Intelligent liquid dispensing device

Abstract

The present application relates to the field of ornamental green plant maintenance, and provides an intelligent liquid preparation device, which comprises a mixing cavity for collecting raw liquid and water to prepare liquid; a raw liquid bin for storing raw liquid and being communicated with the mixing cavity through a liquid feeder, the liquid feeder being provided with a piezoelectric atomization module; a water inlet pipeline for conveying water to the mixing cavity, the water inlet pipeline being provided with a water inlet valve; a discharge pipeline for discharging liquid in the mixing cavity, the discharge pipeline being provided with a flow meter; an interaction module for obtaining a set concentration and a set flow rate input by a user; a control module for controlling the opening degree of the water inlet valve according to the set flow rate, collecting the feedback flow rate of the flow meter, inputting the set concentration, the feedback flow rate and the raw liquid concentration into an atomization amount calculation model, calculating a target set atomization amount after local fine adjustment and global adjustment of the atomization amount calculation model, and generating corresponding piezoelectric control parameters based on the target set atomization amount to instruct the piezoelectric atomization module to perform an atomization action.

Description

Technical Field

[0001] This invention relates to the field of ornamental plant care, and more particularly to an intelligent liquid preparation device. Background Technology

[0002] Fertilization is an essential part of the daily maintenance of garden and courtyard plants. Liquid fertilizers, with their easy absorption and high utilization rate, are a commonly used type of fertilizer in the field of garden and courtyard plant maintenance. Existing integrated water and fertilizer irrigation systems can also achieve intelligent mixing and application of liquid fertilizers. However, these irrigation systems are only suitable for commercial applications, adapted to large parks and urban greenbelts, and are bulky and expensive, making them unsuitable for residential use. On the other hand, the maintenance of ornamental plants in residential courtyards and terraces is relatively more difficult and more sensitive to the concentration of liquid fertilizers, requiring higher mixing precision. Currently, there is no intelligent equipment on the market suitable for mixing and applying liquid fertilizers in such scenarios. Summary of the Invention

[0003] The main objective of this invention is to provide an intelligent liquid dispensing device that can meet the requirements of high dispensing accuracy.

[0004] To achieve the above objectives, the present invention provides an intelligent liquid preparation device, comprising:

[0005] The mixing chamber is used to collect the original solution and water to prepare the mixture;

[0006] The raw liquid chamber is used to store the raw liquid and is connected to the mixing chamber through a liquid dispenser. The liquid dispenser has a piezoelectric atomization module.

[0007] The water inlet pipe is used to supply water to the mixing chamber, and the water inlet pipe is equipped with a water inlet valve;

[0008] The discharge pipe is used to discharge the liquid material in the mixing chamber, and the discharge pipe is equipped with a flow meter;

[0009] The interactive module is used to obtain the user's input of the set concentration and set flow rate;

[0010] The control module controls the opening of the inlet valve based on the set flow rate and collects the feedback flow rate from the flow meter. It inputs the set concentration, feedback flow rate, and stock solution concentration into the atomization volume calculation model. After local fine-tuning and global adjustment of the atomization volume calculation model, the target set atomization volume is calculated. Based on the target set atomization volume, corresponding piezoelectric control parameters are generated to instruct the piezoelectric atomization module to perform atomization actions. By adding a conductivity sensor, pH sensor, and camera, a multimodal sensing system is constructed. A convolutional neural network is used to identify plant varieties and growth status, and the optimal nutrient solution concentration is inferred in real time by combining sensor data from the conductivity and pH sensors. Growth status includes leaf color and lesions.

[0011] In some embodiments, the set concentration, feedback flow rate, and original liquid concentration are input into the atomization volume calculation model. After local fine-tuning and global adjustment of the atomization volume calculation model, the target set atomization volume is calculated, including: normalizing the set concentration, feedback flow rate, and original liquid concentration according to the input layer of the preset atomization volume calculation model to obtain normalized numerical variables corresponding to the set concentration, feedback flow rate, and original liquid concentration; performing feature processing on the normalized numerical variables according to the intermediate layer of the preset atomization volume calculation model to obtain a feature vector; mapping the feature vector to an initial set atomization volume according to the output layer of the preset atomization volume calculation model; performing local fine-tuning on the output layer of the preset atomization volume calculation model based on the initial set atomization volume according to a pre-constructed reward function, and remapping the feature vector according to the locally fine-tuned output layer to obtain a candidate set atomization volume; and performing global adjustment on the output layer of the preset atomization volume calculation model based on the locally fine-tuned values ​​according to a pre-constructed federated learning algorithm, and remapping the feature vector according to the globally fine-tuned output layer to obtain the target set atomization volume.

[0012] In some embodiments, based on a pre-constructed reward function, the output layer of a preset atomization amount calculation model is locally fine-tuned according to the initial set atomization amount. The feature vector is then remapped based on the locally fine-tuned output layer to obtain a candidate set atomization amount. This includes: evaluating the effect of the initial atomization amount using a pre-constructed reward function and calculating the gradient feedback of the reward function with respect to the initial atomization amount to obtain the direction for improving the current output quality; based on the gradient feedback of the reward function, locally fine-tuning the weights and bias parameters of the output layer using a local learning rate, the purpose of which is to quickly improve the direction of the current output without affecting the global parameters of the entire model; and inputting the feature vector into the locally fine-tuned output layer and performing forward propagation again to obtain the updated candidate set atomization amount.

[0013] In some embodiments, based on a pre-built federated learning algorithm, the output layer of a preset atomization calculation model is globally adjusted based on local fine-tuning values. The locally fine-tuned output layer then remaps the feature vector to obtain the target atomization amount. This includes: each participating device performing a local fine-tuning operation of the output layer based on a reward function locally and calculating its local parameter update amount relative to the original output layer parameters; all clients uploading their respective local parameter update amounts to the server; the server aggregating the local parameter update amounts of all clients using a federated averaging algorithm to generate a set of globally updated output layer global update parameters, enabling the model to integrate the optimization experience of multiple devices while maintaining privacy; the server-updated global output layer parameters are used to recalculate the feature vector to output the target atomization amount.

[0014] In some embodiments, when the server aggregates the local parameter update amounts of all clients using a federated averaging algorithm, it is further configured to perform the following steps: obtain the local parameter update amount uploaded by each client and its corresponding device runtime data; calculate the weight coefficient of each client based on the device runtime data, wherein the client with a longer runtime corresponds to a higher weight coefficient; perform a weighted average of the local parameter update amounts of each client according to the weight coefficient to generate the globally updated output layer global update parameters; before distributing the global update parameters to each client, encrypt the global update parameters, wherein the encryption process includes using a homomorphic encryption algorithm to perform layered encryption of the parameters, so that each client can only obtain the parameter portion that matches the local model structure when decrypting.

