Intelligent constant-temperature water valve based on AI small model and control method of intelligent constant-temperature water valve

Abstract

According to the intelligent constant-temperature water valve based on the AI small model and the control method of the intelligent constant-temperature water valve, the temperature of a cold water inlet and the temperature of a hot water inlet are monitored in real time through multiple temperature sensors, the opening degree of a valve element is dynamically adjusted in combination with an intelligent sensing AI small model algorithm, and a cold water pre-discharging module is integrated to solve the problem of initial cold water waste; pure temperature feedback closed-loop control is adopted, and the problem of unstable temperature caused by water pressure fluctuation is solved; real-time parameters are introduced for the AI small model through nonlinear flow monitoring and sensor hardware, the parameters are intelligently adjusted in combination with the neural network technology, and the adjusting precision and the multi-scene application capacity are improved. In addition, the cold water pre-drainage module triggers drainage through a temperature difference threshold value, water resource waste is reduced, and the system is suitable for the fields of household shower, commercial hot water systems and the like and has the advantages of being high in cost performance and intelligent.

Description

Technical Field

[0001] This invention relates to the technical field of intelligent water valves, and in particular to an intelligent thermostatic water valve based on an AI small model and its control method. Background Technology

[0002] A smart water valve is a device that automatically adjusts the fluid flow rate based on a set temperature using technologies such as electric drive or thermal elements, thereby controlling the water temperature and output. It offers numerous advantages, including precise temperature control, energy efficiency, ease of use, and a high degree of automation.

[0003] However, existing smart water valves have the following defects: (1) When the difference between hot and cold water pressure is too large or the household water pressure is too low, the smart water valve may automatically shut off or the water temperature regulation may be unstable; (2) Cold water that cannot be discharged in time can easily accumulate in the cavity of the smart water valve. During the water discharge process, the water valve needs to be opened for a long time to wait for the water temperature to be adjusted to the preset value, resulting in the waste of water resources; (3) The smart water valve may be incompatible with some smart devices and cannot achieve seamless docking and linkage control. Summary of the Invention

[0004] The purpose of this invention is to address the shortcomings of existing technologies by providing an intelligent thermostatic water valve and its control method based on an AI small model. This valve monitors the inlet temperatures of hot and cold water in real time using multiple temperature sensors, dynamically adjusts the valve core opening using an intelligent sensing AI small model algorithm, and integrates a pre-draining cold water module to solve the problem of initial cold water waste.

[0005] To achieve the above objectives, the technical solution provided by this invention is: an intelligent thermostatic water valve based on an AI small model, comprising:

[0006] A mixing chamber is provided with a cold water inlet / outlet channel, a hot water inlet / outlet channel, a mixed water inlet / outlet channel, and a drainage channel. The cold water inlet / outlet channel is connected to a cold water pipe, and the hot water inlet / outlet channel is connected to a hot water pipe. Both the cold water inlet / outlet channel and the hot water inlet / outlet channel are equipped with pressure sensors.

[0007] A multi-channel temperature sensing module, comprising at least one temperature sensor, which is embedded in the inner wall of the cold water inlet and outlet channels and the hot water inlet and outlet channels and is located at the cold water pipeline and the hot water pipeline respectively, for real-time acquisition of the cold and hot water inlet temperatures;

[0008] A flow sensor module, which is installed at the inlet and outlet channels of the mixed water, includes at least one water flow sensor for detecting the flow rate of the water valve;

[0009] An actuating adjustment mechanism is provided, comprising a micro motor, a gear reduction group, and a valve core assembly. The valve core assembly is disposed within the mixing chamber and is linked to the output shaft of the micro motor via the gear reduction group to adjust the opening ratio of the cold and hot water inlet and outlet channels.

[0010] A pre-drained cold water module includes a drain valve and a retractable drain rod. The drain valve is located at the inlet and outlet of the drain channel. One end of the retractable drain rod is linked to the valve stem of the drain valve, and the other end of the retractable drain rod extends to the bottom of the mixing chamber to drain the low-temperature water accumulated in the mixing chamber.

[0011] The control module includes a first main controller, an AI mini-model unit, a signal processing circuit, a wireless transceiver, and a human-machine interaction unit. The first main controller is connected to the control terminals of a multi-channel temperature sensing module, a flow sensor module, a micro motor, and a drain valve through the signal processing circuit to receive temperature and flow signals. The first main controller is communicatively connected to the AI ​​mini-model unit to calculate the valve core assembly opening, drainage trigger command, and dynamic flow distribution strategy. The wireless transceiver is communicatively connected to the AI ​​mini-model unit to receive water quality, pollution, and pH parameters and feed them back to the AI ​​mini-model unit. The human-machine interaction unit is connected to the first main controller to display real-time water quality data.

[0012] Furthermore, the intelligent thermostatic water valve includes:

[0013] The sensing and monitoring module includes a second main controller, an information monitoring module, and a wireless transceiver. The second main controller is communicatively connected to the information monitoring module. The information monitoring module is equipped with a multi-sensor group and an edge computing terminal. The multi-sensor group can form M network combinations, where M≥1. The multi-sensor group includes a heavy metal sensor, a chemical component detection sensor, a pH sensor, a residual chlorine sensor, and a water quality sensor. The information monitoring module is interconnected with the wireless transceiver of the control module through a wireless network to realize real-time monitoring of water quality parameters, alarm for exceeding standards, and uploading of monitoring data to the cloud.

