level) is stored, forming a health record for the entire life cycle of the valve. Based on historical data, a degradation trend model of the valve can be established to predict its remaining service life, providing a scientific basis for users to formulate replacement plans.

[0103] A user's home heating system has been running for 5 years, and the first electric shut-off valve V1 has not been activated for a long time. Due to the deposition of impurities in the water and the natural aging of the sealing ring, the valve has begun to show slight sticking and poor sealing, but it has not yet affected normal use.

[0104] The actual operation process of the valve health self-inspection procedure of this invention: Step 1: Self-check cycle triggered The preset self-check cycle is "every Sunday at 1:00 AM during the heating season". At 1:00 AM this Sunday, the system will automatically start the valve health self-check program. (During the heating season, valve self-checks will also be performed during the pressure test; during the non-heating season, valves will self-check monthly.) Step 2: Current Curve Acquisition and Analysis The edge computing control unit sends a "close" command to the first electric shut-off valve V1, while simultaneously monitoring the motor drive current in real time at a high sampling rate.

[0105] The generated current curve is as follows: Historical baseline curve: Start-up peak current 0.8A → Operating current 0.5A → Target current 0.3A → Zero; The measured curves are as follows: peak starting current 1.3A (62.5% higher than the reference) → operating current fluctuates between 0.7-0.9A (abnormal fluctuation) → current at the bottom is 0.4A (higher than the reference) → returns to zero.

[0106] Current coefficient generation: The system matches the measured current curve with the built-in "hysteresis mode" feature library, with a matching degree of 85%, and generates a current coefficient of 0.7 (out of 1.0).

[0107] Step 3: Motion Time Monitoring Record the time from receiving the command to completing the action for V1: the historical benchmark valve closing time is 6.2 seconds; the actual measured valve closing time is 8.5 seconds (37% longer).

[0108] Time coefficient generation: Based on the deviation from the reference time, a time coefficient of 0.6 is generated.

[0109] Step 4: Sealing rating test (followed by pressure testing) After the self-test is completed, the system uses a subsequent slow leak detection procedure (or a specially arranged pressure holding test) to evaluate the sealing performance of V1.

[0110] The pressure holding test revealed that after V1 was closed, the pressure dropped by 0.08 bar within 10 minutes, and the sealing grade was rated as C (the benchmark is A).

[0111] Sealing coefficient generation: Based on the sealing level, a sealing coefficient of 0.5 is generated.

[0112] Step 5: Health Score Calculation The default base score is 100, and the weighting coefficients are α=0.3 (time), β=0.4 (current), and γ=0.3 (sealing).

[0113] Calculation: Valve health score = 100 × (0.6 × 0.3 + 0.7 × 0.4 + 0.5 × 0.3) = 100 × (0.18 + 0.28 + 0.15) = 61 points; Step 6: Output health status and maintenance recommendations Rating level: 61 points is at the "Caution" level (yellow warning).

[0114] Output: "First electric shut-off valve V1 health score 61 points (Caution level). Increased starting current, prolonged action time, and decreased sealing performance were detected, suggesting possible jamming and aging of the sealing ring. It is recommended to contact a professional for cleaning and maintenance or replacement of the sealing ring in the near future to avoid affecting the emergency shut-off function." After receiving the alert, the user contacted a repair technician at a convenient time. The technician disassembled and inspected the valve and found a small amount of scale buildup on the valve core and slight hardening of the sealing ring. Cleaning and maintenance were performed, and the sealing ring was replaced. After maintenance, the system automatically checked: the current curve returned to near-baseline, the action time was 6.5 seconds, the sealing grade returned to A level, and the health score rose to 95 points (excellent).

[0115] By proactively maintaining the valve before it completely fails, users can avoid the catastrophic consequences of the valve being unable to shut off in the event of a future leak.

[0116] The edge computing control unit is also used to perform intelligent learning and adaptive scene mode switching. Specifically, during the preset learning period after installation, it collects and records historical data of the multi-sensor fusion module to establish a personalized benchmark library reflecting pressure range and flow range; it monitors the current ambient temperature in real time and dynamically adjusts the sensitivity threshold of leak detection according to the current ambient temperature; when it detects that the flow data of the inlet ultrasonic sensor U1 and the outlet ultrasonic sensor U2, as well as the temperature data of the inlet temperature sensor T1 and the outlet temperature sensor T2, show a synchronous increase during the winter heating season, it increases the sensitivity of leak detection; conversely, it reduces the sensitivity or puts the monitoring function into a dormant state during the non-heating season.

[0117] This step involves collecting and recording historical data from the multi-sensor fusion module during a pre-set learning period (e.g., 14 days) after installation, establishing a benchmark library reflecting the personalized pressure and flow ranges of each user's household. Different households have significantly different heating systems (e.g., floor level affects pressure, apartment size affects flow, and water usage habits affect fluctuation patterns). Through personalized learning, the system can accurately grasp "what is the normal state for this user," fundamentally reducing false alarm rates.

[0118] The procedure explicitly states that "the current ambient temperature is monitored in real time, and the sensitivity threshold for leak detection is dynamically adjusted based on the current ambient temperature."

