Laboratory automation control system based on ai model

Abstract

The application belongs to the technical field of temperature and humidity control, and particularly relates to a laboratory automation control system based on an AI model. In view of the problems such as difficulty in independent regulation and control of temperature and humidity, and neglect of the influence of personnel activities in traditional laboratory environment control technology, the system integrates a multi-modal environment perception module, a personnel behavior modeling module, an intelligent decision and control module, and a false touch prevention verification module. The multi-modal environment perception module deploys multiple sensors to collect data and construct a three-dimensional environment thermal map to monitor temperature and humidity in real time; the personnel behavior modeling module identifies personnel positions and activity intensity and quantifies the influence of personnel on temperature and humidity; the intelligent decision and control module optimizes regulation and control according to a man-unmanned mode switching strategy; and the false touch prevention verification module uses LSTM to identify operation intention and prevent false operation. The application realizes accurate regulation and control of laboratory temperature and humidity, improves the quality and safety of the experimental environment, and guarantees the accuracy of experimental results and the stability of equipment.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of temperature and humidity control, and particularly relates to a laboratory automation control system based on an AI model. BACKGROUND

[0002] In the field of laboratory environment control, traditional technologies have many problems. First, independent temperature and humidity regulation is difficult to achieve. The requirements for temperature and humidity in the laboratory are often fine and complex, and different experimental projects may require specific temperature and humidity combinations. However, existing technologies cannot accurately and efficiently regulate temperature and humidity independently and accurately, often resulting in a trade-off between the two, affecting the accuracy of experimental results and the normal operation of experimental equipment. Second, the influence of personnel activities on temperature and humidity in the laboratory has long been ignored. Personnel's metabolic heat production, respiratory moisture, and other behaviors can change the temperature and humidity environment of the laboratory. However, traditional control systems do not consider personnel factors in the operation process, resulting in deviations in temperature and humidity regulation and difficulty in maintaining a stable experimental environment. SUMMARY

[0003] The present application proposes a laboratory automation control system based on an AI model to address the technical problems in the background technology.

[0004] To achieve the above purpose, the technical solution adopted by the present application is as follows: a multi-modal environment perception module, a personnel behavior modeling module, an intelligent decision and control module, and a false touch prevention verification module are included.

[0005] The multi-modal environment perception module includes deployment of sensors and data fusion.

[0006] Deployment of sensors: distributed temperature and humidity sensors, airflow sensors, and human perception sensors are deployed.

[0007] Data fusion: multi-source data is fused using flow field coupling Kalman to construct a three-dimensional environmental thermal map of the laboratory, and the temperature and humidity conditions are monitored in real time.

[0008] The personnel behavior modeling module includes behavior recognition and impact quantification.

[0009] Behavior recognition: the position of personnel is tracked in real time through the YOLO algorithm, and the activity intensity of personnel is identified in combination with pressure sensors.

[0010] Impact quantification: personnel metabolic heat production and respiratory moisture functions are constructed according to activity intensity, and the contribution value of personnel to temperature and humidity is dynamically calculated.

[0011] The intelligent decision and control module performs dynamic strategy switching, including an unmanned model and a manned mode.

[0012] Unmanned mode: model predictive control (MPC) is used to maintain constant temperature and humidity.

[0013] Person mode: Establish a multi-objective genetic algorithm targeting temperature and humidity accuracy, energy consumption and personnel comfort for multi-objective optimization;

[0014] The anti-mis-touch verification module includes behavior intention recognition and mis-touch protection.

[0015] Behavior intention recognition: LSTM is used to learn and analyze multi-dimensional data such as hand motion trajectory, touch time and touch position of the operator, to determine whether the behavior intention of the operator is normal operation or mis-touch.

[0016] Mis-touch protection: When mis-touch is determined, mis-operation is prevented through a voice confirmation mechanism, and secondary verification is required.

[0017] As preferred, the multi-modal environment perception module, the specific implementation of constructing a laboratory three-dimensional environment thermal map in data fusion includes state vector expansion and flow field model embedding.

[0018] State vector expansion: introduce air flow velocity vector and air moisture content , construct a seven-dimensional state vector: , wherein is the air temperature at the current time, is the relative humidity at the current time, is the temperature and humidity gradient vector at the current time t.

[0019] Flow field model embedding: predict the temperature and moisture content at the next time through flow field model embedding, the calculation method is: , , wherein is random noise.