[0015] In some embodiments, generating corresponding piezoelectric control parameters based on the target atomization amount includes: acquiring a parameter set reflecting the atomization characteristics of the original liquid at the operating temperature; performing function fitting between the parameter set and the piezoelectric control parameters and the atomization amount to establish an atomization parameter control model; and calculating the set piezoelectric control parameters corresponding to the target atomization amount according to the atomization parameter control model, wherein the piezoelectric control parameters include a driving voltage, and the parameter set includes the viscosity and surface tension of the original liquid.

[0016] In some embodiments, the system further includes: a concentration detection sensor configured in the discharge pipeline; the control module also collects the feedback concentration from the concentration detection sensor, calculates the deviation between the feedback concentration and the set concentration and compares it with the set range; when the deviation exceeds the set range, the control module issues an abnormality warning message for the raw liquid through an interactive module; the concentration detection sensor includes a conductivity sensor.

[0017] In some embodiments, the discharge pipe has a discharge direction, the mixing chamber has an upper end face located upstream of the discharge direction, a lower end face located downstream of the discharge direction, and a side wall connecting the upper end face and the lower end face. The projected area of ​​the upper end face along the discharge direction is larger than the projected area of ​​the lower end face along the discharge direction. The water inlet pipe is connected to the mixing chamber at the upper end face, the liquid dispenser is connected to the mixing chamber at the side wall, and the discharge pipe is connected to the mixing chamber at the lower end face.

[0018] In some embodiments, the flow meter includes an electromagnetically coupled rotating body and a detection module. The rotating body is placed inside the discharge pipe and can rotate along a plane perpendicular to the discharge direction. The detection module includes a Hall sensor.

[0019] In some embodiments, the rotating body includes a driving member and a driven member configured with a magnet. The driving member includes a rotating shaft portion parallel to the discharge direction and a plurality of fan blade portions surrounding the outside of the rotating shaft portion. The fan blade portions have concave arc surfaces intersecting the discharge direction. The driven member is connected to the discharge direction end of the rotating shaft portion. The flow meter also includes a guide plate located on the side of the driving member opposite to the discharge direction. The guide plate is provided with a plurality of guide holes for guiding the liquid material to contact the concave arc surface.

[0020] As can be seen from the above scheme, the aforementioned mixing device uses a piezoelectric atomization module for liquid mixing, which helps to accurately control the liquid concentration and improve the mixing efficiency. The control module can control the dilution ratio of the stock solution, i.e., the concentration of the liquid, by adjusting the opening of the water inlet valve and the atomization volume. The feedback flow rate collected by the flow meter can reflect the actual flow rate of the liquid. The control module calculates the target atomization volume based on the feedback flow rate and the set concentration and set flow rate collected by the interactive module, thereby accurately controlling the liquid concentration. Furthermore, based on the target atomization volume, the corresponding piezoelectric control parameters can be calculated to precisely control the atomization action.

[0021] Addressing the high sensitivity of ornamental plants in residential courtyards and terraces to liquid fertilizer concentration, the system utilizes a misting calculation model (local fine-tuning + global adjustment) in the control module, combined with the micron-level misting precision of the piezoelectric misting module, to achieve a concentration control error within ±0.3%. This avoids root burn or malnutrition caused by concentration deviations, significantly improving the safety of home plant care.

[0022] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and do not limit this application. Attached Figure Description

[0023] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0024] To gain a more complete understanding of this application and its beneficial effects, the following description will be provided in conjunction with the accompanying drawings, wherein the same reference numerals in the following description denote the same parts.

[0025] Figure 1 This is an external structural diagram of an embodiment of the intelligent liquid preparation device of the present invention.

[0026] Figure 2 This is an internal structural diagram of an embodiment of the intelligent liquid preparation device of the present invention.

[0027] Figure 3 yes Figure 2 Cross-sectional structural diagram.

[0028] Figure 4 yes Figure 3 Enlarged structural diagram at point A in the middle.

[0029] Figure 5 This is a partial structural diagram of the flow meter in an embodiment of the intelligent liquid preparation device of the present invention.

[0030] Figure 6 yes Figure 5 The structure explodes diagram.

[0031] Reference numerals: 1. Mixing chamber; 11. Upper end face; 12. Lower end face; 13. Side wall; 2. Raw material tank; 3. Adder; 31. Piezoelectric atomization module; 4. Water inlet pipe; 41. Water inlet valve; 5. Discharge pipe; 6. Flow meter; 61. Rotating body; 611. Driving component; 611a. Rotating shaft; 611b. Fan blade; 611c. Concave arc surface; 612. Driven component; 612a. Magnet; 62. Detection module; 63. Guide plate; 631. Guide hole; 64. Support; 7. Interactive module; 71. Display screen; 72. Knob; 73. Button; 8. Solar cell module; 9. Concentration detection sensor.

[0032] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and do not limit this application. Detailed Implementation

[0033] The preferred embodiments of this application will now be described in detail with reference to the accompanying drawings. In the following description, the same reference numerals are used for the same components, and repeated descriptions are omitted. Furthermore, the drawings are merely schematic diagrams, and the proportions of the components or their shapes may differ from actual dimensions. It should be noted that all directional indications (such as up, down, left, right, front, back, etc.) in the embodiments of this application are only used to explain the relative positional relationships and movement of the components in a specific posture. If the specific posture changes, the directional indications will also change accordingly.

[0034] It should also be noted that when a component is described as "fixed to" or "set on" another component, it can be directly on the other component or there may be an intervening component present. When a component is described as "connected to" another component, it can be directly connected to the other component or there may be an intervening component present.

[0035] Fertilization is an essential part of the daily maintenance of garden and courtyard plants. Liquid fertilizers, with their easy absorption and high utilization rate, are a commonly used type of fertilizer in the field of garden and courtyard plant maintenance. Existing integrated water and fertilizer irrigation systems can also achieve intelligent mixing and application of liquid fertilizers. However, these irrigation systems are only suitable for commercial applications, adapted to large parks and urban greenbelts, and are bulky and expensive, making them unsuitable for residential use. On the other hand, the maintenance of ornamental plants in residential courtyards and terraces is relatively more difficult and more sensitive to the concentration of liquid fertilizers, requiring higher mixing precision. Currently, there is no intelligent equipment on the market suitable for mixing and applying liquid fertilizers in such scenarios.

