Machine learning-based two-stage detection method for constant temperature air conditioning supply disturbance

Abstract

The present application relates to a kind of two-stage detection methods of constant temperature air conditioning air supply disturbance based on machine learning, method includes the following steps: S1, the data set of constant temperature air conditioning air supply system is constructed, multiple disturbance data are included in the data set, the time series temperature of each measuring point under single disturbance, disturbance type label and disturbance degree;S2, the time series temperature is as input, disturbance type label is as expected output data and carries out the training of disturbance type diagnosis network, obtains first-stage disturbance type diagnosis model;S3, second-stage disturbance degree diagnosis model is obtained by training;S4, actual disturbance time series temperature data is obtained, actual disturbance time series temperature data is input first-stage disturbance type diagnosis model and second-stage disturbance degree diagnosis model, and actual disturbance type and actual disturbance degree are obtained.Compared with prior art, the present application has the advantages that the type and degree of disturbance of constant temperature air conditioning air supply system can be detected in time.

Description

Technical Field

[0001] This invention relates to the technical field of constant temperature air conditioning supply air disturbance detection, and in particular to a two-stage detection method for constant temperature air conditioning supply air disturbance based on machine learning. Background Technology

[0002] Industries such as semiconductor manufacturing and high-energy physics research require indoor environments with high air stability, often needing to achieve high-precision levels (-0.5℃ ≤ ΔT ≤ 0.5℃) or even ultra-high-precision levels (-0.1℃ ≤ ΔT ≤ 0.1°C). For constant-temperature rooms at high-precision levels and below, airflow organization needs to be specifically designed to meet the demands of high-precision temperature control, such as horizontal unidirectional flow, and employing a "small temperature difference, large flow rate" air supply method to reduce the disturbance effect of the air conditioning system. However, during operation, air conditioning systems often still experience various disturbances generated by the air conditioning system itself. For high-precision or ultra-high-precision constant-temperature creation requirements, if these disturbances are not eliminated in time, they can cause irreversible damage to processes and equipment. Rapid diagnosis of disturbances in the air conditioning system can help to quickly and effectively take corresponding measures to eliminate or reduce them.

[0003] However, current perturbation detection methods cannot detect the type and degree of the corresponding perturbation, resulting in insufficient accuracy of perturbation detection and hindering timely adjustment and maintenance. Summary of the Invention

[0004] The purpose of this invention is to provide a two-stage detection method for air supply disturbance of a constant temperature air conditioning system based on machine learning, in order to detect the type and degree of disturbance in the constant temperature air conditioning system in a timely manner.

[0005] The objective of this invention can be achieved through the following technical solutions:

[0006] A two-stage detection method for air supply disturbance in a constant temperature air conditioner based on machine learning, the method includes the following steps:

[0007] S1. Construct a dataset for a constant temperature air conditioning air supply system. The dataset includes multiple disturbance data, which are the time-series temperature, disturbance type label, and disturbance degree of each measuring point under a single disturbance in different operating conditions.

[0008] S2. Using the time-series temperature as input and the perturbation type label as the expected output data, train the perturbation type diagnosis network to obtain the first-level perturbation type diagnosis model.

[0009] S3. Using the time-series temperature and perturbation type label as input and the perturbation level as the expected output data, train the perturbation level diagnosis network to obtain the second-level perturbation level diagnosis model.

[0010] S4. Obtain the actual disturbance time-series temperature data, input the actual disturbance time-series temperature data into the first-level disturbance type diagnostic model to obtain the actual disturbance time-series temperature data with actual disturbance type, input the actual disturbance time-series temperature data with actual disturbance type into the second-level disturbance degree diagnostic model to obtain the actual disturbance type and actual disturbance degree corresponding to the actual disturbance time-series temperature data.

[0011] Furthermore, the specific steps for constructing the dataset of the constant temperature air conditioning supply system are as follows:

[0012] A simulated constant temperature air conditioning air supply system is constructed, which includes a constant temperature room, an air supply system, an internal equipment heat source, and a temperature monitoring system.