[0119] The pressure and flow parameters of a heating system naturally change with ambient temperature (e.g., water pressure naturally decreases at night when temperatures are low). Traditional fixed thresholds are prone to false alarms at night or missed alarms during the day, especially during seasons with large diurnal temperature variations. This invention uses dynamic compensation based on ambient temperature to ensure that the detection threshold always matches the current environmental conditions, guaranteeing accurate detection around the clock.

[0120] The procedure specifically emphasizes that when the flow data of the inlet ultrasonic sensor U1 and the outlet ultrasonic sensor U2, as well as the temperature data of the inlet temperature sensor T1 and the outlet temperature sensor T2, show a synchronous increase during the winter heating season, the sensitivity of leak detection will be automatically increased; conversely, the sensitivity will be reduced or the monitoring function will be put into a dormant state during the non-heating season.

[0121] During the heating season, pipelines are pressurized and flowing with water, which is a high-risk period for leaks. At this time, increasing sensitivity can ensure that there are no blind spots in safety protection. Low power consumption / sleep mode during non-heating season: During the non-heating season, the system is dry or stagnant, and the risk of leakage is extremely low. Reducing sensitivity or entering sleep mode can significantly reduce equipment power consumption, extend sensor life, and avoid false alarms caused by normal phenomena such as thermal expansion and contraction.

[0122] The system automatically determines the start of the heating season by "synchronizing" the flow rate and temperature, without requiring manual settings from the user, truly achieving "unattended operation and intelligent switching".

[0123] The personalized benchmark library of this invention is dynamically updated (and can be continuously fine-tuned in subsequent operation), and the detection threshold always matches the current actual state of the system, adapting to changes throughout the system's entire life cycle and maintaining high accuracy over the long term.

[0124] By putting the monitoring function into sleep mode or reducing its sensitivity during the non-heating season, the standby power consumption of the equipment and the workload of the sensors are significantly reduced. This extends the lifespan of the equipment, reduces unnecessary energy consumption, and aligns with green environmental protection principles. At the same time, the reduced consumption of backup batteries during the non-heating season ensures sufficient power when truly needed (such as in the event of a sudden power outage).

[0125] The entire learning and adaptation process is fully automated, requiring no user setup or operation. The elimination of complex configuration processes significantly enhances the user experience.

[0126] For example, at the beginning of the heating season: 1. Date: November 15 (local unified heating begins).

[0127] 2. Feature recognition: The flow data of the inlet ultrasonic sensor U1 and the outlet ultrasonic sensor U2 are simultaneously improved: the system detects that there is a continuous water flow in the pipe (heating circulation is on), and the flow rate is stable at 500-800L / h (non-water usage period).

[0128] The temperature data from the inlet water temperature sensor T1 and the outlet water temperature sensor T2 are updated simultaneously: the water temperature gradually rises from 15℃ to 45℃ and 50℃ respectively.

[0129] Based on comprehensive assessment, the system detected a simultaneous increase in both flow rate and temperature, indicating that the heating season has begun.

[0130] 3. Adaptive Actions: Leak detection sensitivity is automatically improved: the pressure drop detection threshold is tightened from 0.3 bar / second to 0.2 bar / second; the flow difference detection threshold is tightened from 5 L / h to 2 L / h.

[0131] Monitoring frequency increased: The data acquisition and analysis cycle was shortened from 100ms to 50ms.

[0132] During the heating season, pipelines are more prone to leaks due to factors such as thermal expansion and contraction and water pressure fluctuations. Improved sensitivity enhances the system's ability to detect even minor leaks, thus ensuring safety.

[0133] During the heating season: 1. Scenario: On a certain day in December, the outdoor temperature is -5℃ and the indoor temperature is 20℃ during the day; the outdoor temperature is -15℃ and the indoor temperature is 18℃ at night.

[0134] 2. Dynamic adjustment: The system monitors the ambient temperature sensor in real time (or indirectly senses water temperature changes through inlet water temperature sensor T1 and outlet water temperature sensor T2).

[0135] As the ambient temperature drops at night, the system predicts that the water temperature may decrease (due to thermal expansion and contraction). It automatically widens the pressure decay alarm threshold appropriately (e.g., from 0.2 bar to 0.25 bar) to compensate for the natural pressure reduction caused by the temperature drop.

[0136] This avoids false alarms caused by a natural drop in pressure due to lower nighttime temperatures. Users can sleep soundly without being disturbed by false alarms.

[0137] Non-heating season: 1. Time: April 15th of the following year (end of the heating season).

[0138] 2. Feature recognition: The flow data of the inlet ultrasonic sensor U1 and the outlet ultrasonic sensor U2 decreased simultaneously: the continuous water flow stopped, and there was only occasional momentary water flow.

[0139] The inlet water temperature sensor T1 and the outlet water temperature sensor T2 simultaneously decrease in temperature: the water temperature gradually drops from above 50℃ to normal temperature (15-20℃).

[0140] Based on comprehensive assessment, the system detected a simultaneous decrease in both flow rate and temperature, indicating that the heating season has ended.

[0141] 3. Adaptive Actions: Leak detection sensitivity is automatically reduced: the pressure drop detection threshold is relaxed to 0.5 bar / second; the flow difference detection threshold is relaxed to 10 L / h.

[0142] Some monitoring functions have entered dormancy: the frequency of traffic collection has been reduced (from continuous collection to intermittent collection), and AI inference tasks are running at a lower level.