[0020] As preferred, the behavior recognition YOLO algorithm in the personnel behavior modeling module uses YOLOv8 model architecture, combined with improved Deformable DETR to improve positioning accuracy in personnel occlusion scenarios, the improvement of the Deformable DETR is to integrate space-time sequence memory, including a 10-frame long history trajectory buffer, storing the three-dimensional coordinates and occlusion state of the personnel; when the target is completely occluded, the next frame position is predicted through linear interpolation, the prediction error is ≤15 cm; At the same time, calculate the occlusion state transition probability, assign a 2 times attention search weight to the area within 50 cm around the historical trajectory.

[0021] As preferred, the pressure sensor is arranged on the ground of the experiment table and the passage, and the pressure signals are collected by a distributed array, and the rate of change of the pressure data, i.e. the amount of change of the pressure per unit time, is calculated as a characteristic of the activity intensity; the extracted pressure characteristics are used as input by the random forest, and the activity intensity score of the personnel is output.

[0022] As preferred, in the impact quantification of the personnel behavior modeling module, the metabolic heat production and respiratory moisture dissipation functions of the personnel are constructed according to the activity intensity, and the contribution value of the personnel to the temperature and humidity is dynamically calculated, which specifically includes:

[0023] The contribution value of the metabolic heat production of the personnel to the temperature is calculated by the following formula: wherein K is the activity intensity score of the personnel, M is the heat production coefficient per unit activity intensity, is the activity duration of the personnel, is the target temperature preset for the laboratory, is the fluctuation range of the temperature of the laboratory;

[0024] The contribution value of the respiratory moisture dissipation to the humidity is calculated by the following formula: wherein D is the moisture dissipation coefficient per unit activity intensity, H is the actual humidity of the current laboratory, is the maximum allowable value of the humidity of the laboratory.

[0025] As preferred, the final laboratory temperature and humidity standard to be set is calculated according to the contribution values of the temperature and humidity, and the influence factor of the temperature on the humidity is added when the standard is set, to obtain the final laboratory temperature and humidity standard to be set, including the final temperature setting and the final humidity setting;

[0026] The calculation method of the final temperature setting is as follows: wherein is the specific heat capacity of air at constant pressure, V represents the volume of the laboratory, is the air density, is the temperature regulation compensation coefficient;

[0027] The final humidity setting is realized by coupling the influence of the temperature, and the calculation method is as follows: wherein represents the target humidity preset for the laboratory, represents the saturation moisture content at the temperature ; is the compensation range of the saturation humidity with the change of the temperature.

[0028] As preferred, the specific steps of the multi-objective genetic algorithm optimization are as follows:

[0029] Decision variables: determine the operating parameters of the equipment for controlling the temperature and humidity;

[0030] Encoding: encode the decision variables as chromosomes;

[0031] Initializing population: randomly generate a certain number of chromosomes to form an initial population, and the gene values of each chromosome are randomly generated within the value range of the decision variables;

[0032] Adaptive evaluation: adopt non-dominated sorting to obtain a set of current optimal Pareto optimal solutions, and then perform selection, crossover and mutation operations to find the optimal Pareto optimal solution set in each iteration;

[0033] Termination condition judgment: when the fitness of the population is less than the fitness change threshold, the iteration is terminated, and the optimal Pareto optimal solution set is output.

[0034] Compared with the prior art, the advantages and positive effects of the present application are that the multi-modal environment perception module accurately monitors temperature and humidity; the personnel behavior modeling module takes into account personnel activities and dynamically calculates their impact on temperature and humidity; the intelligent decision and control module optimizes the control strategy for the presence and absence of people, balancing temperature and humidity, energy consumption and comfort; the anti-misoperation verification module uses LSTM to identify operation intent, combined with voice confirmation and secondary verification, to avoid misoperation. The overall implementation of the laboratory temperature and humidity accurate independent regulation improves the stability and safety of the experimental environment, and overcomes the shortcomings of traditional technology. DETAILED DESCRIPTION

[0035] In order to enable more clear understanding of the above-mentioned purpose, features and advantages of the present application, the present application will be further described below in combination with embodiments. It should be noted that the embodiments of the present application and the features in the embodiments can be combined with each other without conflict.

[0036] In the following description, many specific details are set forth in order to provide a thorough understanding of the present application, however, the present application can also be implemented in other ways different from those described herein, therefore, the present application is not limited to the specific embodiments disclosed in the following description.