[0036] See Figure 1-6 The intelligent liquid mixing device provided by this invention includes a mixing chamber 1 for collecting stock solution and water to make a mixing solution; a stock solution tank 2 for storing stock solution and communicating with the mixing chamber 1 through a liquid dispenser 3, the liquid dispenser 3 having a piezoelectric atomization module 31; a water inlet pipe 4 for supplying water to the mixing chamber 1, the water inlet pipe 4 being equipped with a water inlet valve 41; a discharge pipe 5 for discharging the liquid in the mixing chamber 1, the discharge pipe 5 being equipped with a flow meter 6; an interactive module 7 for acquiring the set concentration and set flow rate input by the user; and a control module for controlling the opening degree of the water inlet valve 41 according to the set flow rate and collecting the feedback flow from the flow meter 6. The system calculates the target atomization volume by inputting the set concentration, feedback flow rate, and stock solution concentration into the atomization volume calculation model. After local fine-tuning and global adjustment of the atomization volume calculation model, the target atomization volume is calculated. Based on the target atomization volume, corresponding piezoelectric control parameters are generated to instruct the piezoelectric atomization module to perform atomization actions. By adding a conductivity sensor, pH sensor, and camera, a multimodal sensing system is constructed. A convolutional neural network is used to identify plant varieties and growth status, and the current optimal nutrient solution concentration is inferred in real time by combining the sensor data corresponding to the conductivity sensor and pH sensor. Growth status includes leaf color and lesions.

[0037] The mixing device uses a piezoelectric atomization module 31 to atomize the stock solution before feeding it into the mixing chamber 1 to mix with the transport water. This helps to precisely control the concentration of the stock solution and improve the preparation efficiency. The control module can control the dilution ratio of the stock solution, i.e., the concentration of the stock solution, by adjusting the opening of the inlet valve 41 and the atomization amount. The feedback flow rate collected by the flow meter 6 reflects the actual flow rate of the stock solution. The control module calculates the target atomization amount based on the feedback flow rate and the set concentration and set flow rate collected by the interaction module 7, using the atomization amount calculation model, thereby accurately controlling the concentration of the stock solution. Furthermore, based on the target atomization amount, the corresponding piezoelectric control parameters can be calculated to precisely control the atomization action.

[0038] The intelligent liquid dispensing device provided by this invention is an electromechanical integrated solution, featuring a housing. The display screen 71, knob 72, and button 73, which are components of the interactive module 7, are all disposed on the housing. In addition, the interface portions of the water inlet pipe 4 and the discharge pipe 5 are also exposed outside the housing. The water inlet pipe 4 can be directly connected to a faucet, and the discharge pipe 5 can be connected to a garden sprinkler. The raw liquid tank 2 is disposed on the upper side of the housing. The main bodies of the water inlet pipe 4 and the discharge pipe 5, as well as the water inlet valve 41, the liquid dispenser 3, and the flow meter 6, are all disposed inside the housing.

[0039] The discharge pipe 5 has a discharge direction D1. The mixing chamber 1 has an upper end face 11 located upstream of the discharge direction D1, a lower end face 12 located downstream of the discharge direction D1, and a side wall 13 connecting the upper end face 11 and the lower end face 12. The projected area of ​​the upper end face 11 along the discharge direction D1 is larger than the projected area of ​​the lower end face 12 along the discharge direction D1. The water inlet pipe 4 is connected to the mixing chamber 1 at the upper end face 11. The liquid dispenser 3 is connected to the mixing chamber 1 at the side wall 13. The discharge pipe 5 is connected to the mixing chamber 1 at the lower end face 12.

[0040] The mixing chamber 1 is roughly funnel-shaped. The water inlet pipe 4 and the discharge pipe 5 are connected to the upper and lower ends of the mixing chamber 1, respectively. The connection between the water inlet pipe 4 and the mixing chamber 1 is located upstream of the connection between the liquid dispenser 3 and the mixing chamber 1. Therefore, turbulence can be used to improve the mixing efficiency of the original liquid and water and improve the uniformity of the concentration of the liquid.

[0041] The flow meter 6 includes an electromagnetically coupled rotating body 61 and a detection module 62. The rotating body 61 is placed inside the discharge pipe 5 and can rotate along a plane perpendicular to the discharge direction D1. The detection module 62 includes a Hall sensor.

[0042] The flow meter 6 is a turbine flow meter with high detection accuracy. The rotating body 61 placed in the discharge pipe 5 can promote the mixing of the raw liquid and water, thus playing a certain auxiliary role in the preparation of the liquid.

[0043] The rotating body 61 includes a driving member 611 and a driven member 612 with a magnet 612a. The driving member 611 includes a rotating shaft 611a parallel to the discharge direction D1 and a plurality of fan blades 611b surrounding the outside of the rotating shaft 611a. The fan blades 611b have concave arc surfaces 611c intersecting the discharge direction D1. The driven member 612 is connected to the discharge direction D1 end of the rotating shaft 611a. The flow meter 6 also includes a guide plate 63 located on the side of the driving member 611 opposite to the discharge direction D1. The guide plate 63 is provided with a plurality of guide holes 631 for guiding the liquid to contact the concave arc surface 611c.

[0044] The guide plate with multiple guide holes 631 can guide the liquid to fully interact with the driving component 611, enabling the liquid to drive the rotating body 61 better and allowing the rotational speed of the rotating body 61 to better match the flow rate of the liquid, thereby helping to improve the detection accuracy of the flow meter 6. The cooperation between the guide plate 63 and the rotating body 61 can also promote the mixing of the raw liquid and water, thereby helping to improve the concentration uniformity of the liquid.

[0045] In addition, the flow meter also includes a support portion 64, which includes a horizontal bar and vertical bars located at both ends of the horizontal bar. The support portion 64 is connected to the guide plate 63 via the vertical bars. The lower end of the rotating shaft portion 611a passes through the driven member 612 and is connected to the horizontal bar portion of the support portion 64, while the upper end of the rotating shaft portion 611a is connected to the guide plate 63. Two magnets 612a are symmetrically arranged in the two grooves of the driven member.

[0046] The intelligent liquid preparation device also includes a power supply module, which includes a solar cell module 8. Considering that outdoor environments such as courtyards and terraces have ample sunlight but are inconvenient to access electricity, the solar cell module 8 can serve as a supplement to the conventional power supply method of the intelligent liquid preparation device. When the power is large enough, the solar cell module 8 can also serve as the main power supply method for the intelligent liquid preparation device. In this embodiment, the solar cell module 8 is disposed on the top surface of the casing.

[0047] The intelligent liquid preparation device also includes a concentration detection sensor 9 configured in the discharge pipeline 5. The control module also collects the feedback concentration from the concentration detection sensor 9, calculates the deviation between the feedback concentration and the set concentration, and compares it with the set range. When the deviation exceeds the set range, the control module issues an abnormality warning message for the original liquid through the interactive module 7. The concentration detection sensor includes a conductivity sensor.