[0143] The overall power consumption of the device has been reduced from 5W in working mode to 1.5W in standby mode.

[0144] During the non-heating season, the pipes are empty or stagnant, resulting in an extremely low risk of leakage. Reduced sensitivity avoids false alarms caused by normal phenomena such as thermal expansion and contraction in summer and occasional water usage; power consumption is significantly reduced, extending equipment life and saving power for backup batteries, ensuring sufficient power to cope with emergencies during sudden power outages.

[0145] It also includes a communication module connected to the edge computing control unit, the edge computing The control unit is used to achieve unified management of multiple properties through the communication module: Establish a hierarchical account system with a main account and at least one sub-account. Each account is associated with one or more devices in different locations. The main account has global management permissions, while the sub-accounts have viewing and control permissions within a preset range. A unified dashboard interface is generated through a cloud-based intelligent platform. This interface aggregates and displays the real-time status, key health indicators, and current alarm information of all associated properties in the form of property cards on the same display screen. The key health indicators include leakage confidence score, valve health score, blockage diagnosis results, and total system health score. Receive quick operation instructions from users on any property card on the unified dashboard, generate and send remote control commands to the corresponding remote device, the remote control commands include emergency shutdown, valve self-test, sensitivity adjustment and mode switching; The cloud-based intelligent platform connects to the edge computing control units of each remote property through an encrypted communication link, supports the aggregation, storage, and synchronization of status data from multiple devices, and employs a breakpoint resume mechanism to retransmit historical data when the network recovers.

[0146] By establishing an account system with master and sub-accounts, each account can be linked to devices in multiple properties located in different locations. All linked properties are aggregated and displayed on a unified dashboard in the form of property cards, providing real-time status, key health indicators, and current alarm information. Compared to existing technologies that manage individual devices independently, this invention addresses the pain point of users with multiple properties being unable to centrally monitor the heating status of each residence. Users only need to log in with one account to have comprehensive control, significantly improving management convenience.

[0147] By establishing a hierarchical system of master and sub-accounts, access control and management among family members can be achieved. Simultaneously, users can execute quick operation commands on any property card from a unified dashboard, and the system instantly generates and sends remote control commands to the corresponding remote device. Compared to existing technologies lacking hierarchical access control and with complex operation paths, this invention not only ensures account security but also enables rapid remote intervention upon receiving alarms from remote properties, effectively improving emergency response efficiency.

[0148] It also includes a dual power supply system, which includes: The main power module is used to connect to the external AC power grid and convert it to DC power supply; the backup lithium battery module uses lithium iron phosphate cells to seamlessly switch power supply when the main power supply is interrupted. The edge computing control unit is also used to monitor the main power supply status and backup battery power in real time. The system measures the amount of power supplied and records interruption events when the main power supply is interrupted, controls the device to enter a low-power mode, and generates a low-power warning when the battery level is below a preset threshold.

[0149] By cooperating with a backup lithium battery module using lithium iron phosphate cells, the main power module can seamlessly switch to backup battery power in the event of a main power outage. Compared with existing technologies that rely on a single power source, this invention avoids the risk of equipment failure due to sudden power outages, ensuring that the equipment can still monitor, make decisions, and shut down normally in emergencies such as leaks, significantly improving the reliability and safety of the system.

[0150] The edge computing control unit monitors the main power status and backup battery level in real time. When the main power is interrupted, it automatically controls the device to enter a low-power mode to extend battery life, and generates a low-battery warning when the battery level falls below a preset threshold. Compared to existing technologies without power status monitoring, this invention allows users to promptly understand the device's power supply status and take appropriate measures, avoiding the safety hazard of monitoring function failure due to battery depletion.

[0151] The edge computing control unit has a preset priority-based conflict resolution strategy, which has multiple The hierarchical decision-making architecture includes: The data acquisition layer is used to acquire raw data from the multi-sensor fusion module at a period of 10ms. According to the data, data quality checks and caching are performed; The feature extraction layer is used to filter and calibrate the preprocessed data, and to calculate the difference / rate of change. Calculation and spectral analysis to extract multi-dimensional features; The model inference layer runs at least several AI models, including a leak detection model, a blockage diagnosis model, and a valve health model, and performs independent inference based on the multi-dimensional features, outputting their respective confidence scores or diagnostic results. The decision arbitration layer receives the inference results from each model, user instructions, and scheduled task requests. The decision is made based on a preset priority matrix, where leakage events have a higher priority than user commands and periodic self-check tasks, in order to determine the final system action to be executed.