[0037] In modern laboratory operation, accurate environmental control is crucial for the accuracy of experimental results, the stability of experimental equipment and the comfort of personnel. However, traditional laboratory environmental control technology has many problems, such as difficulty in independent regulation of temperature and humidity, inability to effectively consider the impact of personnel activities on the environment, etc., which seriously restricts the efficient operation of the laboratory. In order to achieve the effect of accurate independent regulation of laboratory temperature and humidity, fully consider the impact of personnel activities on the environment, improve the stability and safety of the experimental environment, and solve the shortcomings of traditional laboratory environmental control technology, the present application adopts an AI model-based laboratory automation control system, which specifically includes a multi-modal environment perception module, a personnel behavior modeling module, an intelligent decision and control module, and an anti-misoperation verification module.

[0038] The multimodal environmental sensing module includes sensor deployment and data fusion. Sensor deployment includes distributed temperature and humidity sensors, airflow sensors, and human body sensors. Specifically, the distributed temperature and humidity sensors are evenly distributed throughout the laboratory, installed near lab benches, corners, and ventilation openings, to comprehensively collect temperature and humidity data from different areas and accurately grasp spatial differences in temperature and humidity. Airflow sensors are installed inside ventilation ducts and at key airflow nodes. Sensors in the ventilation ducts monitor parameters such as temperature, humidity, and velocity of incoming and outgoing airflow, while sensors at key nodes capture changes in airflow within the room, providing a basis for analyzing the impact of airflow on temperature and humidity. Human body sensors are deployed at the laboratory entrance, in corridors, and around lab benches. Sensors at the entrance detect personnel entering and exiting, those in corridors track personnel movement, and those around lab benches sense personnel activity during experiments, thus promptly detecting potential changes in environmental temperature and humidity caused by personnel activity.

[0039] Data fusion: A flow-field coupled Kalman fusion method is used to fuse multi-source data to construct a three-dimensional thermal map of the laboratory environment, enabling real-time monitoring of temperature and humidity. Specifically, the multimodal environmental sensing module's implementation of constructing the three-dimensional thermal map of the laboratory environment during data fusion includes state vector extension and flow field model embedding. State vector extension involves introducing airflow velocity vectors. and air humidity Construct a seven-dimensional state vector: ,in The current air temperature. The relative humidity at the current moment. This represents the temperature and humidity gradient vector at the current time t; Flow field model embedding: The temperature and humidity at the next time step are predicted by embedding the flow field model, and the calculation method is as follows: , ,in The noise is random. When constructing the 3D environmental heat map of the laboratory, a flow-field coupled Kalman filter algorithm is used to fuse multi-source data containing a seven-dimensional state vector. By continuously updating the data in the state vector, information such as temperature, humidity, airflow velocity, and air moisture content collected by sensors is effectively integrated. Combined with the predicted temperature and humidity values ​​for the next time step obtained from the flow field model embedding, it is transformed into an intuitive 3D heat map. The heat map displays the temperature and humidity conditions at various locations in the laboratory using different colors and brightness levels, achieving real-time, visual monitoring of the laboratory's temperature and humidity.

[0040] The personnel behavior modeling module comprises behavior recognition and influence quantification. The behavior recognition comprises: tracking the position of the personnel in real time through a YOLO algorithm, and recognizing the activity intensity of the personnel in combination with a pressure sensor. A YOLOv8 model architecture is adopted, and the positioning accuracy of the personnel in a shielding scene is improved in combination with an improved Deformable DETR. The improvement of the Deformable DETR is integration of space-time sequence memory, which comprises a historical trajectory buffer zone with a length of 10 frames, and stores the three-dimensional coordinates and shielding state of the personnel. When the target is completely shielded, the next frame position is predicted through linear interpolation, and the prediction error is less than or equal to 15 cm. Meanwhile, the shielding state transition probability is calculated, and the regions within a range of 50 cm near the historical trajectory are given a search weight of attention that is 2 times. The pressure sensor is arranged on the ground of the experiment table and the passage, and the pressure signal is collected through a distributed array. The change rate of the pressure data, that is, the change amount of the pressure per unit time, is calculated as a feature of the activity intensity. The pressure features are used as input into a random forest, and the activity intensity score of the personnel is output. Specifically, in the laboratory scene, the pressure sensor array is arranged on the ground of the experiment table and the passage for accurately recognizing the activity intensity of the personnel. The sensor collects pressure signals in real time, and the change rate of the pressure data is obtained by calculating the change amount of the pressure per unit time, which is used as a feature of the activity intensity of the personnel. Then, the pressure features are input into the random forest model. The random forest outputs the activity intensity score of the personnel based on the comprehensive judgment of the multiple decision trees inside the random forest. This process guarantees the comprehensiveness of data collection through distributed deployment, effectively reflects the activity intensity by using the pressure change rate, and realizes accurate evaluation of the activity intensity of the personnel by virtue of the strong learning ability and stability of the random forest model.