[0014] Furthermore, the multi-channel temperature sensing module includes a first water temperature sensor and a second water temperature sensor; the first water temperature sensor is encapsulated in the inner wall of the cold water inlet / outlet channel and is provided with a sealed sleeve, the end of the sealed sleeve being provided with a filter screen for filtering impurities; the second water temperature sensor is encapsulated in the inner wall of the hot water inlet / outlet channel and is provided with a sealed sleeve, the end of the sealed sleeve being provided with a filter screen for filtering impurities; in the water usage nodes composed of multiple sets of cold water inlet / outlet channels and hot water inlet / outlet channels, there are N sets of first water temperature sensors and second water temperature sensors, where N>1, forming a cold and hot mixed water supply network to realize multi-node temperature monitoring and differentiated water supply; the first water temperature sensor and the second water temperature sensor are communicatively connected to the signal processing circuit of the control module, the signal processing circuit including a constant current source module for converting voltage signals into digital temperature values, and the sampling frequency of the signal processing circuit is not less than 10Hz.

[0015] Furthermore, the drainage channel includes:

[0016] The main drain pipe has a spiral guide groove on its inner wall to prevent water accumulation.

[0017] A filter screen is provided at the outlet end of the drain valve of the pre-drained cold water module;

[0018] A backwashing port is provided at the end of the main drain pipe and is used to connect an external high-pressure water pipe to backwash the drain channel.

[0019] A control method for an intelligent thermostatic water valve based on an AI mini-model is provided. This method employs a pure temperature feedback closed-loop control in the control module, using multiple temperature sensors to monitor the temperatures of the cold water inlet and outlet channels and the hot water inlet and outlet channels in real time. The control module inputs real-time monitoring parameters from multiple temperature sensor modules, flow sensor modules, the actuator, and the pre-drain cold water module into its AI mini-model unit. This AI mini-model unit uses a lightweight neural network to perceive and monitor the data in real time, dynamically adjusting the flow distribution strategy to regulate the valve core assembly opening. Furthermore, the AI ​​mini-model unit is capable of self-learning and anomaly self-diagnosis. The control module triggers drainage from the pre-drain cold water module based on a temperature difference threshold, discharging the low-temperature wastewater remaining in the mixing chamber. The multi-sensor group built into the sensing and monitoring module is further configured into a network for monitoring the water quality at each water-using node.

[0020] Furthermore, the control module acquires real-time monitoring parameters from the multi-channel temperature sensing module, flow sensor module, actuator adjustment mechanism, and pre-drained cold water module and inputs them into its AI mini-model unit. The AI ​​mini-model unit uses a lightweight neural network to perceive the monitoring data in real time, dynamically adjust the flow distribution strategy, and thereby adjust the valve core assembly opening, including the following steps:

[0021] S1. Input parameter fusion: The AI ​​small model unit receives multi-source input data, including temperature difference, water pressure difference, historical water use habit data, and water quality parameters fed back by the sensing and monitoring module.

[0022] S2. Model Training and Inference: Pre-store water usage scenario data, extract features from the water usage scenario data, build an AI mini-model based on LSTM or lightweight CNN, match the closest historical scenario based on real-time input data, call the corresponding AI mini-model for inference, and output the corrected flow distribution coefficient K, that is, the ratio of hot water to cold water, where K is a variable, K∈[0.3,1.2].

[0023] S3. Flow distribution execution: Define a multi-dimensional input vector based on temperature difference, pressure difference, historical water usage data, and water quality parameters; then perform feature fusion on the multi-dimensional input vector using a nonlinear activation function to generate a comprehensive decision vector; define the flow ratio formula based on the comprehensive decision vector; finally, map the flow ratio to the valve core displacement command to dynamically adjust the flow distribution strategy, thereby regulating the valve core assembly opening.

[0024] Furthermore, the AI ​​mini-model unit is capable of self-learning and anomaly self-diagnosis, including:

[0025] User habit learning: By recording the user's preset temperature setpoint T_set, water usage duration, and flow rate adjustment operations within a preset time period, a personalized water usage curve is generated, and the parameters of the basic piecewise function f(ΔT) are automatically optimized, where ΔT is the temperature difference between the hot and cold water inlets.

[0026] Scene Adaptation: When the sensing and monitoring module detects abnormal water quality and the water quality threshold set by TDS, the AI ​​mini-model automatically reduces the proportion of hot water, that is, reduces the K value, to ensure water quality safety; when the water pressure fluctuation ΔP is detected to be greater than the set water pressure fluctuation threshold, the threshold of ΔT is dynamically adjusted to avoid frequent drainage.

[0027] Fault self-diagnosis: (1) Temperature sensor fault: If the difference between the first water temperature sensor and the second water temperature sensor exceeds the preset temperature threshold, the first water temperature sensor is marked as abnormal, and the second water temperature sensor is switched to the backup and a prompt is issued; (2) Motor stall fault: If the micro motor fails to reach the target displacement for the preset duration, the fault is recorded and the drainage protection is triggered; (3) Communication interruption fault: If the wireless connection with the sensing and monitoring module is interrupted for more than the preset connection duration, the pre-drainage cold water function is turned off and the hot water ratio K is limited to prevent high temperature scalding.

[0028] Furthermore, the control module controls the pre-drainage cold water module to trigger drainage based on the temperature difference threshold, thereby discharging the low-temperature wastewater remaining in the mixing chamber, including:

[0029] When the initial mixing water temperature T in the mixing chamber mix When the difference between the temperature and the first set temperature T_set1 reaches the drainage threshold ΔT1, the first main controller sends an opening command to the drainage valve.

[0030] Drainage process control: The drain rod achieves micro-displacement control via a micro-motor driven gear set, with a displacement accuracy of ±0.1mm and a telescopic stroke range of 0-30mm; when the mixed water temperature T mix When the temperature is not lower than the second set temperature T_set2, the first main controller closes the drain valve and stops draining;

[0031] Low temperature protection forced drainage: If the temperature sensed by the first water temperature sensor remains below the preset low temperature threshold during the drainage process, the first main controller will extend the drainage time of the drain valve to avoid low temperature water residue.

