AI-based biodegradation instrument multi-channel dehumidification control method and system
By combining a multi-stage condensation and silica gel adsorption structure with an AI intelligent monitoring module, the problem of unstable dehumidification efficiency and frequent maintenance of the biodegradation dehumidifier under complex multi-channel operating conditions is solved, achieving an adaptive, low-energy-consumption intelligent control effect.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- CHINESE ACAD OF INSPECTION & QUARANTINE
- Filing Date
- 2026-05-23
- Publication Date
- 2026-07-10
AI Technical Summary
Existing biodegradation dehumidifiers are difficult to adapt to complex multi-channel operating conditions, have unstable dehumidification efficiency, require frequent silica gel replacement, cannot intelligently diagnose abnormalities such as frost and blockage, rely on manual experience, and have high energy consumption.
Combining a multi-stage condensation and silica gel adsorption structure, an AI intelligent monitoring module based on physical information neural networks is introduced to form a closed-loop control process. It predicts dehumidification effect, optimizes operating parameters, provides early warning of faults, and provides feedback updates through real-time data.
It achieves adaptive dehumidification of the multi-channel biodegradation instrument, reduces maintenance frequency, improves detection stability and reliability, reduces energy consumption, and can intelligently identify anomalies and optimize control strategies.
Smart Images
Figure CN122351995A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of gas detection and dehumidification control technology for biodegradable instruments, and in particular to a multi-channel dehumidification control method and system for biodegradable instruments that combines a multi-channel condensation-adsorption dehumidification structure with an artificial intelligence-based intelligent control method. It belongs to the fields of biodegradable instrument dehumidification, industrial gas treatment, process control, and physical information neural network application technology. Background Technology
[0002] During the experimental testing process of the biodegradation instrument, microbial metabolism produces humid and hot gases containing a large amount of water vapor. If these humid and hot gases enter the carbon dioxide, methane, infrared analysis, or other gas detection modules directly without effective dehumidification, it can easily cause fogging of the optical window, sensor corrosion, signal drift, and detection errors. In severe cases, it can damage the precision detection devices at the back end, affecting the accuracy of the biodegradation rate measurement results and the long-term stability of the equipment.
[0003] Existing dehumidification devices typically employ single-stage condensation, single-stage desiccant adsorption, or fixed-program control methods. While these methods can remove some moisture under stable operating conditions, they suffer from problems such as large fluctuations in dehumidification load, frequent silica gel replacement, reliance on manual experience for operating parameters, and delayed fault response in scenarios involving simultaneous operation of multiple reaction channels, significant variations in gas production and moisture content across channels, and long experimental cycles. Particularly in multi-channel biodegradation experiments, where flow rates and humidity levels may differ between channels, fixed cooling power or fixed drainage cycles make it difficult to balance dehumidification efficiency, energy consumption, and maintenance cycles.
[0004] In recent years, artificial intelligence technology has begun to be applied to the fields of industrial control and process optimization, such as trend prediction based on historical data or rule-based temperature and humidity regulation. However, traditional data-driven models mostly remain at the back-end fitting level, making it difficult to fully reflect the physical mechanisms such as condensation heat exchange, gas flow, adsorption material saturation, and energy consumption balance. In dehumidification scenarios with small sample sizes or large variations in operating conditions, problems such as a lack of physical rationality in prediction results, unstable control strategies, and inaccurate maintenance warnings are prone to occur.
[0005] Therefore, it is necessary to provide a technical solution that can fully utilize the multi-stage dehumidification structure and automate the dehumidification process through AI or intelligent methods. This would enable the system to automatically predict dehumidification effects, optimize operating parameters, provide silica gel life warnings, diagnose faults, and update feedback based on real-time sensor data and physical constraints, thereby achieving adaptive, low-maintenance, and unattended operation of the multi-channel biodegradable dehumidifier dehumidification process. Summary of the Invention
[0006] The present invention aims to provide an AI-based multi-channel dehumidification control method and system for biodegradable instruments, in order to solve the problems of existing biodegradable dehumidification devices that rely on fixed programs or manual maintenance, are difficult to adapt to complex multi-channel operating conditions, cannot predict the lifespan of adsorbent materials, and cannot intelligently diagnose abnormalities such as frost and blockage.