[0048] The raw liquid may deteriorate due to its own quality issues or improper storage, resulting in partial degradation and inactivation of its fertilizing components. This manifests as a significant deviation between the actual concentration of fertilizing components in the liquid and the set concentration. Installing a concentration detection sensor 9 in the discharge pipeline 5 introduces a feedback mechanism for the liquid concentration in the intelligent liquid mixing device. When the deviation between the feedback concentration and the set concentration exceeds the set range, the control module can determine that the raw liquid has deteriorated and prompt the user to replace the liquid via the interactive module 7, thus effectively addressing liquid problems caused by raw liquid deterioration. Considering that most of the fertilizing components in the raw liquid are ionic compounds, a conductivity sensor can be selected as the concentration detection sensor 9.

[0049] A multimodal sensing system was constructed by adding an EC (electrical conductivity) sensor, a pH sensor, and a camera to the device. A convolutional neural network (CNN) was used to identify plant varieties and growth status (such as leaf color and lesions), and the optimal nutrient solution concentration was inferred in real time by combining sensor data, forming a closed-loop self-learning system of "perception-decision-execution".

[0050] Multimodal data acquisition uses EC / pH sensors to monitor the conductivity (reflecting ion concentration) and pH of the effluent from the mixing chamber in real time, while cameras periodically capture images of plant leaves.

[0051] The data fusion model includes: designing a multi-branch neural network: image branch: CNN extracts leaf features (such as chlorophyll content and leaf area); sensor branch: fully connected layer processes EC / pH / flow data; fusion layer: weighted fusion of two types of features through an attention mechanism, outputting a recommended concentration C recommendation.

[0052] The self-learning mechanism records the plant's growth feedback a week after each fertilization (by comparing changes in leaf state through images), and uses backpropagation to update the fusion model parameters, forming a closed loop of "fertilization effect - model optimization".

[0053] No user input of plant varieties is required; the device autonomously identifies and matches the optimal concentration (e.g., automatically increasing iron fertilizer concentration when rose leaves turn yellow), lowering the barrier to entry and making it suitable for gardening beginners. It dynamically adjusts the formula by combining physiological indicators (EC value reflecting ion concentration) and phenotypic characteristics (leaf condition), overcoming the blind application of traditional "fixed concentration + experience-based fertilization" and promoting healthy plant growth (e.g., reducing root burn caused by improper concentration). By continuously learning from plant care data from different users (within the context of federated learning to protect privacy), the model can identify rare varieties and generate suitable solutions, building an ever-evolving plant nutrition knowledge base.

[0054] In some embodiments, generating corresponding piezoelectric control parameters based on a target atomization amount includes: acquiring a parameter set reflecting the atomization characteristics of the original liquid at the operating temperature; performing function fitting between the parameter set and the piezoelectric control parameters and the atomization amount to establish an atomization parameter control model; and calculating the set piezoelectric control parameters corresponding to the target atomization amount based on the atomization parameter control model. The piezoelectric control parameters include a driving voltage, and the parameter set includes the viscosity and surface tension of the original liquid.

[0055] The atomization volume of a piezoelectric atomization module is related to both the piezoelectric control parameters and the specific physicochemical parameters of the stock solution at the operating temperature. For details, please refer to the following formula: Q=(C*V 2 *D*A / γ) / (f*n) 0.5 Where Q is the piezoelectric atomization amount, V is the driving voltage, D is the diameter of the piezoelectric element, A is the nozzle area, f is the vibration frequency, η is the viscosity of the original liquid, γ is the surface tension of the original liquid, and C is the equipment characteristic constant.

[0056] To accurately control the atomization quantity Q, an atomization parameter control model needs to be established based on piezoelectric control parameters and the specific physicochemical parameters of the concentrate at the operating temperature. Referring to the above formula, the driving voltage V is used as the piezoelectric control parameter, and the viscosity η and surface tension γ of the concentrate are used as its physicochemical parameters. The piezoelectric element diameter D, nozzle area A, and device characteristic constant C can be combined into a single constant. The physicochemical parameters of the concentrate, as a dataset, are related to the type, concentration, and operating temperature of the concentrate. The amount of relevant data required for function fitting is substantial. This data can be stored locally in a relevant database or data model, or retrieved from the cloud as needed.

[0057] In some embodiments, the set concentration, feedback flow rate, and original liquid concentration are input into the atomization volume calculation model. After local fine-tuning and global adjustment of the atomization volume calculation model, the target set atomization volume is calculated, including: normalizing the set concentration, feedback flow rate, and original liquid concentration according to the input layer of the preset atomization volume calculation model to obtain normalized numerical variables corresponding to the set concentration, feedback flow rate, and original liquid concentration; performing feature processing on the normalized numerical variables according to the intermediate layer of the preset atomization volume calculation model to obtain a feature vector; mapping the feature vector to an initial set atomization volume according to the output layer of the preset atomization volume calculation model; performing local fine-tuning on the output layer of the preset atomization volume calculation model based on the initial set atomization volume according to a pre-constructed reward function, and remapping the feature vector according to the locally fine-tuned output layer to obtain a candidate set atomization volume; and performing global adjustment on the output layer of the preset atomization volume calculation model based on the locally fine-tuned values ​​according to a pre-constructed federated learning algorithm, and remapping the feature vector according to the globally fine-tuned output layer to obtain the target set atomization volume.

[0058] The core processing flow of the atomization volume calculation model includes input layer normalization, intermediate layer feature processing, output layer initial mapping, and local fine-tuning based on the reward function and global adjustment using a federated learning algorithm. Input layer processing normalizes the set concentration, feedback flow rate, and original solution concentration, converting them into numerical variables to fit the model input. Intermediate layer feature processing extracts and processes features from the normalized variables using pre-defined intermediate layers (such as hidden layers in a neural network), generating feature vectors. Output layer initial mapping maps the feature vectors to the initial set atomization volume. Local fine-tuning evaluates the effect of the initial atomization volume based on the reward function, adjusts the output layer parameters, and generates candidate set atomization volumes. Global adjustment aggregates local adjustment experience from multiple devices using a federated learning algorithm, optimizes the output layer parameters, and finally obtains the target set atomization volume.

[0059] The normalization process employs minimum-maximum normalization or Z-score normalization methods to convert the set concentration (range 0-100%), feedback flow rate (unit L / min), and stock solution concentration (percentage) into numerical variables within the range of [0,1] or [-1,1], thus eliminating the influence of dimensions.