[0041] The influence quantification comprises: constructing personnel metabolic heat production and respiratory moisture emission functions according to the activity intensity, and dynamically calculating the contribution value of the personnel to the temperature and humidity. The contribution value of the personnel metabolic heat production to the temperature is calculated according to the following formula: wherein K is the activity intensity score of the personnel, M is the heat production coefficient per unit activity intensity, is the activity duration of the personnel, is the target temperature preset in the laboratory, is the fluctuation range of the temperature of the laboratory. The contribution value of the respiratory moisture emission to the humidity is calculated according to the following formula: wherein D is the moisture emission coefficient per unit activity intensity, H is the actual humidity of the current laboratory, is the maximum allowable value of the humidity of the laboratory. The final laboratory temperature and humidity standard to be set is calculated according to the contribution values of the temperature and humidity, and a temperature influence factor on the humidity is added when the standard is set, to obtain the final laboratory temperature and humidity standard to be set, including the final temperature setting and the final humidity setting. The final temperature setting is calculated according to the following formula: wherein V represents the volume of the laboratory for the specific heat capacity of air at constant pressure, V represents the volume of the laboratory for the density of air, V represents the volume of the laboratory for the density of air, V represents the volume of the laboratory for the density of air, V represents the volume of the laboratory for the density of air, V represents the volume of the laboratory for the density of air, V represents the volume of the laboratory for the density of air, V represents the volume of the laboratory for the density of air, By constructing the metabolic heat production and respiratory moisture dissipation function, the contribution value of personnel activity intensity to temperature and humidity can be dynamically and accurately calculated, and the traditional extensive temperature and humidity control is abandoned. Not only the influence of personnel activity is considered, but also the influence of temperature on humidity is added, so that the set standard is more in line with the actual environmental requirements and can be adjusted according to the real-time activity of personnel. The dynamic and accurate regulation of laboratory temperature and humidity is realized, and a more stable and comfortable experimental environment is created.

[0042] The intelligent decision and control module performs dynamic strategy switching, including an unmanned model and a manned model. Unmanned model: adopt model predictive control (MPC) to maintain constant temperature and humidity. Specifically, in the unmanned mode of the laboratory intelligent decision and control module, model predictive control (MPC) is used to maintain constant temperature and humidity. First, collect the temperature and humidity environment data of the laboratory, and construct a mathematical model that accurately describes the dynamic changes of the environment. Based on this model, set the target value of constant temperature and humidity and the constraint conditions such as the allowable fluctuation range of temperature and humidity. Then, use the model to combine the current environmental data to predict the future trend of the laboratory temperature and humidity within a certain period of time. According to the prediction results, considering the target and constraint conditions, the optimal control strategy is calculated by using the optimization algorithm, such as adjusting the operating parameters of air conditioning and ventilation equipment. Then execute the control strategy and continuously monitor the environmental data. Compare the actual monitoring results with the prediction, if there is a deviation, adjust the control strategy in time. Through such continuous prediction, control, feedback and adjustment, the laboratory environment is always stable within the set temperature and humidity range.

[0043] Somebody mode: a multi-objective genetic algorithm is established to target the accuracy of temperature and humidity, energy consumption, and personnel comfort for multi-objective optimization. Specifically, in the somebody mode of the laboratory intelligent decision and control module, a multi-objective genetic algorithm is used to achieve multi-objective optimization to control temperature and humidity, and the process of ultimately obtaining control parameters is as follows: First, the operating parameters of the equipment used to control temperature and humidity are determined as decision variables. Then, these decision variables are encoded into chromosomes, and the chromosomes are composed of genes, with each gene corresponding to the value of a decision variable. The population is then initialized, and a certain number of chromosomes are randomly generated to form the initial population. The gene values of each chromosome are randomly generated within the value range of the decision variables, which ensures the diversity of the population. When evaluating the fitness, a non-dominated sorting is used to find a set of current best Pareto optimal solutions. Non-dominated sorting can distinguish the superiority and inferiority of different solutions and divide the solutions into different levels. Then, selection, crossover, and mutation operations are performed on these solutions. Selection selects individuals with high fitness, crossover exchanges part of the genes to generate new individuals, and mutation randomly changes some gene values to constantly find better Pareto optimal solution sets in each iteration. After each iteration, the termination condition is judged. When the fitness of the population is less than the fitness change threshold, it means that the iteration has converged, at which point the iteration is terminated, and the optimal Pareto optimal solution set is output. Finally, from the optimal Pareto optimal solution set, according to the actual needs and preferences, the most suitable solution is selected, and the gene values in the solution are decoded into the actual values of the decision variables. These actual values are the final control parameters for controlling the operation of the temperature and humidity equipment.