[0007] The technical solution of this invention is not merely about setting up a device structure such as a condenser bottle and a dehumidifier bottle, but rather about introducing an AI intelligent monitoring module based on a physical information neural network on the basis of the aforementioned device structure. This module combines multi-dimensional data such as temperature, humidity, flow rate, pressure, condensate volume, silica gel state, and energy consumption with physical constraints such as heat exchange, gas flow, adsorption kinetics, and energy balance to form a closed-loop control process of "acquisition-analysis-decision-execution-feedback".
[0008] The present invention provides an AI-based multi-channel dehumidification control method for a biodegradable device, comprising: collecting operating parameters of key nodes in the multi-channel dehumidification system; preprocessing and fusing the collected data; inputting the processed data into an AI model based on a physical information neural network, wherein the AI model predicts the outlet humidity or dew point, dehumidification efficiency, remaining lifespan of the silica gel, optimal energy consumption parameters, and abnormal states; generating control commands based on the prediction results; adjusting the cooling power, circulating pump speed, valve opening, automatic drainage frequency, or maintenance warning status by the control unit; and transmitting the execution results back to the AI model to continuously optimize the control strategy.
[0009] This invention also provides an AI-based multi-channel dehumidification control system for a biodegradation apparatus, comprising a multi-channel gas distribution unit, a primary condensation dehumidification unit, a secondary adsorption dehumidification unit, a sensing and acquisition unit, an AI intelligent monitoring module, a control unit, a controllable actuator, a data storage and feedback optimization unit, and a remote APP interaction unit. The multi-channel gas distribution unit receives humid and hot gases generated by multiple biodegradation reaction bottles; the primary condensation dehumidification unit condenses and removes most of the water vapor; the secondary adsorption dehumidification unit deeply dries the condensed gas; the AI intelligent monitoring module performs prediction, diagnosis, and decision-making based on a physical information neural network; and the control unit executes the control strategy output by the AI model.
[0010] The specific technical solution adopted in this invention is an AI-based multi-channel dehumidification control method for a biodegradable instrument, applied to a dehumidification system for a biodegradable instrument including a multi-channel gas distribution unit, a primary condensation dehumidification unit, a secondary adsorption dehumidification unit, a sensing and acquisition unit, an AI intelligent monitoring module, a control unit, and controllable actuators. The method includes: the AI intelligent monitoring module collecting temperature, humidity, flow rate, pressure, condensate volume, silica gel state, and energy consumption parameters of the humid and hot gas in each channel during the dehumidification process; performing noise reduction, normalization, outlier removal, and multi-source fusion on the collected data; the AI intelligent monitoring module inputting the processed data into an AI model based on a physical information neural network; the AI model, combining heat conduction and phase change, gas flow, silica gel adsorption kinetics, and energy balance constraints, predicting the outlet humidity or dew point, dehumidification efficiency, remaining silica gel lifespan, channel flow distribution, and abnormal states; generating control commands based on the prediction results to adjust the cooling power, circulating pump speed, channel valve opening, automatic drainage frequency, or adsorption material maintenance status; the control unit driving the corresponding actuators to execute the control commands and feeding back the executed operating status to the AI intelligent monitoring module to update the dehumidification control strategy.
[0011] Furthermore, the loss function of the AI model based on the physical information neural network includes data fitting loss and physical constraint loss, and satisfies the following core computational relationship:
[0012]
[0013] in, The total loss of the model, For data fitting loss, For physical constraint loss, and The weighting coefficient is used; the physical constraint loss includes at least heat exchange constraint, gas flow constraint, adsorption kinetics constraint and energy consumption balance constraint.
[0014] Furthermore, the heat exchange constraint includes the following relationship:
[0015]
[0016] in, This represents the total heat released by the humid gas inside the condenser flask, corresponding to the original technical expression. ; This indicates the total heat absorbed by the cooling water, corresponding to ; Indicates the outlet temperature of the condenser flask, corresponding to ; Indicates the partial pressure of intake water vapor The calculated dew point temperature; when the AI model predicts If the condensation target is not met or the outlet dew point is higher than the set threshold, a control command is generated to increase the compressor's cooling power or increase the cooling water circulation speed.
[0017] Furthermore, the gas flow constraints include multi-channel mass conservation relationships and pipeline pressure loss relationships:
[0018]
[0019] in, For the first Channel gas flow rate Total intake airflow This is the friction factor along the pipeline. The length-to-diameter ratio of the pipe, For gas density, The gas flow rate is given; the AI model calculates the airflow distribution ratio of each channel based on the flow rate and pressure loss of each channel, and generates corresponding channel valve opening or bypass switching commands.