[0032] Anti-clogging protection: If resistance exceeds the threshold when the drain rod extends, drainage will stop immediately and a fault alarm will be triggered.

[0033] Furthermore, the control module has anti-interference and redundancy mechanisms, including:

[0034] Multi-sensor redundancy verification: The difference between the first water temperature sensor and the second water temperature sensor is compared in real time. If the difference exceeds the redundancy threshold, a fault alarm is triggered and the system switches to the backup sensor.

[0035] Water pressure fluctuation compensation: The mixing chamber has a built-in pressure sensor, which corrects the flow ratio formula using the pressure signal. The compensation coefficient Cp = 1 + 0.01 × (P hot -P cold ), where P hot P cold These are the inlet pressures for hot and cold water, respectively.

[0036] Motor stall protection: If the micro motor fails to reach the target displacement within a preset time, the first main controller immediately stops the power supply to the micro motor and records the fault;

[0037] Scalding Prevention: The mixing water inlet and outlet channels are equipped with two sets of water temperature sensors. When the mixed water temperature exceeds the preset temperature 'a' degrees Celsius, the hot water flow is reduced. If the mixed water temperature exceeds the preset temperature 'b' degrees Celsius, the hot water is completely shut off to prevent scalding. Specifically, 'a'... <b;

[0038] Power failure memory: The first controller has a built-in memory that stores the user-set target temperature T_set, drainage threshold ΔT1, and historical fault records.

[0039] Furthermore, the sensing and monitoring module performs the following operations:

[0040] M1, Multi-source data acquisition:

[0041] M11 collects water quality parameters in real time through a group of heavy metal sensors, chemical composition detection sensors, pH sensors, turbidity sensors and water quality sensors.

[0042] M12, synchronously acquire environmental parameters and user behavior data, and cover water-using nodes through distributed sensor groups M≥1;

[0043] M2, Data Preprocessing and Feature Extraction:

[0044] M21. Based on the edge computing terminal, water quality parameters are filtered, denoised, dynamically calibrated, and normalized to generate standardized data packets.

[0045] M22. Extract key feature parameters from standardized data packets and compress the data volume using a lightweight feature encoding algorithm;

[0046] M3, Local Anomaly Detection and Alarms:

[0047] M31. Deploy a lightweight rule engine at the edge to make real-time judgments based on preset thresholds and trigger local alarms;

[0048] M32. If a water pollution event is detected, the drainage protection mechanism is activated, the pre-discharge cold water module is activated to discharge the polluted water, and the event timestamp and pollution parameters are recorded.

[0049] M4, Cloud-based Collaborative Analytics:

[0050] M41. Data is uploaded to the cloud server via a wireless transceiver, and a federated learning framework is used to achieve cross-regional water quality data sharing and train a water pollution prediction model.

[0051] M42: The cloud generates a water quality risk assessment report based on multi-dimensional data and pushes it to the user's terminal.

[0052] M5, Dynamic Feedback and Adaptive Adjustment:

[0053] M51 receives control commands from the cloud and adjusts the valve core opening through the control module to optimize the mixed water quality;

[0054] M52. When abnormal sensor data is detected, a preset self-calibration procedure is triggered to calibrate based on standard solutions or cross-validate with redundant sensors, and the fault node is marked.

[0055] M6. Historical Data Management and Maintenance:

[0056] M61, the local storage module records at least 12 months of water quality time-series data, which can be retrieved by time axis and pollutant type;

[0057] M62. Regularly perform self-check procedures, generate maintenance logs, and encrypt and upload them to the blockchain evidence storage platform to ensure that the data is tamper-proof.

[0058] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0059] This invention employs pure temperature feedback closed-loop control, resolving the temperature instability issue caused by water pressure fluctuations. Through nonlinear flow monitoring and sensor hardware, real-time parameters are introduced into the AI ​​mini-model, and combined with neural network technology, these parameters are intelligently adjusted, improving regulation accuracy and multi-scenario application capabilities. Furthermore, the pre-discharge cold water module triggers drainage based on temperature difference thresholds, reducing water waste. It is suitable for applications such as home showers and commercial hot water systems, offering high cost-effectiveness and intelligent advantages. Attached Figure Description

[0060] Figure 1 This is a schematic diagram of the present invention. Detailed Implementation

[0061] The present invention will be further described below with reference to specific embodiments.

[0062] Example 1

[0063] See Figure 1 As shown, the intelligent thermostatic water valve based on an AI small model provided in this embodiment includes:

[0064] 1) A mixing chamber 1, comprising a cold water inlet / outlet channel 101, a hot water inlet / outlet channel 102, a mixed water inlet / outlet channel 103, and a drainage channel. The cold water inlet / outlet channel 101 is connected to a cold water pipe, and the hot water inlet / outlet channel 102 is connected to a hot water pipe. Both the cold water inlet / outlet channel 101 and the hot water inlet / outlet channel 102 are equipped with pressure sensors. The drainage channel includes a main drain pipe, a filter screen, and a backwash port. The inner wall of the main drain pipe is provided with a spiral guide groove to prevent water accumulation. The filter screen is located at the outlet end of the drain valve of the pre-drained cold water module. The backwash port is located at the end of the main drain pipe and is used to connect an external high-pressure water pipe for backwashing the drainage channel.