[0044] The anti-mis-touch verification module includes behavior intention recognition and mis-touch protection. Behavior intention recognition: using LSTM to learn and analyze multi-dimensional data such as hand movement trajectory, touch time, and touch position of the operator, to determine whether the operator's behavior intention is normal operation or mis-touch. Mis-touch protection: when mis-touch is determined, prevent mis-operation through a voice confirmation mechanism and require secondary verification. Specifically, in the anti-mis-touch verification module based on large-screen touch screen for temperature and humidity adjustment, behavior intention recognition is the core link. First, high-precision sensing components are integrated into the touch layer of the large screen to collect multi-dimensional data such as hand movement trajectory, touch time, and touch position of the operator in real time. These data are transmitted to the trained long short-term memory (LSTM) model. In the training stage, the model uses a large amount of operation data in real scenarios, including normal temperature and humidity adjustment operations and various possible mis-touch situations, to learn the differences between normal operation and mis-touch in hand movement patterns. When the operator performs touch operations on the large screen, the LSTM model quickly analyzes the real-time data. For example, normal temperature and humidity adjustment operations usually have consecutive and purposeful hand movements, such as precise sliding or clicking in the temperature and humidity value area; while mis-touch often appears as short and random touch. The model determines whether this operation is the intention of normal temperature and humidity adjustment or mis-touch according to these features.

[0045] If the LSTM model determines that it is a false touch, the system immediately starts the false touch protection mechanism. The large screen will pop up a voice interaction window and issue a clear voice prompt: "This operation may be a false touch. Please confirm whether you want to adjust the temperature and humidity." If the operator replies "no" by voice, the system will not make any changes to the temperature and humidity settings, maintaining the current temperature and humidity control state. If the operator replies "yes" by voice, indicating that he has the intention to adjust the temperature and humidity, the system will trigger a secondary verification. The secondary verification uses a dynamic password method, and the large screen will display a randomly generated four-digit password, and at the same time, the operator is required to speak the password by voice. Only when the operator accurately speaks the password, the system will determine that this operation is a valid temperature and humidity adjustment instruction, and then adjust the temperature and humidity according to the operator's operation on the large screen, to ensure the stability and precision control of the laboratory temperature and humidity environment.

[0046] The above is only a preferred embodiment of the present application, and is not intended to limit the present application in other forms. Any skilled person in the art can use the disclosed technical content to make changes or modifications to equivalent embodiments applied to other fields, but any simple modification, equivalent change and modification made on the basis of the technical essence of the present application to the above embodiments without departing from the technical solution content of the present application shall still fall within the protection scope of the present application.