[0020] Furthermore, the remaining lifetime of the silica gel is predicted using the following adsorption kinetics relationship:
[0021]
[0022] in, The adsorption rate constant is . To balance the adsorption capacity of silica gel, for The actual amount of silica gel adsorbed at any given time. This represents the maximum saturation adsorption capacity of silica gel; when achieve When the preset ratio or predicted remaining lifespan is lower than the preset time threshold, the AI intelligent monitoring module sends a silicone replacement warning to the user terminal.
[0023] Furthermore, when the amount of condensate generated decreases abnormally and the outlet temperature of the condenser coil is abnormal, the AI intelligent monitoring module determines that there is a risk of frosting and triggers defrosting. When the pipeline pressure loss does not match the gas flow rate, the AI intelligent monitoring module determines that there is a risk of blockage or leakage and generates an alarm or channel switching command. When there is a logical contradiction between multiple sensor parameters that violates physical constraints, the AI intelligent monitoring module determines that there is a sensor abnormality and generates a verification or maintenance prompt.
[0024] Furthermore, the AI intelligent monitoring module records the outlet humidity, dew point, dehumidification efficiency, cooling energy consumption, pumping energy consumption, condensate volume, silica gel status, fault handling results, and user maintenance results after each control command is executed. This data is then input into the AI intelligent monitoring module as historical operating data to iteratively optimize the model parameters, early warning thresholds, or multi-channel control strategies within the AI intelligent monitoring module.
[0025] A multi-channel dehumidification control system for an AI-based biodegradation apparatus includes: a multi-channel gas distribution unit for receiving humid and hot gases generated by multiple biodegradation reaction channels and distributing them to corresponding dehumidification channels; a primary condensation dehumidification unit for performing phase change condensation dehumidification of the humid and hot gases through a condensation bottle and condensation coil; a secondary adsorption dehumidification unit for deep drying of the condensed gas through a dehumidification bottle filled with adsorption material; a sensing and acquisition unit for acquiring parameters such as temperature, humidity, flow rate, pressure, condensate volume, silica gel state, and energy consumption; an AI intelligent monitoring module for analyzing the parameters based on a physical information neural network and outputting prediction results and control strategies; a control unit for receiving the control strategies and controlling controllable actuators; and controllable actuators for adjusting the refrigeration compressor, circulating pump, automatic drain valve, gas path switching valve, or adsorption bottle maintenance interface.
[0026] Furthermore, the primary condensation dehumidification unit includes a condenser bottle, a spiral or serpentine condenser coil disposed inside the condenser bottle, a compressor refrigeration unit connected to the condenser coil, a circulating pump, an automatic drain valve, and a condensate collection tank; the secondary adsorption dehumidification unit includes an adsorption bottle, a porous support screen, and a silica gel adsorption layer, wherein the lower part of the adsorption bottle forms a gas buffer zone, and the upper part forms an adsorption drying zone.
[0027] Furthermore, the AI intelligent monitoring module also includes a remote interaction unit, which communicates with a mobile terminal via Wi-Fi or Bluetooth to display the dehumidification efficiency, operating parameters, remaining material lifespan, and fault alarm information of each channel, and to receive dehumidification target humidity, dew point threshold, maintenance threshold, or operating mode settings.
[0028] Compared with existing technologies, the present invention has at least the following beneficial effects: First, by combining multi-stage condensation and silica gel adsorption structures with AI closed-loop control, it can automatically adjust refrigeration, circulation, valve, and drainage parameters when humidity load changes; Second, by embedding heat conduction, phase change condensation, gas flow, adsorption kinetics, and energy consumption balance constraints into a physical information neural network, it improves the interpretability and reliability of prediction results; Third, it can predict the remaining lifespan of silica gel and push replacement warnings in advance, reducing manual inspections and emergency maintenance; Fourth, it can identify frost, blockage, leakage, or sensor anomalies based on sensor data and physical constraints; Fifth, it can continuously update the model through feedback data, enabling the system to adapt to different experimental channels, different gas production loads, and different environmental conditions. Attached Figure Description
[0029] Figure 1 This is a schematic diagram of a multi-channel dehumidification control system for an AI-based biodegradation instrument.
[0030] Figure 2This is a flowchart of an AI-based multi-channel dehumidification control method.
[0031] Figure 3 This is a schematic diagram of the PINN intelligent control model structure, which is a physical information neural network.
[0032] Figure 4 This is a block diagram of the AI intelligent dehumidification control system module.