[0152] The multi-evidence fusion algorithm is specifically as follows: (1) Obtain scoring data from at least two types of evidence sources, including: a flow difference evidence score X generated based on flow data collected by inlet ultrasonic sensor U1 and outlet ultrasonic sensor U2, wherein the flow difference evidence score X is graded according to the flow difference ΔQ = |U1 flow - U2 flow| per unit time: when 0.5L / h≤ΔQ<2L / h, the score X is 1-30 points; when 2L / h≤ΔQ<5L / h, the score X is 30-60 points; when ΔQ≥ 5L / h, the score X is 60-100 points; a pressure drop evidence score Y generated based on pressure data collected by inlet pressure sensor P1 and outlet pressure sensor P2, wherein the pressure drop evidence score Y is calculated based on the pressure change rate dP / dt within a preset time window and the inlet and outlet pressure difference ΔP = |P1 - P2|. (2) Calculate the leak detection confidence score S using a weighted scoring system: Where ω1 is the weighting coefficient of the flow difference evidence, ω2 is the weighting coefficient of the pressure evidence, and ω1+ω2 = 1; (3) Perform multi-level decision-making based on the confidence score S: When S < 60 points, it is considered a normal state, and routine monitoring continues; When 60 points ≤ S < 80 points, the situation is judged as a suspected leak, triggering an early warning mechanism and controlling... The valve is put into slow leak detection mode for verification; When S ≥ 80, the leak is confirmed, and a shutdown command is generated within ≤3 seconds. The first electric shut-off valve V1 and the second electric shut-off valve V2 are synchronously controlled to perform shut-off actions.

[0153] This architecture divides the decision-making process into four logically clear and independently responsible levels: Data acquisition layer: Responsible for acquiring raw data, quality checking, and caching to ensure the accuracy and integrity of the data source; Feature extraction layer: responsible for data preprocessing and feature engineering, transforming raw data into high-dimensional features that the model can understand; Model inference layer: Responsible for running multiple AI models in parallel inference and outputting their respective diagnostic results; Decision Arbitration Layer: Responsible for receiving requests from multiple parties (model results, user instructions, scheduled tasks) and making a final decision based on the priority matrix.

[0154] This layered design aligns with the software engineering principles of high cohesion and low coupling. Each layer has clearly defined responsibilities and evolves independently, facilitating system development, debugging, maintenance, and upgrades. Modifying the feature extraction algorithm does not affect the model inference layer, and adding new models does not affect the decision-making and arbitration layer.

[0155] The model inference layer "runs at least several AI models, including a leak detection model, a blockage diagnosis model, and a valve health model," and it uses parallel inference.

[0156] Multiple models run simultaneously without blocking each other, avoiding the cumulative latency caused by serial processing. The leak detection model can complete inference in milliseconds, while the congestion diagnosis model (which requires a longer time window) runs concurrently without affecting each other, ensuring that emergencies can be responded to quickly.

[0157] The data acquisition layer collects raw data at a 10ms cycle and has data quality checks and caching functions. The high sampling rate of 10ms can capture transient events such as sudden pressure drops and current surges, ensuring that key features are not missed; data quality checks can remove outliers and fill in missing values; the caching mechanism ensures that data is not lost when the instantaneous processing pressure is too high, providing a reliable data foundation for upper-level decision-making.

[0158] The feature extraction layer performs filtering calibration, difference / rate of change calculation, and spectrum analysis to extract multi-dimensional features.

[0159] Raw sensor data often contains noise and redundant information. By filtering and denoising, calculating differences (such as the pressure difference between P1 and P2), calculating rates of change (such as dP / dt), and performing ultrasonic spectrum analysis, the raw data is transformed into features with greater physical meaning and discriminative power, significantly improving the accuracy and efficiency of subsequent AI model inference.

[0160] Each model (leakage, blockage, health) is based on the same fundamental features, but performs independent reasoning according to its own algorithm logic, and outputs its own confidence level or diagnostic results.

[0161] Independent reasoning avoids interference between models. The leakage model focuses on pressure drops and flow differences, the blockage model focuses on flow trends and bubble flow, and the health model focuses on current curves. Furthermore, when the results of multiple models corroborate each other (e.g., a high confidence level in leak detection coupled with a low valve health score suggests poor valve sealing may lead to leakage), richer diagnostic information can be provided.

[0162] The decision-making arbitration layer adjudicates based on a pre-defined priority matrix, clearly stipulating that "leakage events have higher priority than user commands and periodic self-check tasks." When multiple events occur simultaneously (such as a user clicking "open valve" on the app while the system detects a leak), the system can make the correct judgment based on the priority matrix: safety comes first. The leakage event is handled first, and immediate shutdown is executed, while user commands are temporarily suspended or rejected. This design ensures that in the most urgent moments, the system will not delay critical protective actions due to executing secondary tasks, fundamentally protecting the safety of family property.

[0163] When a high-priority task preempts a task, a low-priority task can be interrupted, delayed, or suspended until the high-priority task is completed. This avoids system deadlock or resource contention caused by multiple concurrent tasks. For example, if a self-check task (such as valve action) is interrupted by a leak event, the system records the current state and resumes the self-check after the leak is resolved, ensuring that all tasks are eventually executed.

[0164] Scenario Setting: One Sunday afternoon, multiple events occur in a user's home, triggering multiple tasks simultaneously, causing the system to face multi-task concurrency conflicts. Specific events are as follows: Event A (Timed Self-Check Task): According to the preset cycle (every Sunday afternoon at 3:00), the system automatically starts the "Valve Health Self-Check" program, prepares to send an action command to the first electric shut-off valve V1, and collects current curves and monitors action time.

[0165] Event B (Remote User Command): When a user is away from home, they click the "Open Valve" button on their mobile app to remotely open valves V1 and V2 in preparation for heating upon their return.

[0166] Event C (Emergency Leakage Event): At almost the same time, the radiator in the house cracked due to an accidental impact. The inlet water pressure sensor P1 and the outlet water pressure sensor P2 detected a sudden drop in pressure, and the inlet water ultrasonic sensor U1 and the outlet water ultrasonic sensor U2 detected a sharp increase in the flow rate difference.