[0013] The types and degrees of disturbance in the simulated constant temperature air conditioning supply system are adjusted under various operating conditions. The time-series temperature at each measuring point under different types and degrees of disturbance is obtained. The actual data is used to adjust the time-series temperature at each measuring point under different types and degrees of disturbance. Finally, the data sets of the constant temperature air conditioning supply system are integrated.

[0014] Furthermore, the simulated constant temperature air conditioning air supply system is constructed based on fluid dynamics simulation software.

[0015] Furthermore, each measuring point is located inside the constant temperature chamber.

[0016] Furthermore, the time-series temperature at each measuring point is obtained based on the temperature monitoring system under different disturbance types and disturbance degrees.

[0017] Furthermore, the temperature monitoring system is a temperature sensor array.

[0018] Furthermore, the air supply system, internal equipment heat source, and temperature monitoring system are located inside the constant temperature room.

[0019] Furthermore, the disturbance type labels include supply air temperature fluctuation, supply air volume fluctuation, and cable heat dissipation fluctuation.

[0020] Furthermore, the perturbation type diagnostic network uses the K-means algorithm for machine learning clustering analysis to classify perturbation types and is trained using the backpropagation algorithm.

[0021] Furthermore, the perturbation level diagnosis network is a BP neural network, trained using the backpropagation algorithm.

[0022] Compared with the prior art, the present invention has the following beneficial effects:

[0023] This invention leverages the advantages of machine learning technology in handling nonlinear multi-input problems. It uses machine learning clustering and regression algorithms to perform hierarchical diagnosis of the type and degree of disturbance in a constant temperature air conditioning system based on monitoring data. The two-level diagnostic model obtains the type and degree of disturbance, improving diagnostic accuracy. It also fills the gap in the lack of disturbance conditions in the measured data. The constructed dataset can summarize various disturbance conditions, improving the completeness of the dataset and enabling the detection of the type and degree of disturbance under different conditions. Attached Figure Description

[0024] Figure 1 This is a flowchart of the present invention;

[0025] Figure 2 This is a schematic diagram of signal transmission according to the present invention;

[0026] Figure 3 This is a prediction accuracy diagram of the disturbance type diagnostic model of the present invention;

[0027] Figure 4 This is a graph showing the prediction accuracy of the disturbance level diagnostic model of the present invention;

[0028] In the figure, 1. A constant temperature room containing an internal heat source; 2. A temperature sensor array; 3. A disturbance type diagnostic model; 4. A disturbance degree diagnostic model; 5. A validated constant temperature room CFD model. Detailed Implementation

[0029] The present invention will now be described in detail with reference to the accompanying drawings and specific embodiments. These embodiments are based on the technical solution of the present invention and provide detailed implementation methods and specific operating procedures. However, the scope of protection of the present invention is not limited to the following embodiments.

[0030] This invention establishes a two-level diagnostic model for air supply disturbance in constant temperature air conditioning systems, and utilizes this model to achieve real-time and rapid diagnosis of air supply disturbance, thereby further reducing or eliminating disturbance and improving the temperature control accuracy of the constant temperature air conditioning system. The flowchart of this invention is as follows: Figure 1 As shown.

[0031] The purpose of this invention is to provide a two-level detection method for air supply disturbance in constant temperature air conditioning systems based on machine learning, in order to solve the technical problems existing in the prior art and achieve rapid diagnosis of disturbance in constant temperature air conditioning systems. The technical approach of this invention is as follows: First, a CFD simulation model, verified by actual measurements, is established for a specific constant temperature room, and the indoor temperature and corresponding disturbance datasets under different disturbance conditions are obtained using the simulation model; then, a diagnostic model capable of determining the type of disturbance is constructed using the dataset based on a machine learning clustering algorithm; finally, a diagnostic model capable of determining the degree of disturbance is constructed using the dataset based on a machine learning regression algorithm, thereby constructing a two-level diagnostic model that enables rapid diagnosis of disturbance in constant temperature air conditioning systems.