[0033] Explanation of reference numerals in the attached diagram: 1 is the biodegradation reaction system; 2 is the multi-channel gas distribution main; 3 is the primary condensation dehumidification module; 3-1 is the condenser bottle; 3-2 is the condenser coil or heat exchanger; 3-3 is the automatic drain valve; 3-4 is the condensate collection tank; 3-5 is the circulating pump; 3-6 is the refrigeration compressor; 4 is the secondary adsorption dehumidification module; 4-1 is the adsorption bottle; 4-2 is the silica gel adsorption layer; 4-3 is the supporting screen; 5 is the gas detection or analysis module; 6 is the AI intelligent monitoring module; 7 is the controllable actuator; T is the temperature sensor; H is the humidity sensor; P is the pressure sensor; F is the flow sensor; L is the condensate level or metering sensor; S is the silica gel status sensor. Detailed Implementation
[0034] The embodiments of the present invention will be described below with reference to the accompanying drawings. It should be understood that the following embodiments are only for illustrating the technical solutions of the present invention and are not intended to limit the scope of protection of the present invention. Without departing from the technical concept of the present invention, those skilled in the art can make equivalent substitutions or improvements to the number of channels, sensor arrangement, condensation structure, adsorption material, AI model structure, and control strategy.
[0035] Example 1: System Structure
[0036] like Figure 1As shown, the input end of the multi-channel gas distribution manifold 2 is connected to multiple biodegradation reaction systems 1, and its output end is connected to the primary condensation and dehumidification module 3 of each independent channel. The gas outlet of the primary condensation and dehumidification module 3 is then connected to the secondary adsorption and dehumidification module 4, and the gas outlet of the secondary adsorption and dehumidification module 4 is finally connected to the gas detection or analysis module 5. In the above main gas path, the sensing and acquisition units are distributed near the inlet, condensation bottle, adsorption bottle, outlet and controllable actuator, and are used to collect data such as temperature, humidity, flow rate, pressure, condensate level and silica gel status, and transmit these signals to the AI intelligent monitoring module 6. After analyzing and predicting the data, the AI intelligent monitoring module 6 outputs the control strategy to the control unit, which drives the controllable actuator 7 (including refrigeration compressor, circulating pump, valve, automatic drain valve, etc.) to adjust the primary condensation and dehumidification module 3 and the secondary adsorption and dehumidification module 4. At the same time, the execution results are fed back to the AI intelligent monitoring module 6 and the data storage and feedback optimization unit for continuous model optimization. The remote APP interaction unit is connected to the control unit or data storage unit wirelessly to realize status viewing, parameter setting and alarm push.
[0037] The humid and hot gas generated by the multiple biodegradation reaction bottles provided in this embodiment enters the corresponding condensation-adsorption dehumidification channel through the multi-channel gas distribution manifold. Each channel can be independently configured with complete condensation bottles and adsorption bottles, or a combination of a shared refrigeration unit and an independent channel actuator can be used in specific scenarios.
[0038] The primary condensation dehumidification module 3 includes a condenser bottle 3-1, a spiral or serpentine condenser coil disposed within the condenser bottle 3-1, a refrigeration compressor 3-6 connected to the condenser coil, a circulation pump 3-5, an automatic drain valve 3-3, and a condensate collection tank 3-4. After entering the condenser bottle 3-1, the humid and hot gas exchanges heat with the low-temperature condenser coil, causing water vapor to undergo a phase change and condense. The condensate is discharged into the condensate collection tank through the automatic drain valve. The condenser coil can be a heat exchanger.
[0039] The secondary adsorption dehumidification module 4 includes an adsorption bottle 4-1, a porous support screen 4-3 located in the lower part of the adsorption bottle 4-1, and a silica gel adsorption layer 4-2 located in the upper part. The condensed gas enters the lower buffer zone of the adsorption bottle 4-1. After the airflow is evenly distributed through the support screen 3, it passes through the silica gel adsorption layer, where residual moisture is further adsorbed, ultimately forming dry gas that meets the detection requirements and enters the gas detection or analysis module 5.
[0040] The sensor acquisition units are distributed near the air inlet, condenser bottle, adsorption bottle, air outlet, and controllable actuator, and are used to collect data such as temperature, humidity, flow rate, pressure, condensate volume, silica gel state, refrigeration power, and circulating pump operating status. This data is not only used for real-time display but also serves as input for the AI intelligent monitoring module 6 to make predictions and decisions.
[0041] Example 2: AI-based dehumidification control process
[0042] like Figure 2 As shown, the control method provided in this embodiment includes the following steps.