[0065] 2) Multi-channel temperature sensing module 2, comprising at least one temperature sensor embedded in the inner wall of the cold water inlet / outlet channel and the hot water inlet / outlet channel, respectively located at the cold water pipeline and the hot water pipeline, for real-time acquisition of the cold and hot water inlet temperatures; the multi-channel temperature sensing module 2 includes a first water temperature sensor and a second water temperature sensor; the first water temperature sensor is encapsulated in the inner wall of the cold water inlet / outlet channel 101 and is provided with a sealed sleeve, the end of which is provided with a filter screen for filtering impurities; there are at least two second water temperature sensors, one as the main sensor for measuring the hot water / mixed water temperature, and the other as a backup sensor for measuring the hot water / mixed water temperature. The second water temperature sensor is encapsulated in the inner wall of the hot water inlet / outlet channel 102 and is provided with a sealed sleeve. The end of the sealed sleeve is provided with a filter screen for filtering impurities. In the water-using node composed of multiple sets of cold water inlet / outlet channels 101 and hot water inlet / outlet channels 102, there are N sets of first water temperature sensors and second water temperature sensors, where N>1, forming a cold and hot mixed water supply network to realize multi-node temperature monitoring and differentiated water supply. The first water temperature sensor and the second water temperature sensor are communicatively connected to the signal processing circuit 603 of the control module. The signal processing circuit includes a constant current source module for converting voltage signals into digital temperature values. The sampling frequency of the signal processing circuit is not less than 10Hz.

[0066] 3) Flow sensor module 3, which is installed at the inlet and outlet channels of the mixed water and includes at least one water flow sensor for detecting the flow rate of the water valve.

[0067] 4) The adjustment mechanism 4 includes a micro motor 401, a gear reduction group 402 and a valve core assembly 403. The valve core assembly 403 is disposed in the mixing chamber and is linked with the output shaft of the micro motor 401 through the gear reduction group 402 to adjust the opening ratio of the cold and hot water inlet and outlet channels.

[0068] 5) Pre-draining cold water module 5, the pre-draining cold water module 5 includes a drain valve and a retractable drain rod, the drain valve is set at the inlet and outlet of the drain channel, one end of the retractable drain rod is linked to the valve stem of the drain valve, and the other end of the retractable drain rod extends to the bottom of the mixing chamber for draining the low-temperature water accumulated in the mixing chamber.

[0069] 6) Control module 6, which includes a first main controller 601, an AI mini-model unit 602, a signal processing circuit 603, a wireless transceiver 604, and a human-machine interaction unit 605. The first main controller 601 is connected to the control terminals of the multi-channel temperature sensing module 2, the flow sensor module 3, the micro motor 401, and the drain valve through the signal processing circuit 603, and is used to receive temperature and flow signals. The first main controller 601 is communicatively connected to the AI ​​mini-model unit 602 and is used to calculate the opening degree of the valve core assembly 403, the drain trigger command, and the dynamic flow distribution strategy. The wireless transceiver 604 is communicatively connected to the AI ​​mini-model unit 602 and is used to receive water quality, pollution, and pH parameters and feed them back to the AI ​​mini-model unit 602. The human-machine interaction unit 605 is a display screen connected to the first main controller 601 and is used to display real-time water quality data.

[0070] The control module includes the following functions:

[0071] Automatic mode: After the user sets T_set, the system collects T1 and T2 in real time through dual temperature sensors, and automatically adjusts the valve core opening and triggers pre-drainage of cold water in combination with dynamic flow AI control algorithm.

[0072] Manual mode: The valve core opening can be directly adjusted by the external button preset on the valve body surface of the intelligent thermostatic water valve. At this time, the pre-draining cold water function is turned off and the flow ratio is fixed at 1:1.

[0073] Water-saving mode: When the water flow rate is detected to be below 0.5L / min for 5 consecutive minutes, the hot water inlet opening is automatically reduced by 10% to reduce hot water consumption.

[0074] Child lock mode: Activated by pressing and holding the external button for 3 seconds, it prevents any mode switching and parameter modification, thus preventing accidental operation.

[0075] Bluetooth / Wi-Fi wireless communication: Integrated BLE 5.0 or ESP8266 module, allowing remote setting of T_set, viewing of real-time temperature, flow data and drainage records via smart device APP.

[0076] Voice control: The control module can communicate with a smart speaker and supports voice commands such as "increase the temperature by 2°C" and "start pre-draining cold water".

[0077] Water usage statistics: Calculate daily / monthly water consumption based on cumulative flow data, generate water-saving reports, and push them to user terminals.

[0078] Self-learning mode: Automatically adjusts T_set and pre-drained cold water threshold ΔT1 based on user water usage habits.

[0079] 7) Sensing and monitoring module 7, which includes a second main controller 701, an information monitoring module, and a wireless transceiver 703. The second main controller 701 is communicatively connected to the information monitoring module. The information monitoring module includes a safety monitoring module 7021 and a parameter monitoring module 7022. It is equipped with a multi-sensor group and an edge computing terminal. The multi-sensor group can form M combined networks, where M≥1. The multi-sensor group includes a heavy metal sensor, a chemical component detection sensor, a pH sensor, a residual chlorine sensor, and a water quality sensor. The information monitoring module is interconnected with the wireless transceiver 604 of the control module 6 through the wireless network of the wireless transceiver 703 to realize real-time monitoring of water quality parameters, alarm for exceeding standards, and uploading of monitoring data to the cloud.