[0060] Feature processing can be designed as a multilayer perceptron (MLP) or convolutional neural network (CNN) through intermediate layers. Nonlinear features are extracted through weighted summation and activation functions (such as ReLU) to generate feature vectors containing the concentration-flow correlation.

[0061] The initial mapping involves a linear output layer, where the feature vector is mapped to the initial atomization level using a weight matrix. The formula is: Initial atomization level = Feature vector × W + b; where W is the weight and b is the bias. Local fine-tuning is achieved by defining the reward function as a negative function of the concentration error (e.g., ...). Where R is the value of the reward function, C is the set concentration, such as the target liquid concentration input through the interactive module (such as the display screen, knob, etc.), representing the concentration value of the nutrient solution or preparation solution expected by the user (usually expressed as a percentage or conductivity, etc.). C is actually the actual concentration of the liquid detected in real time by the concentration detection sensor (such as the conductivity sensor) in the device, reflecting the true concentration value of the current preparation result. It is collected in real time by the concentration detection sensor in the discharge pipeline and directly feeds back the actual ion concentration or other concentration indicators of the liquid discharged from the current mixing chamber. The output layer parameters are adjusted by gradient descent to quickly optimize the atomization calculation accuracy of the current device.

[0062] Global adjustments are made by uploading the differences in parameters after local adjustments from each device. The server then aggregates data from multiple devices through federated learning to update the global model parameters, thereby improving the universality across devices.

[0063] By normalizing input parameters of different dimensions, the model ensures consistent processing capabilities for data such as concentration and flow rate. Feature processing in the intermediate layers uncovers the nonlinear correlation between concentration and flow rate, improving the theoretical accuracy of atomization calculations. Combining local fine-tuning and global adjustment satisfies the real-time optimization needs of single devices while leveraging multi-device data to enhance the model's generalization ability, avoiding overfitting or underfitting. Through reward functions and federated learning, the model can dynamically adjust to the high-precision requirements of small scenarios such as residential courtyards, solving the "one-size-fits-all" adaptation problem of traditional devices.

[0064] In some embodiments, based on a pre-constructed reward function, the output layer of a preset atomization amount calculation model is locally fine-tuned according to the initial set atomization amount. The feature vector is then remapped based on the locally fine-tuned output layer to obtain a candidate set atomization amount. This includes: evaluating the effect of the initial atomization amount using a pre-constructed reward function and calculating the gradient feedback of the reward function with respect to the initial atomization amount to obtain the direction for improving the current output quality; based on the gradient feedback of the reward function, locally fine-tuning the weights and bias parameters of the output layer using a local learning rate, the purpose of which is to quickly improve the direction of the current output without affecting the global parameters of the entire model; and inputting the feature vector into the locally fine-tuned output layer and performing forward propagation again to obtain the updated candidate set atomization amount.

[0065] The initial atomization level is evaluated using a reward function, and gradient feedback is calculated to adjust the weights and bias parameters of the output layer, rapidly optimizing the atomization level calculation accuracy of the current device. The reward function evaluation quantifies the concentration error corresponding to the initial atomization level, generating gradient feedback to indicate the direction of parameter adjustment. Local parameter adjustment uses a learning rate independent of the global model, adjusting only the output layer parameters to avoid affecting the general feature extraction capabilities of intermediate layers. Forward propagation update recalculates the candidate atomization level based on the adjusted output layer parameters, achieving real-time optimization. The reward function is designed as a loss function defined as the square of the concentration error. Where R is the value of the reward function, C is the set concentration, such as the target concentration of the nutrient solution input through the interactive module (such as the display screen, knob, etc.), which represents the concentration value of the nutrient solution or preparation solution expected by the user (usually expressed as a percentage or conductivity). C is actually the actual concentration of the nutrient solution detected in real time by the concentration detection sensor (such as the conductivity sensor) in the device, reflecting the true concentration value of the current preparation result. It is collected in real time by the concentration detection sensor in the discharge pipeline and directly feeds back the actual ion concentration or other concentration indicators of the current mixing chamber discharge solution.

[0066] in (Assuming the atomization rate is much smaller than the water flow rate, it can be simplified to) ).

[0067] Gradient calculation involves computing the gradient of the reward function with respect to the output layer weights W and bias b. ;

[0068] ;

[0069] The updated formula is:

[0070] ;

[0071] Where αlocal is the learning rate independent of the global model, and is usually set to a large value to accelerate local optimization; R is the value of the reward function. W R is the gradient of the reward function with respect to the output layer weights W, W′ is the updated output layer weight matrix used to calculate the fogging amount in the next forward propagation, and W is the output layer weight matrix of the current iteration (its value before the update). α is local. W R represents the direction and magnitude of the weight update, adjusted in the opposite direction of the reward function gradient (gradient descent) to minimize the concentration error (maximize the reward R). b′ is the updated output layer bias vector. b is the output layer bias vector of the current iteration (its value before the update). Similar to the weight update, the bias adjustment compensates for the initial atomization deviation, quickly correcting the concentration error. b R is the gradient of the reward function with respect to the output layer bias vector b of the current iteration.

[0072] Candidate atomization amount generation involves inputting the feature vector into the updated output layer and calculating the candidate atomization amount through forward propagation.

[0073] .

[0074] By directly linking concentration error to the reward function, the device can quickly adjust the atomization volume for the current irrigation scenario, solving the concentration sensitivity problem in small-batch fertilization (such as the poor tolerance of terrace plants to high-concentration fertilizers). Only the output layer parameters are adjusted, avoiding retraining the entire model, reducing computational load and energy consumption, and adapting to the low computing power requirements of civilian equipment. Each device can be optimized independently to adapt to the differentiated concentration requirements of different users' plant varieties (such as succulents and flowers), improving the dispensing accuracy of a single device.

[0075] In some embodiments, based on a pre-built federated learning algorithm, the output layer of a preset atomization calculation model is globally adjusted based on local fine-tuning values. The locally fine-tuned output layer then remaps the feature vector to obtain the target atomization amount. This includes: each participating device performing a local fine-tuning operation of the output layer based on a reward function locally and calculating its local parameter update amount relative to the original output layer parameters; all clients uploading their respective local parameter update amounts to the server; the server aggregating the local parameter update amounts of all clients using a federated averaging algorithm to generate a set of globally updated output layer global update parameters, enabling the model to integrate the optimization experience of multiple devices while maintaining privacy; the server-updated global output layer parameters are used to recalculate the feature vector to output the target atomization amount.

[0076] By leveraging federated learning algorithms to aggregate local parameter updates from multiple devices, global model parameters are optimized while protecting user privacy, enabling the atomization calculation model to incorporate maintenance experience from different scenarios.