[0043] Step S1: Multi-channel dehumidification parameter acquisition. The system collects parameters such as inlet air temperature, outlet air temperature, inlet and outlet humidity, gas flow rate, pipeline pressure, condensate adsorption status, cooling power, and circulation pump status for each channel through temperature sensors, humidity sensors, flow sensors, pressure sensors, condensate level or metering sensors, and silica gel status sensors.
[0044] Step S2, Data Preprocessing. The data processing unit performs noise reduction, outlier removal, unit unification, normalization, and multi-source data fusion on the collected data, and constructs the model input dataset by combining historical operating data and simulation data.
[0045] Step S3: State analysis and dehumidification effect prediction based on physical information neural network. The AI model predicts the outlet humidity, dew point temperature, dehumidification efficiency, remaining life of silica gel, optimal energy consumption parameters, and abnormal risks for each channel based on real-time input data and physical constraints.
[0046] Step S4: Generate control strategies based on the prediction results. When the model predicts that the outlet humidity or dew point does not meet the target, generate instructions to increase the cooling power, increase the circulation pump speed, or adjust the opening of the corresponding channel valve; when the condensate volume reaches the threshold, generate a drainage instruction; when the remaining life of the silica gel is lower than the threshold, generate a replacement warning; when frost, blockage, or sensor malfunction occurs, generate a fault handling instruction.
[0047] Step S5: Perform multi-channel dehumidification control. The control unit drives the refrigeration compressor, circulating pump, gas path switching valve, automatic drain valve, adsorption bottle maintenance interface, or alarm device to perform corresponding actions according to the control strategy output by the AI model.
[0048] Step S6: Determine whether the air humidity or dew point meets the target. If not, the system feeds back the current execution status to the AI model for re-analysis and further adjusts the operating parameters; if it meets the target, proceed to the lifespan prediction, fault diagnosis, and operation result recording process.
[0049] Steps S7 to S9 involve silica gel lifetime prediction, fault diagnosis, and model updates. The system predicts the remaining lifetime of the silica gel based on its adsorption state, gas humidity, and flow rate. It identifies frost formation, blockage, or sensor malfunctions based on physical constraints and writes the operating data and control results as new samples into the data storage and feedback optimization unit for subsequent online model updates or periodic retraining.
[0050] Example 3: Formula Preservation, Physical Constraints, and AI Model
[0051] like Figure 3 As shown, in this embodiment, the AI intelligent monitoring module preferably uses a Physical Information Neural Network (PINN) as its core analysis model. The input layer of this model receives preprocessed multidimensional data, including the condenser flask inlet temperature. Condenser outlet temperature Cooling water inlet and outlet temperatures and Gas mass flow rate Cooling water mass flow rate Multi-channel gas mass flow rate Pipeline pressure loss Gas flow rate Real-time adsorption capacity of silica gel Partial pressure of gas and water vapor Compressor cooling capacity Circulating pump head and cooling water volume flow rate wait.
[0052] In this embodiment, the following formulas are retained in the specific implementation as physical constraints, prediction criteria, and control decision criteria for the AI model. These formulas do not limit the model to a single network structure, but rather serve to ensure that the AI model's predictions conform to the fundamental physical laws of the dehumidification process.
[0053] Heat exchange and dew point constraints can be expressed as:
[0054]
[0055]
[0056] in, This represents the total heat released by the humid gas inside the condenser flask, corresponding to the original technical expression. This includes sensible heat and latent heat of water vapor phase change; This indicates the total heat absorbed by the cooling water, corresponding to ; Indicates the outlet temperature of the condenser flask, corresponding to ; This indicates the partial pressure of water vapor at the inlet of the condenser flask, corresponding to... ; This represents the dew point temperature calculated based on the partial pressure of water vapor in the intake air. The energy conservation relationship described above is used to ensure that the condensation process meets the energy conservation requirement, while the dew point judgment relationship is used to determine whether water vapor has the temperature conditions for phase change and condensation, and to provide a basis for the AI model to adjust the refrigeration power and circulation pump speed.
[0057] Gas flow constraints can be expressed as:
[0058]
[0059]
[0060] in, Indicates the number of reaction channels. Indicates the channel number. Indicates the first Channel gas flow rate Indicates the total intake airflow; Indicates pipeline pressure loss. This represents the friction coefficient along the pipeline. Indicates the length-to-diameter ratio of the pipe. Indicates gas density, This represents the gas flow rate. The above mass conservation relationship is used to ensure that the multi-channel gas distribution meets the mass conservation principle, and the pressure loss relationship is used to quantify the relationship between pipeline pressure loss and flow rate, and can be used to identify blockages, leaks, or abnormal flow rates.