[0080] Example 2

[0081] The control method for the intelligent thermostatic water valve based on the AI ​​small model described in Embodiment 1 provided in this embodiment is as follows: The control module adopts pure temperature feedback closed-loop control, and monitors the temperature of the cold water inlet and outlet channels and the hot water inlet and outlet channels in real time through multiple temperature sensors. The control module obtains real-time monitoring parameters from the multiple temperature sensor module, flow sensor module, actuator, and pre-drain cold water module and inputs them into its AI small model unit. The AI ​​small model unit adopts a lightweight neural network to perceive the monitoring data in real time, dynamically adjust the flow distribution strategy, and then adjust the opening of the valve core assembly. The AI ​​small model unit can perform self-learning and anomaly self-diagnosis. The control module controls the pre-drain cold water module to trigger drainage according to the temperature difference threshold, and discharges the low-temperature wastewater remaining in the mixing chamber. Through the multi-sensor group built into the sensing and monitoring module, it is further configured into a combined network for monitoring the water quality of each water-using node.

[0082] 1) The control module acquires real-time monitoring parameters from the multi-channel temperature sensor module, flow sensor module, actuator, and pre-drained cold water module and inputs them into its AI mini-model unit. The AI ​​mini-model unit uses a lightweight neural network to perceive the monitoring data in real time, dynamically adjust the flow distribution strategy, and thereby adjust the valve core assembly opening, including the following steps:

[0083] S1. Input parameter fusion: The AI ​​small model unit receives multi-source input data, including temperature difference, water pressure difference, historical water use habit data, and water quality parameters fed back by the sensing and monitoring module.

[0084] S2. Model Training and Inference: Pre-store millions of water usage scenario data; extract features from the water usage scenario data, build an AI mini-model based on LSTM or lightweight CNN, match the closest historical scenario based on the real-time input data, call the corresponding AI mini-model for inference, and output the corrected flow distribution coefficient K, that is, the ratio of hot water to cold water, where K is a variable, K∈[0.3,1.2].

[0085] S3, Traffic Allocation Execution:

[0086] S3.1 Input Parameter Space: Define a multidimensional input vector X = [ΔT, ΔP, H, P], where: ΔT is the temperature difference between the hot and cold water inlets, ΔT = |T_hot - T_cold|; ΔP is the pressure difference between the hot and cold water inlets, ΔP = P hot -P cold The data is collected by a pressure sensor; H is the historical water use habit feature vector, which includes time period, frequency and temperature preference; P is the real-time parameter vector fed back by the sensing and monitoring module, and the water quality parameters include water quality, flow fluctuation and ambient temperature.

[0087] S3.2 Feature Fusion Layer: The input vector is fused using a non-linear activation function σ(·) to generate a comprehensive decision vector D = σ(W1·X + b1), where W1 is the trainable weight matrix, b1 is the bias vector, and X is the input vector.

[0088] S3.3, Dynamic Allocation Function: Defines the flow ratio Q hot / Q cold = f(D), where f(D) is a differentiable nonlinear function that satisfies: f(D) = α·tanh(β·D) + γ·ln(1 + exp(δ·D)), where α, β, γ, and δ are the model parameters, which are dynamically optimized through reinforcement learning;

[0089] S3.4, Execution Control: Map the flow ratio to the valve core displacement command S = g(f(D)), where g() is the valve core displacement-flow ratio conversion function, which is determined by the dynamic characteristics of the actuator.

[0090] 2) The AI ​​mini-model unit is capable of self-learning and anomaly self-diagnosis, including:

[0091] User habit learning: By recording the user's temperature setting value T_set, water usage duration, and flow rate adjustment operations over 30 days, a personalized water usage curve is generated, and the parameters of the basic piecewise function f(ΔT) are automatically optimized.

[0092] Scene Adaptation: When the sensing and monitoring module detects abnormal water quality (TDS > 500 ppm), the AI ​​mini-model automatically reduces the proportion of hot water, i.e., the K value decreases by 0.2, to ensure water quality safety; when water pressure fluctuation (ΔP > 0.1 MPa) is detected, the threshold of ΔT is dynamically adjusted to avoid frequent drainage.

[0093] Fault self-diagnosis: (1) Temperature sensor failure: If the difference between the first water temperature sensor and the second water temperature sensor continues to exceed 1℃, it is marked as an abnormality of the first water temperature sensor, and the second water temperature sensor is switched to the backup and a prompt is issued; (2) Motor stall failure: If the micro motor fails to reach the target displacement for 5 consecutive seconds, the fault is recorded and the drainage protection is triggered; (3) Communication interruption failure: If the wireless connection with the sensing and monitoring module is interrupted for more than 30 seconds, the pre-drainage cold water function is turned off and the hot water ratio K is limited to ≤0.8 to prevent high temperature burns.

[0094] 3) The control module controls the pre-drainage cold water module to trigger drainage based on the temperature difference threshold, thereby discharging the low-temperature wastewater remaining in the mixing chamber, including:

[0095] When the initial mixing water temperature T in the mixing chamber mix When the difference between the temperature and the set temperature T_set reaches the drainage threshold ΔT1, the first main controller sends an opening command to the drainage valve:

[0096] The initial mixed water temperature T in the mixing chamber mix The calculation formula is as follows:

[0097]

[0098] Where T1 is the first water temperature sensor, T2 is the second water temperature sensor, and Q... cold Q is the cold water flow rate. hot This refers to the hot water flow rate.

[0099] Drainage process control: The drainage rod achieves micro-displacement control through a micro-motor driven gear set, with a displacement accuracy of ±0.1mm and a telescopic stroke range of 0-30mm; when T mix When T_set is greater than or equal to T, the first master controller closes the drain valve and stops draining.

[0100] Low temperature protection forced drainage: If the temperature T1 of the cold water inlet and outlet channels remains below the preset low temperature threshold during the drainage process, the first main controller will extend the drainage time of the drain valve to avoid low temperature water residue.