[0032] This invention mainly includes the following steps:

[0033] S1. Construct a dataset for a constant temperature air conditioning air supply system. The dataset includes multiple disturbance data, which are the time-series temperature, disturbance type label, and disturbance degree of each measuring point under a single disturbance in different operating conditions.

[0034] S2. Using the time-series temperature as input and the perturbation type label as the expected output data, train the perturbation type diagnosis network to obtain the first-level perturbation type diagnosis model.

[0035] S3. Using the time-series temperature and perturbation type label as input and the perturbation level as the expected output data, train the perturbation level diagnosis network to obtain the second-level perturbation level diagnosis model.

[0036] S4. Obtain the actual disturbance time-series temperature data, input the actual disturbance time-series temperature data into the first-level disturbance type diagnostic model to obtain the actual disturbance time-series temperature data with actual disturbance type, input the actual disturbance time-series temperature data with actual disturbance type into the second-level disturbance degree diagnostic model to obtain the actual disturbance type and actual disturbance degree corresponding to the actual disturbance time-series temperature data.

[0037] In S1 of this invention, a dataset of a constant temperature air conditioning supply system is constructed. A simulation model including a constant temperature room, air outlet, internal equipment heat source, etc. is established based on fluid dynamics (CFD) simulation software. The model is then calibrated using monitoring system data of the simulated object to obtain a verified CFD simulation model. Subsequently, step changes are set for supply air temperature, supply air volume, and indoor heat source intensity to simulate the disturbances generated by the constant temperature air conditioning supply system during operation. Considering that the probability of multiple disturbances occurring simultaneously in the supply air system is low, the dataset construction part only considers the effect of a single disturbance.

[0038] Secondly, considering that temperature data is readily available and can effectively reflect indoor air thermal effects, CFD post-processing software was used to set up several virtual temperature measurement points (a total of k points) within the constant temperature room model based on the results of a large number of simulations. The time-series data of each temperature measurement point after the occurrence of disturbances were output. Combined with the type and degree of disturbances under this operating condition, a dataset for subsequent machine learning was constructed. The structure of the dataset is shown below:

[0039] Case 1: {[Time series data of measuring point 1: 21.1℃, 21.2℃, 21.3℃, ...], [Time series data of measuring point 2: 21.1℃, 21.2℃, 21.3℃, ...], [Time series data of measuring point 3: 21.1℃, 21.2℃, 21.3℃, ...], ...[Time series data of measuring point k: 21.1℃, 21.2℃, 21.3℃, ...], Disturbance type label, Disturbance degree};

[0040] Case 2: {[Time series data of measuring point 1: 21.1℃, 21.2℃, 21.3℃, ...], [Time series data of measuring point 2: 21.1℃, 21.2℃, 21.3℃, ...], [Time series data of measuring point 3: 21.1℃, 21.2℃, 21.3℃, ...], ...[Time series data of measuring point k: 21.1℃, 21.2℃, 21.3℃, ...], Disturbance type label, Disturbance degree}; ...

[0042] Case N: {[Time series data of measurement point 1: 21.1℃, 21.2℃, 21.3℃, ...], [Time series data of measurement point 2: 21.1℃, 21.2℃, 21.3℃, ...], [Time series data of measurement point 3: 21.1℃, 21.2℃, 21.3℃, ...], ...[Time series data of measurement point k: 21.1℃, 21.2℃, 21.3℃, ...], Disturbance type label, Disturbance degree}.

[0043] Then, the first-level perturbation type diagnostic model is trained. First, a machine learning clustering analysis algorithm (such as K-means) is selected. The "time-series monitoring data of measurement point 1 to measurement point k" of the aforementioned dataset is used as input, and the "perturbation type label" is used as the expected output data. The dataset is divided into training set, test set, and validation set. Self-learning algorithms such as backpropagation are used to train the model based on the training set. During the training process, the test set data is used to test whether the training has converged. Then, the validation set data is used to evaluate and verify the training effect, thereby establishing a perturbation type diagnostic model with time-series monitoring data of several temperature measurement points as input and perturbation type as output.