[0061] The adsorption kinetic constraints can be expressed as:
[0062]
[0063] in, express Actual amount of silica gel adsorbed at any given time, subscript Indicates time or moment; This represents the silica gel adsorption rate constant; Indicates the equilibrium adsorption capacity of silica gel, subscript Indicates a state of equilibrium; Indicates the maximum saturation adsorption capacity of silica gel, subscript This represents the maximum saturation state. This formula describes the relationship between the silica adsorption rate and the remaining adsorption capacity, enabling the model to predict the remaining lifetime of silica based on the real-time adsorption state. achieve The system sends a replacement warning to the user terminal when the preset ratio, such as 90%, or when the predicted remaining lifespan is less than 7 days.
[0064] Energy balance constraints can be expressed as:
[0065]
[0066]
[0067] in, Indicates the energy consumption of the refrigeration unit, corresponding to ; Indicates cooling capacity, corresponding to ; Indicates the coefficient of performance (COP). Indicates runtime; This indicates the energy consumption of the circulating pump, corresponding to ; Indicates the head of the circulating pump, corresponding to ; Indicates the volumetric flow rate of cooling water, corresponding to ; Indicates the efficiency of the circulating pump, corresponding to By using the energy consumption constraints described above, the model can optimize the cooling power and circulation pump speed while meeting the dehumidification requirements, thus avoiding unnecessary energy consumption under low humidity conditions.
[0068] The total loss function of the PINN model can be expressed as:
[0069]
[0070]
[0071] in, To represent the data fitting loss, the mean square error is preferably used to calculate the deviation between the model's predicted value and the sensor's measured value. This indicates the deviation of the prediction results from the various physical constraint equations; This represents the total loss of the model; and These are the weighting coefficients for data fitting loss and physical constraint loss, respectively. By minimizing the total loss function, the model fits the measured data while adhering to fundamental physical laws such as condensation heat transfer, gas flow, adsorption kinetics, and energy balance.
[0072] The meanings of superscripts, subscripts, and digits involved in the above physical relationships are as follows: Indicates the numbering of different reaction channels or gas branches. Indicates the total number of channels; Indicates runtime or a specific moment in time; This indicates the adsorption equilibrium state; This indicates the maximum saturation state of the adsorbent material; , , , Subscripts are used to distinguish physical quantities on the gas side, water side, circulating pump side, or refrigeration side. To ensure formula editability, variable readability, and interpretability of the implementation process, Latin variable symbols are used in the formulas, and their correspondences with those in the text are indicated. , , , , , The correspondence between the original technical expressions is established. By explicitly defining the above symbols, the implementability and auditability of the formulas in the embodiments can be guaranteed.
[0073] Example 4: Intelligent Decision Making, Fault Diagnosis, and Feedback Optimization
[0074] The AI intelligent monitoring module outputs predictions including dehumidification efficiency, outlet dew point, remaining silica gel life, optimal energy consumption parameters, channel airflow distribution ratio, and fault risk type. The intelligent decision-making unit generates executable control commands based on these results.
[0075] When the model determines that the dew point is higher than the set threshold or the dehumidification efficiency is insufficient, the control unit increases the power of the refrigeration compressor, increases the speed of the cooling water circulation pump, or increases the valve opening and condensation residence time for channels with high moisture load. When the model determines that the flow rate of a certain channel is too high or the pressure loss is abnormal, the control unit can adjust the valve opening of that channel or activate bypass regulation to ensure that the dehumidification effect of multiple channels is consistent.
[0076] When the condensate level or metering sensor detects that the condensate has reached a preset threshold, the control unit controls the automatic drain valve or pumping mechanism to discharge the condensate into the condensate collection tank and records the drainage volume and drainage frequency, which are used to estimate the dehumidification load and correct the dehumidification efficiency.
[0077] When the silica gel status sensor or adsorption kinetics model determines that the remaining lifespan of the silica gel is below a threshold, the system pushes a replacement warning via a remote APP interaction unit. After replacing the silica gel, the user can confirm the maintenance result through the APP. The system stores the maintenance time in association with the dehumidification effect before and after replacement, which is used to correct the lifespan prediction model.