[0101] Anti-clogging protection: If resistance exceeds the threshold when the drain rod extends, drainage will stop immediately and a fault alarm will be triggered.

[0102] 4) The control module has anti-interference and redundancy mechanisms, including:

[0103] Multi-sensor redundancy verification: The difference between the first and second water temperature sensors is calculated in real time. If the difference exceeds the redundancy threshold of 0.5℃, a fault alarm is triggered and the system switches to the backup sensor.

[0104] Water pressure fluctuation compensation: The mixing chamber has a built-in pressure sensor with a range of 0-1 MPa and an accuracy of ±0.5% FS. The flow ratio calculation formula is corrected based on the pressure signal, with a compensation coefficient Cp = 1 + 0.01 × (P hot -P cold ), where P hot P cold These are the inlet pressures for hot and cold water, respectively.

[0105] Motor stall protection: If the micro motor fails to reach the target displacement for 3 consecutive seconds, the first main controller will immediately stop the power supply to the micro motor and record the fault.

[0106] Scalding Prevention: The mixing water inlet and outlet channels are equipped with two sets of water temperature sensors. When the mixed water temperature exceeds the preset temperature 'a' degrees Celsius, the hot water flow is reduced. If the mixed water temperature exceeds the preset temperature 'b' degrees Celsius, the hot water is completely shut off to prevent scalding. Specifically, 'a'... <b;

[0107] Power-off memory: The first controller has a built-in EEPROM memory with a capacity of 24LC256, which stores the user-set target temperature T_set, drainage threshold ΔT1, and historical fault records.

[0108] The sensing and monitoring module performs the following operations:

[0109] M1, Multi-source data acquisition:

[0110] M11 collects water quality parameters in real time through a group of heavy metal sensors, chemical composition detection sensors, pH sensors, turbidity sensors and water quality sensors.

[0111] M12, synchronously acquire environmental parameters and user behavior data, and cover water-using nodes through distributed sensor groups M≥1;

[0112] M2, Data Preprocessing and Feature Extraction:

[0113] M21. Based on the edge computing terminal, water quality parameters are filtered, denoised, dynamically calibrated, and normalized to generate standardized data packets.

[0114] M22. Extract key feature parameters from standardized data packets and compress the data volume using a lightweight feature encoding algorithm, achieving a compression rate of ≥80%.

[0115] M3, Local Anomaly Detection and Alarms:

[0116] M31. Deploy a lightweight rule engine at the edge to make real-time judgments based on preset thresholds and trigger local alarms;

[0117] M32. If a water pollution event is detected, the drainage protection mechanism will be activated immediately, the pre-discharge cold water module will be activated to discharge the polluted water, and the event timestamp and pollution parameters will be recorded.

[0118] M4, Cloud-based Collaborative Analytics:

[0119] M41. Data is uploaded to the cloud server via the wireless communication module, and a federated learning framework is used to achieve cross-regional water quality data sharing and train a water pollution prediction model.

[0120] M42: The cloud generates a water quality risk assessment report based on multi-dimensional data and pushes it to the user's terminal.

[0121] M5, Dynamic Feedback and Adaptive Adjustment:

[0122] M51 receives control commands from the cloud and adjusts the valve core opening through the control module to optimize the mixed water quality;

[0123] M52. When abnormal sensor data is detected, a preset self-calibration procedure is triggered to calibrate based on standard solutions or cross-validate with redundant sensors, and the fault node is marked.

[0124] M6. Historical Data Management and Maintenance:

[0125] M61, the local storage module records at least 12 months of water quality time-series data, which can be retrieved by time axis and pollutant type;

[0126] M62. Regularly perform self-check procedures, generate maintenance logs, and encrypt and upload them to the blockchain evidence storage platform to ensure that the data is tamper-proof.

[0127] The above-described embodiments are merely preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Therefore, any changes made in accordance with the shape and principle of the present invention should be covered within the protection scope of the present invention.

Claims

1. An AI small model-based intelligent thermostatic water valve, characterized in that, The application relates to a water quality intelligent control system, which comprises the following parts: a mixing cavity provided with a cold water inlet and outlet channel, a hot water inlet and outlet channel, a mixed water inlet and outlet channel and a drainage channel, wherein the cold water inlet and outlet channel is connected with a cold water pipeline, the hot water inlet and outlet channel is connected with a hot water pipeline, and the cold water inlet and outlet channel and the hot water inlet and outlet channel are provided with pressure sensors; a multi-path temperature sensing module comprising at least one temperature sensor embedded in the inner walls of the cold water inlet and outlet channel and the hot water inlet and outlet channel and located at the cold water pipeline and the hot water pipeline respectively, which is used for collecting the cold and hot water inlet temperatures in real time; a flow sensor module installed at the mixed water inlet and outlet channel, which comprises at least one water flow sensor and is used for detecting the water valve flow rate; an execution adjusting mechanism comprising a micro motor, a gear reduction set and a valve core assembly, wherein the valve core assembly is arranged in the mixing cavity, is linked with the output shaft of the micro motor through the gear reduction set and is used for adjusting the opening ratio of the cold and hot water inlet and outlet channels; a pre-drainage cold water module comprising a drainage valve arranged at the inlet and outlet ends of the drainage channel and a telescopic drainage rod, one end of the telescopic drainage rod is linked with the valve rod of the drainage valve, the other end of the telescopic drainage rod extends to the bottom of the mixing cavity and is used for draining the low-temperature accumulated water in the mixing cavity; a control module comprising a first main controller, an AI small model unit, a signal processing circuit, a wireless transceiver and a man-machine interaction unit; the first main controller is connected with the multi-path temperature sensing module, the flow sensor module, the micro motor and the control end of the drainage valve through the signal processing circuit respectively and is used for receiving temperature and flow signals; the first main controller is communicatively connected with the AI small model unit and is used for calculating the valve core assembly opening degree, the drainage triggering instruction and the dynamic flow distribution strategy; the wireless transceiver is communicatively connected with the AI small model unit and is used for receiving water quality, pollution and PH value parameters and feeding back the parameters to the AI small model unit; and the man-machine interaction unit is connected with the first main controller and is used for displaying real-time water quality data.