Claims

1. A laboratory automation control system based on an AI model, characterized in that, It includes a multimodal environment perception module, a personnel behavior modeling module, an intelligent decision-making and control module, and an anti-accidental touch verification module; The multimodal environment perception module includes sensor deployment and data fusion; Sensor deployment: Deploy distributed temperature and humidity sensors, airflow sensors, and human body sensing sensors; Data fusion: A flow field coupled Kalman fusion method is used to fuse multi-source data to construct a three-dimensional thermal map of the laboratory environment and monitor temperature and humidity in real time; The personnel behavior modeling module includes behavior recognition and impact quantification; Behavior recognition: Real-time tracking of personnel location using the YOLO algorithm, combined with pressure sensors to identify the intensity of personnel activity; Impact Quantification: Construct metabolic heat production and respiration moisture dissipation functions based on activity intensity, and dynamically calculate the contribution of personnel to temperature and humidity. The intelligent decision-making and control module performs dynamic strategy switching, including unmanned mode and manned mode; Unmanned mode: Model predictive control (MPC) is used to maintain constant temperature and humidity; Humanized mode: Establish a multi-objective genetic algorithm with the goals of temperature and humidity accuracy, energy consumption, and human comfort for multi-objective optimization; The accidental touch prevention verification module includes behavioral intent recognition and accidental touch protection; Behavioral intent recognition: LSTM is used to learn and analyze multi-dimensional data such as the operator's hand movement trajectory, touch time, and touch position to determine whether the operator's behavioral intent is normal operation or accidental touch; Accidental touch protection: When an accidental touch is detected, a voice confirmation mechanism is used to prevent accidental operation, and secondary verification is required; In the personnel behavior modeling module's impact quantification, based on activity intensity, functions for personnel metabolic heat production and respiration moisture dissipation are constructed, and the specific contributions of personnel to temperature and humidity are dynamically calculated, including: The contribution of human metabolic heat production to temperature is calculated using the following formula: Where K is the activity intensity score of the personnel, and M is the heat generation coefficient per unit activity intensity. For the duration of personnel activities, The target temperature preset for the laboratory. This refers to the permissible fluctuation range of laboratory temperature. The contribution of respiration to humidity is calculated as follows: Where D is the moisture dissipation coefficient per unit activity intensity, and H is the actual humidity in the current laboratory. This represents the maximum permissible humidity level in the laboratory. The final laboratory temperature and humidity standards to be set are calculated based on the contribution values ​​of temperature and humidity. When setting the standards, the influence of temperature on humidity is also taken into account to obtain the final laboratory temperature and humidity standards to be set, including the final temperature setting and the final humidity setting. The final temperature setting is calculated as follows: ,in Here, V represents the specific heat capacity of air at constant pressure, and V represents the volume of the laboratory. air density, This is the temperature control compensation coefficient; The final humidity setting is based on the effect of coupled temperature, and the calculation method is as follows: ,in, This represents the target humidity preset by the laboratory. Represents temperature The saturated moisture content below; This represents the compensation range for changes in saturated humidity with temperature.

2. The laboratory automation control system based on an AI model according to claim 1, characterized in that, The specific implementation of the multimodal environment perception module in data fusion to construct a 3D environmental heat map of the laboratory includes state vector extension and flow field model embedding. State vector extension: Introducing airflow velocity vector and air humidity Construct a seven-dimensional state vector: ,in The current air temperature. The relative humidity at the current moment. This is the temperature and humidity gradient vector at the current time t; Flow field model embedding: Predict the temperature and humidity at the next moment by embedding the flow field model. The calculation method is as follows: , ,in It is random noise.

3. The laboratory automation control system based on an AI model according to claim 1, characterized in that, The YOLO algorithm for behavior recognition in the personnel behavior modeling module adopts the YOLOv8 model architecture and combines it with an improved Deformable DETR to improve the localization accuracy in personnel occlusion scenarios. The improvement of Deformable DETR is to integrate spatiotemporal sequence memory, which includes a historical trajectory buffer with a length of 10 frames to store the three-dimensional coordinates and occlusion state of the personnel. When the target is completely occluded, the position of the next frame is predicted by linear interpolation with a prediction error of ≤15cm. At the same time, the occlusion state transition probability is calculated, and the area within 50cm of the historical trajectory is given twice the attention search weight.

4. The laboratory automation control system based on an AI model according to claim 1, characterized in that, The pressure sensors are deployed on the test bench and the floor of the passageway. They collect pressure signals through a distributed array and calculate the rate of change of pressure data, that is, the amount of pressure change per unit time, as a feature of the activity intensity. Using random forest as input, the extracted stress features are used to output a score of the individual's activity intensity.

5. The laboratory automation control system based on an AI model according to claim 1, characterized in that, The specific steps of the multi-objective genetic algorithm optimization are as follows: Decision variables: Determine the operating parameters of the equipment used to control temperature and humidity; Encoding: Encoding decision variables as chromosomes; Population initialization: A certain number of chromosomes are randomly generated to form the initial population. The gene values ​​of each chromosome are randomly generated within the range of the decision variables. Fitness evaluation: The best Pareto optimal solution set is obtained by non-dominated sorting, and then selection crossover mutation operation is performed to find the best Pareto optimal solution set in each iteration; Termination condition judgment: When the fitness of the population is less than the fitness change threshold, the iteration is considered to terminate, and the optimal Pareto optimal solution set is output.

Images