[0167] The actual operation process of the priority-based conflict resolution strategy of this invention: 1. Data Acquisition Layer (10ms cycle) Pressure data from inlet pressure sensor P1 and outlet pressure sensor P2, flow data from inlet ultrasonic sensor U1 and outlet ultrasonic sensor U2, and current data from first electric shut-off valve V1 and second electric shut-off valve V2 are continuously collected at a period of 10ms.

[0168] Data quality checks were normal; all data was cached and passed to the feature extraction layer in real time.

[0169] 2. Feature Extraction Layer The raw data is filtered and denoised, and the pressure change rate dP / dt is calculated (a sudden drop is found), the inlet and outlet water pressure difference ΔP (rapid increase) and the inlet and outlet water flow difference ΔQ (rapid increase) are calculated.

[0170] Spectral analysis was performed on the ultrasonic data from the inlet ultrasonic sensor U1 and the outlet ultrasonic sensor U2 to extract the energy of the leakage characteristic frequency band (significantly increased). Simultaneously, characteristics such as the static current of V1 were also extracted (but at this time, the first electrically operated shut-off valve V1 had not yet activated).

[0171] 3. Model Inference Layer (Parallel Inference) Leakage detection model: Based on the characteristics of pressure drop (dP / dt exceeds the standard), flow difference (△Q=6L / h), and flow characteristics, a leakage detection confidence score of S=92 points is generated (far exceeding the 80-point shut-off threshold).

[0172] Congestion diagnosis model: Based on recent traffic trends and characteristics, the inference result is "no congestion" (normal).

[0173] Valve health model: Based on the static current of the first electric shut-off valve V1 and the most recent self-test data, the inference result is "good health status" (but this self-test has not yet started, and there is no new data).

[0174] 4. Decision-making arbitration level (conflict resolution) The arbitration panel received three requests simultaneously: (1) Model inference results: The leak detection model outputs "S=92 points, leak confirmed", and requests emergency shutdown; (2) User instruction: "Open valve" request from APP; (3) Scheduled task: "Start valve health self-check" request from system scheduling.

[0175] Arbitration layer query priority matrix: Priority 1 (highest): Confirm leakage event; Priority 2: User emergency shutdown command (but the user's current command is "on", not "off"); Priority 3: User routine command; Priority 4: Periodic self-check task; Result: The highest priority event (leakage) was immediately executed: the arbitration level issued an instruction to the execution agency to simultaneously close the first electric shut-off valve V1 and the second electric shut-off valve V2.

[0176] User command (open valve) rejected / suspended: The arbitration layer determines that there is currently a leak and that executing the "open valve" command would exacerbate the leak. Therefore, the command is rejected and a message is pushed to the user's APP: "Indoor leak detected, emergency shutdown has been performed, and the opening command cannot be executed."

[0177] Periodic self-check tasks are interrupted / delayed: Valve health self-check tasks that are being prepared to start are canceled (or marked as "delayed execution") and will be rescheduled for execution after the leak event is handled and the system returns to normal.

[0178] 5. Precise Actuator Control: The actuator synchronously controls V1 and V2 to complete the shutdown action within ≤3 seconds. This triggers a local audible and visual alarm and pushes leakage alarm and shutdown notifications to the user's APP and property management platform via the communication module.

[0179] 6. Result: Although the user issued the "open valve" command at the same time, the system correctly executed the shutdown action because the leakage event had a higher priority.

[0180] The user received an alarm notification a few minutes later. When they returned home, they found that the leak had been stopped and only the area was wet, thus avoiding major damage from flooding the entire house.

[0181] The self-check task is automatically re-executed after the leak is dealt with and the system is restored (e.g., at night), without affecting the overall maintenance plan.

[0182] The present invention also relates to a control method based on the intelligent safety and health diagnosis device of the aforementioned home heating system, comprising the following steps: Step S1: Real-time acquisition of multi-source data: Using a multi-sensor fusion module, the inlet water pressure P1, outlet water pressure P2, inlet water temperature T1, outlet water temperature T2, inlet water flow rate, outlet water flow rate, and flow difference of the heating system are acquired in real-time at a preset sampling period. Simultaneously, the motor drive current and power status of the first electric shut-off valve V1 and the second electric shut-off valve V2 are monitored. This multi-sensor fusion module acquires inlet / outlet water pressure, temperature, and flow rate data in real-time at a preset sampling period, while simultaneously monitoring the valve motor drive current and power status, achieving synchronous acquisition of multi-dimensional physical quantities of the heating system. Compared to traditional methods that only collect single parameters, this invention provides a comprehensive and synchronous data foundation for subsequent fusion diagnosis, avoiding analytical biases caused by data asynchrony.

[0183] Step S2, Edge Feature Extraction and Preprocessing: The edge computing control unit processes the acquired features... The raw data undergoes filtering, calibration, and compensation processing; pressure change rate, inflow / outflow difference, and temperature gradient features are extracted; and spectral analysis is performed on the flow difference data collected by the ultrasonic sensor to extract energy features of a preset frequency band. Filtering, calibration, and compensation processing are completed locally in the edge computing control unit, extracting pressure change rate, flow difference, and temperature gradient features, and performing spectral analysis on the flow data. Compared to methods that directly upload raw data to the cloud for processing, this invention completes quality optimization and feature engineering at the data source, reducing communication bandwidth pressure and eliminating environmental interference through temperature-pressure coupling model compensation, providing high-quality feature input for model inference.