2. The AI small model-based intelligent thermostatic water valve according to claim 1, wherein The application also relates to a water quality intelligent control system, which comprises the following parts: a perception and monitoring module comprising a second main controller and an information monitoring module and a wireless transceiver; the second main controller is communicatively connected with the information monitoring module; the information monitoring module is provided with a multi-sensor group and an edge computing terminal; the multi-sensor group can form M combined networks, M>=1; the multi-sensor group comprises a heavy metal sensor, a chemical component detection sensor, a PH value sensor, a residual chlorine sensor and a water quality sensor; the information monitoring module is interconnected with the wireless transceiver of the control module through a wireless network, realizes real-time monitoring of water quality parameters, over-limit alarm and uploading of monitoring data to the cloud.

3. The AI small model-based intelligent thermostatic water valve according to claim 1, characterized in that: The multi-path temperature sensing module comprises a first water temperature sensor and a second water temperature sensor; the first water temperature sensor is packaged on the inner wall of the cold water inlet and outlet channel and is provided with a sealed sleeve, and the end of the sealed sleeve is provided with a filter screen for filtering impurities; the second water temperature sensor is packaged on the inner wall of the hot water inlet and outlet channel and is provided with a sealed sleeve, and the end of the sealed sleeve is provided with a filter screen for filtering impurities; in a water node formed by multiple groups of cold water inlet and outlet channels and hot water inlet and outlet channels, there are N groups of first water temperature sensors and second water temperature sensors, where N > 1, a cold and hot mixed water supply network is formed, and multi-node temperature monitoring and differentiated water supply are realized; the first water temperature sensor and the second water temperature sensor are in communication connection with a signal processing circuit of the control module, the signal processing circuit comprises a constant current source module for converting a voltage signal into a digital temperature value, and the sampling frequency of the signal processing circuit is not less than 10 Hz.

4. The AI small model-based intelligent thermostatic water valve according to claim 1, wherein The drainage channel comprises: a main drainage pipe, an inner wall of the main drainage pipe is provided with a spiral flow guide groove for avoiding water accumulation; a filter screen cover is arranged at the outlet end of the pre-drainage cold water module; a backwashing interface is arranged at the end of the main drainage pipe, and a high-pressure water pipe is connected to the backwashing interface for reverse flushing of the drainage channel.

5. The control method of the AI small model-based intelligent thermostatic water valve according to any one of claims 1-4, characterized in that: The method is a pure temperature feedback closed-loop control of the control module, and the temperature of the cold water inlet and outlet channel and the hot water inlet and outlet channel is monitored in real time by the multi-path temperature sensor; the control module obtains the real-time monitoring parameters of the multi-path temperature sensing module, the flow sensor module, the execution adjustment mechanism and the pre-drainage cold water module, and inputs the parameters into the AI small model unit of the control module; the AI small model unit adopts a lightweight neural network, and can realize real-time sensing of the monitoring data, dynamic adjustment of the flow distribution strategy, and adjustment of the opening degree of the valve core assembly; the AI small model unit can also realize self-learning and abnormal self-diagnosis; the control module controls the pre-drainage cold water module to trigger drainage according to the temperature difference threshold, so as to drain the low-temperature wastewater remaining in the mixing cavity; the multi-sensor group built in the sensing and monitoring module is further configured in combination with a network, and is used for monitoring the water quality of each water node.

6. The control method of an intelligent thermostatic water valve based on an AI small model according to claim 5, characterized in that, The control module obtains the real-time monitoring parameters of the multi-path temperature sensing module, the flow sensor module, the execution adjustment mechanism and the pre-drainage cold water module, and inputs the parameters into the AI small model unit of the control module; the AI small model unit adopts a lightweight neural network, and can realize real-time sensing of the monitoring data, dynamic adjustment of the flow distribution strategy, and adjustment of the opening degree of the valve core assembly; the AI small model unit can also realize self-learning and abnormal self-diagnosis; the control module controls the pre-drainage cold water module to trigger drainage according to the temperature difference threshold, so as to drain the low-temperature wastewater remaining in the mixing cavity; the multi-sensor group built in the sensing and monitoring module is further configured in combination with a network, and is used for monitoring the water quality of each water node. S1, input parameter fusion: the AI small model unit receives multi-source input data, and the multi-source input data comprises a temperature difference, a water pressure difference, historical water use habit data and water quality parameters fed back by the sensing and monitoring module; S2, model training and reasoning: prestore water use scene data, extract features from the water use scene data, establish an AI small model based on LSTM or a lightweight CNN, match the closest historical scene according to the real-time input data, call the corresponding AI small model for reasoning, and output a corrected flow distribution coefficient K, i.e. the ratio of hot water to cold water, K is a variable, and K ∈ [0.3, 1.2]; S3, flow distribution execution: define a multi-dimensional input vector according to the temperature difference, pressure difference, historical water usage habit data and water quality parameters; then perform feature fusion on the multi-dimensional input vector according to a nonlinear activation function to generate a comprehensive decision vector; define a flow ratio formula according to the comprehensive decision vector; finally, map the flow ratio to a valve core displacement instruction to dynamically adjust the flow distribution strategy, and then adjust the valve core assembly opening degree.