[0078] When the system detects an abnormal decrease in condensate production and an abnormal temperature at the condenser coil outlet, it can determine that there is a risk of frosting and trigger the defrosting procedure; when the pipeline pressure loss increases abnormally and the gas flow rate decreases, it can determine that there is a risk of blockage; when there is a logical contradiction in the data from multiple sensors that violates physical constraints, it can determine that the sensors are abnormal and issue a verification prompt.
[0079] like Figure 4As shown, the system modules of this invention may include a multi-channel dehumidification unit, a sensing and acquisition unit, a data processing unit, a PINN model analysis unit, a decision control unit, an execution unit, a data storage and feedback optimization unit, and a remote APP interaction unit. The multi-channel dehumidification unit completes the process treatment of humid and hot gases; the sensing and acquisition unit acquires the operating status; the data processing unit completes data cleaning, outlier handling, unit standardization, and feature extraction; the PINN model analysis unit completes prediction and diagnosis; the decision control unit generates operating control strategies; the execution unit drives hardware actions; the data storage and feedback optimization unit records operating data and supports model updates; and the remote APP interaction unit enables status viewing, parameter setting, and alarm push notifications.
[0080] In one specific embodiment, the primary condensation dehumidification unit preferentially condenses most of the water vapor in the humid and hot gas, improving the initial dehumidification efficiency by more than 60%. The secondary adsorption dehumidification unit, due to the reduced moisture load entering the silica gel due to the pre-stage condensation, can extend the silica gel replacement cycle from approximately 5 days to approximately 1 month. The above data can be considered as preferred implementation effects; actual values can be adjusted according to experimental conditions, number of channels, gas flow rate, and differences in adsorption materials.
[0081] In another specific embodiment, the system can connect to a mobile terminal via Wi-Fi or Bluetooth. Users can view the temperature and humidity, flow rate, pressure, dehumidification efficiency, condensate volume, remaining silica gel life, fault alarms, and historical trends for each channel in the app. They can also set dew point thresholds, humidity targets, maintenance reminder thresholds, and operating modes. The system can also automatically adjust control strategies based on user settings, thereby enabling remote operation and maintenance in an unattended state.
[0082] Therefore, this invention, through the organic combination of a multi-channel structure, a two-stage collaborative dehumidification structure, an AI intelligent monitoring module, a physical information neural network model, and an execution feedback closed loop, transforms the dehumidification process of the biodegradation instrument from fixed program control to adaptive intelligent control based on real-time status and physical constraints. This can improve dehumidification stability and detection reliability while reducing maintenance frequency and energy consumption.
Claims
1. A multi-channel dehumidification control method for a biodegradable instrument based on AI, applied to a dehumidification system for a biodegradable instrument including a multi-channel gas distribution unit, a primary condensation dehumidification unit, a secondary adsorption dehumidification unit, a sensing and acquisition unit, an AI intelligent monitoring module, a control unit, and a controllable actuator; characterized in that, The method includes: an AI intelligent monitoring module collecting temperature, humidity, flow rate, pressure, condensate volume, silica gel state, and energy consumption parameters of the humid and hot gas in each channel during the dehumidification process; performing noise reduction, normalization, outlier removal, and multi-source fusion on the collected data; the AI intelligent monitoring module inputting the processed data into an AI model based on a physical information neural network, the AI model combining heat conduction and phase change, gas flow, silica gel adsorption kinetics, and energy balance constraints to predict the outlet humidity or dew point, dehumidification efficiency, silica gel remaining lifespan, channel flow distribution, and abnormal states; generating control commands based on the prediction results to adjust the cooling power, circulating pump speed, channel valve opening, automatic drainage frequency, or adsorption material maintenance status; the control unit driving the corresponding actuator to execute the control commands, and feeding back the executed operating status to the AI intelligent monitoring module to update the dehumidification control strategy.
2. The AI-based multi-channel dehumidification control method for biodegradation apparatus according to claim 2, characterized in that, The loss function of the AI model based on the physical information neural network includes data fitting loss and physical constraint loss, and satisfies the following core computational relationship: ； in, The total loss of the model, For data fitting loss, For physical constraint loss, and The weighting coefficient is used; the physical constraint loss includes at least heat exchange constraint, gas flow constraint, adsorption kinetics constraint and energy consumption balance constraint.