[0184] Step S3, Multi-model Parallel Inference: Input the extracted feature data into the preset... Parallel inference in multiple models within an AI coprocessor, said multiple models including at least: A leak detection model generates a leak detection confidence score based on a multi-evidence fusion algorithm. A blockage diagnosis model identifies pipe siltation or air blockage based on flow trends and flow characteristics; The valve health model generates a valve health score based on current curve analysis and action time monitoring. Feature data is input into multiple AI models, including a leak detection model, a blockage diagnosis model, and the valve health model, for parallel inference. Compared to traditional methods that use serial processing or multiple tasks sharing a single model, this invention uses a parallel architecture to allow multiple models to run simultaneously without blocking each other. This ensures real-time leak detection (millisecond-level response) while simultaneously completing blockage diagnosis and health assessment, achieving multi-dimensional intelligent diagnosis.

[0185] Step S4: Priority-based decision arbitration: The decision arbitration layer receives the inference from each model. The system then makes a decision based on a preset priority matrix. Specifically, when the leak detection confidence score exceeds a first preset threshold, an early warning command is generated; when it exceeds a second preset threshold (which is higher than the first), it is designated as a highest priority event, and an emergency shutdown command is immediately generated. The decision arbitration layer receives the inference results from each model, user commands, and scheduled task requests, and makes a decision based on the preset priority matrix, determining that the leak event has a higher priority than user commands and self-check tasks. Compared to existing technologies that process all requests equally, this invention ensures that in emergency situations such as leaks, the system can interrupt or suspend secondary tasks and prioritize the shutdown command, fundamentally protecting family property security.

[0186] Step S5, Precise Control of the Execution Agency: Execute corresponding actions according to the instructions following arbitration. If it is an emergency shutdown command, the first electric shut-off valve V1 will be synchronously controlled within ≤3 seconds. The first and second electric shut-off valves V2 perform shut-off actions, triggering local audible and visual alarms and remote push notifications. If it is a warning or diagnostic command, the corresponding health status, warning information, or maintenance suggestions are output through the local display interface and remote client. Based on the arbitration command, the first and second electric shut-off valves are synchronously controlled to perform shut-off actions within ≤3 seconds, triggering local audible and visual alarms and remote push notifications. Compared to traditional devices with slow response or only local alarms, this invention achieves rapid closed-loop control from detection and decision-making to execution. Simultaneously, multi-dimensional alarms ensure that users can be informed of the danger and confirm the handling results immediately, regardless of their location.

[0187] Step S6, Data Caching and Synchronization: Encrypt and store all runtime data and event logs. The data is stored locally on the local storage medium. When the network is connected, the data is synchronized to the cloud server via the communication module. The synchronization strategy employs a mechanism of resuming interrupted downloads and prioritizing important events. All operational data and event logs are encrypted and stored locally on the local storage medium, and synchronized to the cloud using the same mechanism when the network is connected. Compared to devices that rely on real-time online uploads, this invention can still completely preserve data in the event of a network outage, and intelligently synchronizes it after the network is restored, ensuring the historical traceability and integrity of the data, and providing a reliable basis for fault analysis and system optimization.

[0188] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. An intelligent safety and health diagnostic device for a home heating system, characterized in that, include: The inlet and outlet water pipe assembly cluster includes an inlet water pipe and an outlet water pipe, wherein the inlet water pipe... A first electric shut-off valve (V1) is installed on the road, and a second electric shut-off valve (V2) is installed on the water outlet pipe. The multi-sensor fusion module includes an inlet water pressure sensor (P1), an inlet water temperature sensor (T1), and an inlet water ultrasonic sensor (U1) installed on the inlet water pipe, used to collect pressure, temperature, and flow data of the inlet water pipe of the heating system in real time; and an outlet water pressure sensor (P2), an outlet water temperature sensor (T2), and an outlet water ultrasonic sensor (U2) installed on the outlet water pipe, used to collect pressure, temperature, and flow data of the outlet water pipe of the heating system in real time; An edge computing control unit, electrically connected to the multi-sensor fusion module, the first electrically operated shut-off valve (V1), and the second electrically operated shut-off valve (V2), is used to perform localized AI decisions. The edge computing control unit is used to generate a leak detection confidence score based on the pressure and flow data using a multi-evidence fusion algorithm; when the confidence score exceeds a first preset threshold, an early warning mechanism is triggered. When the threshold exceeds the second preset threshold which is higher than the first preset threshold, a shutdown command is generated within ≤3 seconds and the first electric shut-off valve (V1) and the second electric shut-off valve (V2) are controlled to perform the shutdown action simultaneously.