[0077] Local parameter updates are performed locally on each device to fine-tune the output layer and calculate the parameter update amount (e.g., ...). , W is the initial weight matrix of the output layer of the preset atomization calculation model (initial parameters issued by the global model or parameters updated in the previous round of global updates). It serves as the basic parameter for the linear mapping of the output layer, mapping the feature vectors generated by the intermediate layers to the initially set atomization amount. b is the initial bias vector of the output layer, which, together with the weight matrix W, determines the linear mapping result of the initial atomization amount. W′ is the output layer weight matrix adjusted locally by the device using gradient descent with the reward function. b′ is the locally fine-tuned output layer bias vector. ΔW is the update increment of the weight matrix, equal to the difference between the fine-tuned weights and the original weights. Δb is the update increment of the bias vector, equal to the difference between the fine-tuned bias and the original bias. Parameter aggregation collects the local update amounts from all devices through the server and generates global update parameters using the Federated Avg algorithm. Global model updates are achieved by distributing global parameters to each device, updating the output layer weights and biases, and improving the model's cross-device generalization ability.

[0078] Local update calculations are performed on each device after local fine-tuning, recording the changes in output layer parameters ΔWi, Δbi (where i is the device number). The federated averaging algorithm calculates global update parameters via the server.

[0079] Where N is the number of participating devices.

[0080] The privacy protection mechanism prevents the leakage of user planting data (such as plant varieties and fertilization frequency) by having devices upload only the parameter update amount, rather than the original data or the complete model. Global parameter distribution involves each device receiving the globally updated parameters and then updating its local output layer.

[0081] Federated learning avoids uploading user data to servers, exchanging only parameter updates, thus meeting the privacy and security needs of civilian devices (e.g., preventing the leakage of home garden data). It aggregates the maintenance experience of multiple users (e.g., fertilization habits for different regional climates and plant types), enabling the model to adapt to more complex scenarios and improving global allocation accuracy. It eliminates the need for centralized data training, continuously optimizing the model through real-time feedback from distributed devices, solving the problems of traditional smart devices relying on fixed algorithms and being difficult to update.

[0082] In some embodiments, when the server aggregates the local parameter update amounts of all clients using a federated averaging algorithm, it is further configured to perform the following steps: obtain the local parameter update amount uploaded by each client and its corresponding device runtime data; calculate the weight coefficient of each client based on the device runtime data, wherein the client with a longer runtime corresponds to a higher weight coefficient; perform a weighted average of the local parameter update amounts of each client according to the weight coefficient to generate the globally updated output layer global update parameters; before distributing the global update parameters to each client, encrypt the global update parameters, wherein the encryption process includes using a homomorphic encryption algorithm to perform layered encryption of the parameters, so that each client can only obtain the parameter portion that matches the local model structure when decrypting.

[0083] By incorporating device runtime as a weighting factor into the federated averaging algorithm and adding homomorphic encryption, the rationality of weight allocation and the security of parameter transmission are ensured. Weighting factor calculation allocates weights based on device runtime (reflecting the amount of data accumulated), with devices running longer contributing more to the global model. Weighted parameter aggregation uses the weighting factor to perform a weighted average of local update amounts, increasing the influence of data-rich devices. Layered encryption ensures that devices can only decrypt the parameter portions matching their local model by applying homomorphic encryption to global update parameters, preventing information leakage.

[0084] The weighting coefficient is calculated by defining the runtime of device i as Ti, and the total runtime as Ti. The weighting coefficients are: ωi = Ti / Ttotal. The weighted average aggregation is performed by globally updating the parameters to a weighted sum:

[0085] ;

[0086] Homomorphic encryption uses homomorphic encryption algorithms (such as BFV and CKKS) on the server to encrypt global parameters in layers. Each device only has the private key to decrypt the corresponding layer of the local model (such as the output layer) and cannot obtain the parameters of other layers or update information from other devices.

[0087] After receiving the encrypted global parameters through the device, the parameter distribution and decryption process only decrypts the weights and biases of the output layer, ensuring the privacy of the model structure and parameters.

[0088] Weighting devices based on runtime allows those accumulating more valuable data (such as yards with consistently applied fertilizer) to contribute more significantly to model optimization, improving the overall model's reliability and robustness. Homomorphic encryption ensures that model details and update information from user devices are not leaked during parameter transmission, meeting the high data security requirements of civilian scenarios (e.g., preventing malicious tampering with model parameters). This incentive mechanism encourages long-term device use, creating a positive cycle of "data contribution - model optimization - experience improvement," thus enhancing user engagement.

[0089] In some embodiments, by modeling the liquid preparation process as a Markov decision process (MDP), a reinforcement learning agent (RL agent) is designed to simultaneously optimize three objectives: concentration accuracy, application rate, and concentrate consumption efficiency. Through deep Q-networks (DQN) or policy gradient algorithms, the device autonomously selects the optimal preparation strategy for different plant types (e.g., succulents require low-concentration, slow application, while shrubs require high-concentration, rapid application).

[0090] The state space definition includes: state S = {C_setup, C_real-time, Q_feedback, V_remaining concentrate, T_ambient temperature}, which correspond to set concentration, real-time concentration, flow rate, remaining concentrate, and ambient temperature, respectively.

[0091] The motion space design includes: motion A = {Δ valve opening, Δ atomization frequency}, with discretized step sizes (such as ±5% opening and ±10% atomization frequency) as optional motions.

[0092] The reward function construction includes: multi-objective reward: R = α( |C Error| + β (fertilization speed) + γ (1 / concentrate consumption); where α / β / γ are weighting coefficients (users can customize them through the interactive module, such as α=0.6, β=0.2, γ=0.2 in succulent mode). C error is the value of the set concentration and the real-time detected concentration. The smaller the error, the higher the reward. Fertilization speed is the effective amount of fertilizer applied per unit time. It is the reciprocal of the concentrate consumption, reflecting the effective utilization rate of the concentrate per unit (the less the consumption, the higher the reward). R is the value of the reward function.

[0093] Training and execution include: offline phase: pre-training the DQN network in a simulation environment using historical fertilization data; online phase: balancing exploration and utilization through the Epsilon-Greedy policy and updating the Q-value table in real time.

[0094] This system automatically generates optimal formulation strategies to address the fertilization sensitivities of different plants (e.g., orchids are sensitive to concentrated fertilizers and require low-rate, long-term application), avoiding plant damage caused by manual trial and error. While ensuring concentration accuracy, it improves consumable utilization by reducing waste of concentrate (e.g., prioritizing the use of the current batch when only 50ml of concentrate remains) and accelerating fertilization speed (for non-sensitive plants). The system also incorporates strategies adjusted based on ambient temperature (adjusting strategies to account for changes in plant absorption efficiency at high temperatures), overcoming the shortcomings of traditional algorithms that ignore environmental variables, making it particularly suitable for the seasonal maintenance needs of open-air gardens.