3. The AI-based multi-channel dehumidification control method for a biodegradation instrument according to claim 2, characterized in that, The heat exchange constraints include the following relationships: ； in, This represents the total heat released by the humid gas inside the condenser flask, corresponding to the original technical expression. ; This indicates the total heat absorbed by the cooling water, corresponding to ; Indicates the outlet temperature of the condenser flask, corresponding to ; Indicates the partial pressure of intake water vapor The calculated dew point temperature; when the AI model predicts If the condensation target is not met or the outlet dew point is higher than the set threshold, a control command is generated to increase the compressor's cooling power or increase the cooling water circulation speed.
4. The AI-based multi-channel dehumidification control method for a biodegradation instrument according to claim 2, characterized in that, The gas flow constraints include multi-channel mass conservation relationships and pipeline pressure loss relationships: ； in, For the first Channel gas flow rate Total intake airflow This is the friction factor along the pipeline. The length-to-diameter ratio of the pipe, For gas density, The gas flow rate is given; the AI model calculates the airflow distribution ratio of each channel based on the flow rate and pressure loss of each channel, and generates corresponding channel valve opening or bypass switching commands.
5. The AI-based multi-channel dehumidification control method for a biodegradation instrument according to claim 2, characterized in that, The remaining lifetime of the silica gel is predicted using the following adsorption kinetics relationship: ； in, The adsorption rate constant is . To balance the adsorption capacity of silica gel, for The actual amount of silica gel adsorbed at any given time. This represents the maximum saturation adsorption capacity of silica gel; when achieve When the preset ratio or predicted remaining lifespan is lower than the preset time threshold, the AI intelligent monitoring module sends a silicone replacement warning to the user terminal.
6. The AI-based multi-channel dehumidification control method for a biodegradation instrument according to claim 1, characterized in that, When the amount of condensate generated decreases abnormally and the outlet temperature of the condenser coil is abnormal, the AI intelligent monitoring module determines that there is a risk of frosting and triggers defrosting. When the pipeline pressure loss does not match the gas flow rate, the AI intelligent monitoring module determines that there is a risk of blockage or leakage and generates an alarm or channel switching command. When there is a logical contradiction between multiple sensor parameters that violates physical constraints, the AI intelligent monitoring module determines that there is a sensor malfunction and generates a verification or maintenance prompt.
7. The AI-based multi-channel dehumidification control method for a biodegradation instrument according to claim 1, characterized in that, The AI intelligent monitoring module records the outlet humidity, dew point, dehumidification efficiency, cooling energy consumption, pumping energy consumption, condensate volume, silica gel status, fault handling results, and user maintenance results after each control command is executed. This data is then input into the AI intelligent monitoring module as historical operating data to iteratively optimize the model parameters, early warning thresholds, or multi-channel control strategies within the AI intelligent monitoring module.
8. A multi-channel dehumidification control system for an AI-based biodegradation device, characterized in that, include: A multi-channel gas distribution unit receives humid and hot gases generated by multiple biodegradation reaction channels and distributes them to corresponding dehumidification channels; a primary condensation dehumidification unit performs phase change condensation dehumidification of the humid and hot gases through a condenser bottle and condenser coil; a secondary adsorption dehumidification unit performs deep drying of the condensed gas through a dehumidification bottle filled with adsorption material; a sensing and acquisition unit collects parameters such as temperature, humidity, flow rate, pressure, condensate volume, silica gel state, and energy consumption; an AI intelligent monitoring module analyzes the parameters based on a physical information neural network and outputs prediction results and control strategies; a control unit receives the control strategies and controls controllable actuators; and controllable actuators adjust the refrigeration compressor, circulating pump, automatic drain valve, gas path switching valve, or adsorption bottle maintenance interface.
9. The AI-based multi-channel dehumidification control system for a biodegradation apparatus according to claim 8, characterized in that, The primary condensation dehumidification unit includes a condenser bottle, a spiral or serpentine condenser coil disposed inside the condenser bottle, a compressor refrigeration unit connected to the condenser coil, a circulating pump, an automatic drain valve, and a condensate collection tank; the secondary adsorption dehumidification unit includes an adsorption bottle, a porous support screen, and a silica gel adsorption layer, wherein the lower part of the adsorption bottle forms a gas buffer zone, and the upper part forms an adsorption drying zone.
10. The AI-based multi-channel dehumidification control system for a biodegradation apparatus according to claim 8, characterized in that, The AI intelligent monitoring module also includes a remote interaction unit, which communicates with a mobile terminal via Wi-Fi or Bluetooth to display the dehumidification efficiency, operating parameters, remaining material lifespan, and fault alarm information of each channel, and to receive dehumidification target humidity, dew point threshold, maintenance threshold, or operating mode settings.