2. The intelligent safety and health diagnostic device for a home heating system according to claim 1, characterized in that, The edge computing control unit is also used to perform slow leak detection and fast leak detection. The steps for leak detection are as follows: During the preset static pressure test time, the second electric shut-off valve (V2) and the first electric shut-off valve (V1) are closed sequentially to keep the pipeline system in a pressure-holding state; during the pressure holding period, the pressure decay rate is monitored by the outlet pressure sensor (P2); The pressure attenuation rate is compensated and corrected based on the temperature data collected by the inlet water temperature sensor (T1) and the outlet water temperature sensor (T2) to distinguish between pressure drop caused by cooling and pressure drop caused by leakage; if the corrected pressure attenuation rate exceeds the micro-leakage threshold, it is identified as a micro-leakage and an alarm message is generated. The edge computing control unit performs the following process for rapid leak detection and automatic shutdown: real-time monitoring of sudden pressure drops; when the inlet pressure sensor (P1) and outlet pressure sensor (P2) simultaneously detect a drop and the first electric shut-off valve (V1) is closed, the pressure difference between the inlet pressure sensor P1 and the outlet pressure sensor (P2) is compared; if the pressure of the inlet pressure sensor P1 increases and the outlet ultrasonic sensor (U2) shows backflow characteristics, an indoor leak is identified and the second electric shut-off valve (V2) is shut off; simultaneously, flow data is collected through the inlet ultrasonic sensor (U1) and the outlet ultrasonic sensor (U2), and the difference between the inlet and outlet flow rates is graded and scored; when the graded score exceeds a preset threshold, a slow leak detection mode is activated for secondary confirmation or an alarm is triggered.

3. The intelligent safety and health diagnostic device for a home heating system according to claim 1, characterized in that, The edge computing control unit is also used to perform congestion diagnosis steps, specifically for: Acquire the temperature data of the inlet water temperature sensor (T1) and the outlet water temperature sensor (T2), as well as the flow rate data of the inlet water ultrasonic sensor (U1) and the outlet water ultrasonic sensor (U2); The temperature difference between the inlet and outlet water is calculated and compared with the pre-established historical reference temperature difference. When the temperature difference between the inlet and outlet water is higher than the historical reference temperature difference, and the flow rate change amplitude of the inlet ultrasonic sensor (U1) and the outlet ultrasonic sensor (U2) is lower than a preset threshold, it is determined that the system is airlocked and a diagnostic warning is generated.

4. The intelligent safety and health diagnostic device for a home heating system according to claim 1, characterized in that, The edge computing control unit is also used to perform valve health self-checks and predictive maintenance: During the self-test cycle, action commands are sent sequentially to the first electric shut-off valve (V1) and the second electric shut-off valve (V2), and the motor drive current is monitored in real time to generate a current curve containing the starting peak current, running current and the arrival current; the action time required for the valve to complete the action from receiving the command is recorded. The valve health score is calculated based on the initial health scoring model, which is: Valve Health Score = Base Score × (Time Coefficient × α + Current Coefficient × β + Sealing Coefficient × γ); where the time coefficient is generated based on the deviation between the action time and the reference time, the current coefficient is generated based on the matching degree between the current curve and the abnormal mode characteristics, the sealing coefficient is generated based on the sealing level of the subsequent pressure holding test, and α, β, and γ are preset weighting coefficients; based on the valve health score and its corresponding scoring level, the valve health status and predictive maintenance suggestions are output.

5. The intelligent safety and health diagnostic device for a home heating system according to claim 1, characterized in that, The edge computing control unit is also used to perform intelligent learning and adaptive scene mode switching. Specifically, during the preset learning period after installation, it collects and records historical data of the multi-sensor fusion module and establishes a personalized benchmark library that reflects the pressure range and flow range. The system monitors the current ambient temperature in real time and dynamically adjusts the sensitivity threshold of leak detection based on the current ambient temperature. When the flow data of the inlet ultrasonic sensor (U1) and the outlet ultrasonic sensor (U2), as well as the temperature data of the inlet temperature sensor (T1) and the outlet temperature sensor (T2) are detected to increase synchronously during the winter heating season, the sensitivity of leak detection is increased. Conversely, the sensitivity is reduced or the monitoring function is put into a dormant state during the non-heating season.

6. The intelligent safety and health diagnostic device for a home heating system according to claim 1, characterized in that, It also includes a communication module connected to the edge computing control unit, which is used to achieve unified management of multiple properties through the communication module: establishing a hierarchical account system of a main account and at least one sub-account, with each account associated with the device of one or more properties in different locations, the main account having global management authority, and the sub-account having viewing and control authority within a preset range; A unified dashboard interface is generated through a cloud-based intelligent platform. This interface aggregates and displays the real-time status, key health indicators, and current alarm information of all associated properties in the form of property cards. The key health indicators include leakage confidence score, valve health score, blockage diagnosis results, and overall system health score. The platform receives quick operation commands from users on any property card within the unified dashboard, generates and sends remote control commands to the corresponding remote devices, and performs emergency shutdown, valve self-test, sensitivity adjustment, and mode switching. The cloud-based intelligent platform connects to the edge computing control units of each remote property via an encrypted communication link, supports the aggregation, storage, and synchronization of status data from multiple devices, and employs a breakpoint resume mechanism to re-transmit historical data when the network recovers.