[0095] In some embodiments, a model initialization method based on transfer learning is designed to address the data scarcity problem in new user scenarios. Using a cloud-pre-trained "general plant fertilization model" as a foundation, the output layer parameters are fine-tuned with a small amount of target user data (such as 3-5 fertilization records) to quickly adapt to different garden plant species and soil conditions, solving the problem of poor accuracy during the "cold start" of traditional models.

[0096] The pre-trained model is built by training a general model in the cloud using tens of millions of plant care data (covering a variety of common varieties). It includes an input layer (concentration / flow rate / original solution parameters), an intermediate layer (deep neural network to extract general fertilization rules), and an output layer (prediction of atomization amount).

[0097] The transfer learning process includes: Feature transfer: retaining the input and intermediate layer parameters of the pre-trained model (freezing weights), and only replacing the output layer with a randomly initialized fully connected layer; Fine-tuning training: using a small amount of data from new users (such as plant types and successful concentration cases manually entered by users during the first fertilization), the output layer is fine-tuned using the mean squared error loss function. The compressed size of the pre-trained model is ≤10MB, which is suitable for low-computing-power MCUs (such as STM32) in civilian devices. The fine-tuning process is completed locally without continuous network connection.

[0098] When new users use the device for the first time, it achieves an initial concentration accuracy of over 80% without lengthy training, making it particularly suitable for the rapid maintenance of potted plants purchased on short notice. Adaptation is completed with only 3-5 successful fertilization records, addressing the data sparsity issue in civilian scenarios (families typically only maintain 5-10 types of plants), thus reducing user involvement costs. The pre-trained model covers common fertilization patterns for common plants (e.g., foliage plants prefer nitrogen, flowering plants prefer phosphorus and potassium), and after transfer, the model can quickly adapt to unlisted varieties (e.g., inferring concentration requirements through similar family and genus characteristics).

[0099] In some embodiments, a fuzzy logic controller is designed to address the issue of fuzzy concentration descriptions (such as "a little lighter" or "a little less than last time") by non-professional users. The user's natural language input is converted into fuzzy concentration commands, which are then mapped to precise atomization volume and valve opening control parameters via a fuzzy rule base, reducing operational complexity.

[0100] The definition of fuzzy sets includes: the linguistic variable “concentration preference” is defined as {very light, light, moderate, light, very light}, with a corresponding universe of discourse of [0, 100%], and the membership function adopts a triangular or trapezoidal distribution (e.g., “light” corresponds to the concentration range of 15%-30%, and the membership decreases linearly with the increase of concentration).

[0101] The fuzzy rule base is constructed by establishing rules of "user instruction - concentration range - control parameters", for example: IF concentration preference = dimmer AND plant type = succulent THEN atomization amount = base value × 0.8, valve opening = 80%; IF concentration preference = stronger AND plant type = shrub THEN atomization amount = base value × 1.2, valve opening = 90%.

[0102] Fuzzy reasoning and defuzzification calculate the fuzzy set of control variables using the Mamdani reasoning method, and then use the centroid method to defuzzify and obtain precise values. The formula is: u=∑μ(ui)*ui / ∑μ(ui); where μ(ui) ​​is the membership degree of the control variable, and ui is the control variable. Users do not need to know the concentration percentage; they can directly issue commands through natural language (such as "apply diluted fertilizer to roses"), and the device automatically matches the plant database (with built-in default concentration ranges for 50+ common plants), overcoming the barrier of professional terminology. Users can input imprecise commands (such as "reduce the concentration a little more than last time"), and the device automatically adjusts through historical data memory and fuzzy logic to avoid concentration deviations caused by operational errors, which is especially suitable for middle-aged and elderly users or gardening enthusiasts. Combining users' historical command habits (such as the user-defined "diluted fertilizer" actually corresponding to a 20% concentration), the fuzzy rule base is dynamically corrected to form personalized operating habits and improve user-friendliness.

[0103] It should be understood that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the scope of the application. It should be understood that when an element or layer is referred to as “on,” “adjacent to,” “connected to,” or “coupled to” other elements or layers, it may be directly on, adjacent to, connected to, or coupled to other elements or layers, or there may be intervening elements or layers. Conversely, when an element is referred to as “directly on,” “directly adjacent to,” “directly connected to,” or “directly coupled to” other elements or layers, there are no intervening elements or layers. It should be understood that although the terms first, second, third, etc., may be used to describe various elements, components, areas, layers, and / or portions, these elements, components, areas, layers, and / or portions should not be limited by these terms. These terms are merely used to distinguish one element, component, area, layer, or portion from another element, component, area, layer, or portion. Therefore, without departing from the teachings of this application, the first element, component, area, layer, or portion discussed below may be referred to as a second element, component, area, layer, or portion.

[0104] Spatial relation terms such as “below,” “under,” “below,” “under,” “above,” “above,” etc., are used herein for convenience of description to describe the relationship between one element or feature shown in the figure and other elements or features. It should be understood that, in addition to the orientation shown in the figure, spatial relation terms are intended to also include different orientations of the device in use and operation. For example, if the device in the figure is flipped, then the element or feature described as “below,” “under,” or “below” other elements or features will be oriented “above” other elements or features. Therefore, the exemplary terms “below” and “under” can include both above and below orientations. The device may be otherwise oriented (rotated 90 degrees or otherwise) and the spatial descriptive terms used herein will be interpreted accordingly.

[0105] The terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the scope of this application. When used herein, the singular forms “a,” “an,” and “the” are also intended to include the plural forms unless the context clearly indicates otherwise. It should also be understood that the terms “comprising” and / or “including,” when used in this specification, identify the presence of the stated features, integers, steps, operations, elements, and / or components, but do not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and / or groups. When used herein, the term “and / or” includes any and all combinations of the associated listed items.

[0106] It should also be understood that the term “and / or” as used in this application specification and the appended claims means any combination of one or more of the associated listed items and all possible combinations, and includes such combinations.