7. The control method of an intelligent thermostatic water valve based on an AI small model according to claim 6, characterized in that, The AI small model unit can perform self-learning and abnormal self-diagnosis, including: User habit learning: by recording the temperature set value T_set and water usage time and flow adjustment operation within the user's preset time, a personalized water usage curve is generated, and the parameters of the basic segmented function f(ΔT) are automatically optimized, ΔT being the cold and hot water inlet temperature difference; Scene adaptation: when the sensing and monitoring module detects that the water quality is abnormal TDS set water quality threshold, the AI small model automatically reduces the hot water proportion, that is, reduces the K value, to ensure water quality safety; when the water pressure fluctuation ΔP is greater than the set water pressure fluctuation threshold, the threshold value of ΔT is dynamically adjusted to avoid frequent drainage; Fault self-diagnosis: (1) temperature sensor fault: if the difference between the first and second water temperature sensors exceeds the preset temperature threshold, the first water temperature sensor is marked as abnormal, switched to the standby second water temperature sensor, and a prompt is issued; (2) motor stall fault: if the micro motor still does not reach the target displacement for a preset duration, record the fault and trigger the drainage protection; (3) communication interruption fault: if the wireless connection with the sensing and monitoring module is interrupted for more than the preset connection duration, the pre-drainage cold water function is turned off and the hot water proportion K is limited to prevent high temperature scalding.

8. The control method of an intelligent thermostatic water valve based on an AI small model according to claim 5, characterized in that, The control module controls the pre-drainage cold water module to trigger drainage according to the temperature difference threshold to drain the low-temperature wastewater remaining in the mixing cavity, including: when the initial mixed water temperature T mix when the difference between the first set temperature T_set1 and the initial mixed water temperature T reaches a drain threshold AT1, the first main controller sends an open command to the drain valve; Drainage process control: the drainage rod is driven by a micro motor to realize micro displacement control through a gear set, the displacement accuracy is ±0.1 mm, and the telescopic stroke range is 0-30 mm; when the mixed water temperature T mix is not less than the second set temperature T_set2, the first main controller closes the drainage valve and stops drainage; Low-temperature protection forced drainage: if the sensing temperature of the first water temperature sensor continues to be lower than the preset low-temperature threshold during the drainage process, the first main controller controls the drainage time of the drainage valve to be extended to avoid low-temperature water remaining; Anti-clogging protection: if resistance is detected when the drainage rod is extended, the drainage is immediately stopped and a fault alarm is triggered.

9. The control method of the intelligent thermostatic water valve based on the AI small model according to claim 6, characterized in that, The control module has anti-interference and redundancy mechanisms, including: Multi-sensor redundancy verification: the difference between the first and second water temperature sensors is calculated in real time, and if the difference exceeds the redundancy threshold, a fault alarm is triggered and the standby sensor is switched to; Water pressure fluctuation compensation: the mixing cavity is provided with a pressure sensor, a flow ratio formula is corrected by a pressure signal, and a compensation coefficient Cp=1+0.01x(P hot -P cold ), wherein P hot and P cold are cold and hot water inlet pressures respectively; Motor stall protection: if the micro motor does not reach the target displacement for a preset duration, the first main controller immediately stops the micro motor power supply and records the fault; Anti-scalding: two groups of water temperature sensors are built-in in the mixed water outlet passage, and if the mixed water outlet temperature is higher than a preset a degrees Celsius, the hot water flow is controlled to be reduced, and if the mixed water outlet temperature is higher than a preset b degrees Celsius, the hot water is completely turned off to prevent scalding; wherein a<b; Power-off memory: the first controller is built-in with a memory to store the user's set target temperature T_set, drainage threshold ΔT1 and historical fault records.

10. The control method of an intelligent thermostatic water valve based on an AI small model according to claim 5, characterized in that, The sensing and monitoring module performs the following operations: M1, multi-source data acquisition: M11, Real-time collection of water quality parameters through heavy metal sensors, chemical composition detection sensors, PH value sensors, turbidity sensors, and water quality sensor groups; M12, Synchronous acquisition of environmental parameters and user behavior data, covering water use nodes through distributed deployment of sensor groups M≥1; M2, Data preprocessing and feature extraction: M21, Filtering and denoising, dynamic calibration, and normalization of water quality parameters based on edge computing terminals, generating standardized data packets; M22, Extracting key feature parameters from standardized data packets and compressing data volume through lightweight feature encoding algorithms; M3, Local anomaly detection and alarm: M31, Deploying a lightweight rule engine on the edge, making real-time judgments based on preset thresholds, and triggering local alarms; M32, If water pollution events are detected, start the drainage protection mechanism, link the pre-discharge cold water module to discharge contaminated water, and record the event timestamp and pollution parameters; M4, Cloud collaborative analysis: M41, Upload data to the cloud server through wireless transceivers, implement cross-regional water quality data sharing using a federated learning framework, and train a water pollution prediction model; M42, Cloud-based multi-dimensional data to generate water quality risk assessment reports and push to user terminals; M5, Dynamic feedback and adaptive adjustment: M51, Receive control instructions from the cloud and adjust the valve core opening through the control module to optimize the quality of mixed water; M52, When sensor data anomalies are detected, trigger the preset self-calibration program, calibrate based on standard solutions or cross-verify redundant sensors, and mark faulty nodes; M6, Historical data management and maintenance: M61, Local storage module records at least 12 months of water quality time series data, which can be searched by time axis and pollutant type; M62, Regularly perform self-checking procedures, generate maintenance logs, and upload them to the blockchain evidence platform in encrypted form to ensure data cannot be tampered with.

Images