7. The intelligent safety and health diagnostic device for a home heating system according to claim 1, characterized in that, It also includes a dual power supply system, which includes: a main power supply module for connecting to the external AC power grid and converting it to DC power supply; a backup lithium battery module using lithium iron phosphate cells for seamless power switching when the main power supply is interrupted; the edge computing control unit is also used to monitor the main power supply status and backup battery power in real time, and record the interruption event and control the device to enter a low power mode when the main power supply is interrupted, and generate a low power warning when the battery power is lower than a preset threshold.

8. The intelligent safety and health diagnostic device for a home heating system according to claim 1, characterized in that, The edge computing control unit is pre-configured with a priority-based conflict resolution strategy. Its multi-layer decision architecture includes: a data acquisition layer, which collects raw data from the multi-sensor fusion module at 10ms intervals and performs data quality checks and caching; a feature extraction layer, which performs filtering calibration, difference / rate of change calculation, and spectrum analysis on the preprocessed data to extract multi-dimensional features; a model inference layer, which runs at least several AI models, including a leak detection model, a blockage diagnosis model, and a valve health model, and performs independent inference based on the multi-dimensional features, outputting their respective confidence levels or diagnostic results; and a decision arbitration layer, which receives the inference results of each model, user instructions, and scheduled task requests, and makes a decision based on a pre-configured priority matrix, wherein the priority of leak events is higher than that of user instructions and scheduled self-check tasks, to determine the final system action to be executed.

9. The intelligent safety and health diagnosis device for a home heating system according to claim 1, characterized in that, The multi-evidence fusion algorithm is specifically as follows: (1) Obtaining scoring data from at least two types of evidence sources, including: a flow difference evidence score X generated based on the flow difference data collected by the inlet ultrasonic sensor (U1) and the outlet ultrasonic sensor (U2), wherein the flow difference evidence score X is classified according to the flow difference ΔQ = |U1 flow - U2 flow| per unit time: when 0.5L / h≤ΔQ<2L / h, the score X is 1-30 points; when 2L / h≤ΔQ<5L / h, the score X is 30-60 points; when ΔQ≥ 5L / h, the score X is 60-100 points; a pressure drop evidence score Y generated based on the pressure data collected by the inlet pressure sensor (P1) and the outlet pressure sensor (P2), wherein the pressure drop evidence score Y is calculated based on the pressure change rate dP / dt within a preset time window and the inlet and outlet pressure difference ΔP = |P1 - P2|. (2) Calculate the leak detection confidence score S using a weighted scoring system: Where ω1 is the weighting coefficient for flow difference evidence, and ω2 is the weighting coefficient for pressure evidence, and ; (3) Perform multi-level decision-making based on the confidence score S: when S < 60 points, it is determined to be a normal state and routine monitoring continues; when 60 points ≤ S < 80 points, it is determined to be a suspected leak state, triggering the early warning mechanism and controlling the valve to enter the slow leak detection mode for verification; when S ≥ 80 points, it is determined to be a confirmed leak state, generating a shut-off command within ≤ 3 seconds and simultaneously controlling the first electric shut-off valve (V1) and the second electric shut-off valve (V2) to perform shut-off actions.

10. A control method for an intelligent safety and health diagnostic device for a home heating system according to any one of claims 1 to 9, characterized in that, Includes the following steps: Step S1: Real-time acquisition of multi-source data: Through the multi-sensor fusion module, the inlet water pressure (P1), outlet water pressure (P2), inlet water temperature (T1), outlet water temperature (T2), inlet water flow rate, and outlet water flow rate of the heating system are collected in real time at a preset sampling period, and the motor drive current and power status of the first electric shut-off valve (V1) and the second electric shut-off valve V2 are monitored simultaneously. Step S2, Edge Feature Extraction and Preprocessing: The edge computing control unit filters, calibrates and compensates the collected raw data; extracts the pressure change rate, inlet and outlet flow rate difference, and temperature gradient features; and performs spectrum analysis on the flow data collected by the ultrasonic sensor to extract the energy features of the preset frequency band. Step S3, Multi-model Parallel Inference: The extracted feature data is input into multiple models pre-installed in the AI ​​coprocessor for parallel inference. These multiple models include at least: A leak detection model generates a leak detection confidence score based on a multi-evidence fusion algorithm. A blockage diagnosis model identifies pipe siltation or air blockage based on flow trends and flow characteristics; A valve health model generates a valve health score based on current curve analysis and action time monitoring. Step S4: Priority-based decision arbitration: The decision arbitration layer receives the inference results of each model and makes a decision based on the preset priority matrix; wherein, when the leakage detection confidence score exceeds the first preset threshold, an early warning instruction is generated; when it exceeds the second preset threshold which is higher than the first preset threshold, it is decided as the highest priority event and an emergency shutdown instruction is generated immediately. Step S5, Precise Control of the Actuator: Execute the corresponding action according to the arbitration instruction. If it is an emergency shutdown instruction, the first electric shut-off valve (V1) and the second electric shut-off valve (V2) will be controlled to shut off within ≤3 seconds, and a local audible and visual alarm and a remote push notification will be triggered. If it is a warning or diagnostic instruction, the corresponding health status, warning information or maintenance suggestions will be output through the local display interface and the remote client. Step S6, Data Caching and Synchronization: All running data and event logs are encrypted and stored in local storage media; when the network is connected, the data is synchronized to the cloud server through the communication module. The synchronization strategy adopts the mechanism of resuming interrupted transmission and prioritizing the transmission of important events.

Images