[0107] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in this application, and these modifications or substitutions should all be covered within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. An intelligent liquid preparation device, characterized in that, include: The mixing chamber is used to collect the original solution and water to prepare the mixture; The raw liquid chamber is used to store the raw liquid and is connected to the mixing chamber through a liquid dispenser, the liquid dispenser having a piezoelectric atomization module; A water inlet pipe is used to supply water to the mixing chamber, and the water inlet pipe is equipped with a water inlet valve; A discharge pipe is used to discharge the liquid material in the mixing chamber, and the discharge pipe is equipped with a flow meter; The interactive module is used to obtain the user's input of the set concentration and set flow rate; The control module is used to control the opening degree of the inlet valve according to the set flow rate and to collect the feedback flow rate of the flow meter. The set concentration, feedback flow rate, and original solution concentration are input into the atomization volume calculation model. After local fine-tuning and global adjustment of the atomization volume calculation model, the target set atomization volume is calculated. Based on the target, the atomization amount is set to generate corresponding piezoelectric control parameters to instruct the piezoelectric atomization module to perform atomization actions; by adding a conductivity sensor, a pH sensor, and a camera, a multimodal perception system is constructed, and a convolutional neural network is used to identify plant varieties and growth status, and the current optimal nutrient solution concentration is inferred in real time by combining the sensor data corresponding to the conductivity sensor and pH sensor; the growth status includes leaf color and lesions; The process involves inputting the set concentration, feedback flow rate, and stock solution concentration into the atomization volume calculation model. After local fine-tuning and global adjustment of the atomization volume calculation model, the target set atomization volume is calculated. This includes: normalizing the set concentration, feedback flow rate, and stock solution concentration according to the input layer of the preset atomization volume calculation model to obtain normalized numerical variables corresponding to these values; performing feature processing on the normalized numerical variables according to the intermediate layer of the preset atomization volume calculation model to obtain feature vectors; and mapping the feature vectors to the initial set atomization volume according to the output layer of the preset atomization volume calculation model. The process involves several steps: First, based on a pre-built reward function, the output layer of a preset atomization quantity calculation model is locally fine-tuned according to the initial set atomization quantity. Then, the feature vector is remapped based on the locally fine-tuned output layer to obtain a candidate set atomization quantity. Next, based on a pre-built federated learning algorithm, the output layer of the preset atomization quantity calculation model is globally adjusted based on the locally fine-tuned values. The feature vector is then remapped based on the globally adjusted output layer to obtain the target set atomization quantity. The flowmeter includes an electromagnetically coupled rotating body and a detection module. The rotating body includes a driving component and a driven component equipped with a magnet.

2. The intelligent liquid preparation device as described in claim 1, characterized in that, Based on the pre-constructed reward function, the output layer of the preset atomization amount calculation model is locally fine-tuned based on the initial set atomization amount. The feature vector is then remapped based on the locally fine-tuned output layer to obtain candidate set atomization amounts, including: The effect of the initial atomization amount is evaluated by a pre-built reward function, and the gradient feedback of the reward function with respect to the initial atomization amount is calculated to obtain the direction for improving the current output quality. Based on gradient feedback of the reward function, the weights and bias parameters of the output layer are locally fine-tuned using a local learning rate. The purpose of this adjustment is to quickly improve the direction of the current output without affecting the global parameters of the entire model. The feature vector is input into the locally fine-tuned output layer, and forward propagation is performed again to obtain the updated candidate atomization amount.

3. The intelligent liquid preparation device as described in claim 1, characterized in that, Based on a pre-built federated learning algorithm, the output layer of a preset atomization calculation model is globally adjusted based on local fine-tuning values. The feature vector is then remapped using the globally adjusted output layer to obtain the target atomization amount, including: Each participating device locally performs a local fine-tuning operation of the output layer based on the reward function and calculates its local parameter update amount relative to the original output layer parameters; All clients upload their local parameter updates to the server. The server aggregates the local parameter updates of all clients using a federated averaging algorithm to generate a set of globally updated output layer global update parameters, enabling the model to integrate optimization experience from multiple devices while maintaining privacy. The updated global output layer parameters from the server are used to recalculate the feature vectors, and the output target is set with atomization amount.

4. The intelligent liquid preparation device as described in claim 3, characterized in that, When the server aggregates the local parameter updates of all clients using a federated averaging algorithm, it is also configured to perform the following steps: Obtain the local parameter update amount uploaded by each client and its corresponding device runtime data; The weight coefficient of each client is calculated based on the device runtime data, where the client with a longer runtime corresponds to a higher weight coefficient. The local parameter update amounts of each client are weighted and averaged according to the weighting coefficients to generate the globally updated output layer global update parameters. Before the global update parameters are sent to each client, the global update parameters are encrypted. The encryption process includes using a homomorphic encryption algorithm to perform layered encryption on the parameters, so that each client can only obtain the parameter part that matches the local model structure when decrypting.

5. The intelligent liquid preparation device as described in claim 1, characterized in that, Based on the target, the atomization amount is set to generate corresponding piezoelectric control parameters, including: Obtain a set of parameters reflecting the atomization characteristics of the original solution at the operating temperature; The parameter set is fitted with a function between the piezoelectric control parameters and the atomization amount to establish an atomization parameter control model. The piezoelectric control parameters corresponding to the target atomization amount are calculated based on the atomization parameter control model. The piezoelectric control parameters include the driving voltage, and the parameter set includes the viscosity and surface tension of the original liquid.

6. The intelligent liquid preparation device according to any one of claims 1-4, characterized in that, Also includes: The control module also collects the feedback concentration from the concentration detection sensor configured in the discharge pipeline, calculates the deviation between the feedback concentration and the set concentration, and compares it with the set range. When the deviation exceeds the set range, the control module issues an abnormality prompt message for the raw liquid through the interactive module. The concentration detection sensor includes a conductivity sensor.

7. The intelligent liquid preparation device according to any one of claims 1-4, characterized in that: The discharge pipe has a discharge direction, and the mixing chamber has an upper end face located upstream of the discharge direction, a lower end face located downstream of the discharge direction, and a side wall connecting the upper end face and the lower end face. The projected area of ​​the upper end face along the discharge direction is larger than the projected area of ​​the lower end face along the discharge direction. The water inlet pipe is connected to the mixing chamber at the upper end face, the liquid dispenser is connected to the mixing chamber at the side wall, and the discharge pipe is connected to the mixing chamber at the lower end face.

8. The intelligent liquid preparation device according to any one of claims 1-4, characterized in that: The rotating body is placed inside the discharge pipe and can rotate along a plane perpendicular to the discharge direction. The detection module includes a Hall sensor.

9. The intelligent liquid preparation device as described in claim 8, characterized in that: The driving component includes a rotating shaft portion parallel to the discharge direction and a plurality of fan blade portions surrounding the outside of the rotating shaft portion. The fan blade portions have concave arc surfaces intersecting the discharge direction. The driven component is connected to the discharge direction end of the rotating shaft portion. The flow meter also includes a guide plate located on the side of the driving component opposite to the discharge direction. The guide plate is provided with a plurality of guide holes for guiding the liquid material to contact the concave arc surface.

Images