[0044] Then, a second-level disturbance level diagnostic model is trained. A machine learning prediction model algorithm (such as a backpropagation neural network) is selected. The time-series monitoring data of measurement points 1 to k from the previously constructed dataset, along with the disturbance type, are used as input. The disturbance level is used as the expected output data. The dataset is divided into training, testing, and validation sets. Self-learning algorithms, such as backpropagation, are used for training based on the training set. During training, the test set data is used to test whether training convergence has been achieved. Afterward, the validation set data is used to evaluate and verify the training effect. This establishes a disturbance type diagnostic model corresponding to different disturbance types, using time-series monitoring data from several temperature measurement points as input and disturbance level as output. Combined with the disturbance diagnostic model established above, a two-level disturbance diagnostic model for a specific constant temperature air conditioning system can be obtained.

[0045] The information transmission process of this invention is as follows: Figure 2 As shown, Figure 2 The invention includes the apparatus used in this invention, comprising a constant temperature room, a temperature monitoring system, a data transmission system, a CFD simulation model, and a two-level diagnostic model: 1. a constant temperature room containing an internal heat source and an air supply system; 2. a temperature sensor array; 3. a disturbance type diagnostic model; 4. a disturbance degree diagnostic model; 5. a validated constant temperature room CFD model, and a dataset for training.

[0046] The constant temperature room containing an internal heat source and an air supply system is the controlled object. The temperature sensor array is installed inside the constant temperature room to collect temperature signals in real time, which are then transmitted to the diagnostic model via a signal transmission system. The validated constant temperature room CFD model is used to perform numerous simulations by changing the type and degree of disturbance to obtain a dataset containing the temperature and disturbance type and degree at the location of the temperature sensor array. The disturbance type diagnostic model is trained using the dataset as a training set, with temperature data as input and disturbance type as the expected output, enabling disturbance type diagnosis based on temperature monitoring data. The disturbance degree diagnostic model is trained using the dataset as a training set, with temperature data and disturbance type as input, and disturbance degree as the expected output, enabling disturbance degree diagnosis based on temperature monitoring data and the disturbance type diagnostic results obtained from the preceding model.

[0047] In this embodiment, a CFD numerical simulation model is established for a specific constant-temperature construction scenario, including a constant-temperature room, an air supply system, and an internal heat source. The main disturbance types are selected as air supply temperature fluctuation, air supply volume fluctuation, and cable heat dissipation fluctuation, with fluctuation amplitudes shown in Table 1. In the CFD numerical simulation, the preset fluctuation level is achieved by changing the boundary conditions, covering all disturbance types and levels. Large-scale numerical simulation calculations are performed, and the calculation results are compiled into a dataset that can be used for training. Then, temperature data from the obtained dataset is used as input, and disturbance type labels are used as the expected output. A first-level disturbance type diagnostic model is constructed using the K-means clustering analysis algorithm. After training, the results show that the diagnostic accuracy of the training set is above 94%, and the diagnostic accuracy of the test set is above 86%. Figure 3 As shown. Then, using the temperature data and disturbance type labels from the obtained dataset as input, and the degree of fluctuation as the expected output, a second-level disturbance degree diagnostic model is constructed using a feedforward (BP) neural network regression algorithm. After training, the results show that the regression accuracy (R-squared) for diagnosing the degree of the three disturbances is above 0.96, achieving high regression accuracy, such as... Figure 4 As shown in the figure, the two-level diagnostic model for air supply disturbance of a constant temperature air conditioner constructed in this embodiment, based on machine learning, can utilize temperature data collected by a real-time monitoring system as input when disturbance occurs during actual operation. The first-level disturbance type diagnostic model outputs the type of disturbance present at this time. Then, using the monitored temperature data and the disturbance type diagnostic results obtained from the first-level diagnostic model, the second-level disturbance degree diagnostic model outputs the degree of disturbance effect. This effectively assists maintenance personnel in making timely adjustments and repairs, reducing the impact of disturbance, and improving the temperature control accuracy of the constant temperature room.

[0048] Table 1. Types and Degrees of Isothermal Disturbance

[0049]

[0050] This invention can establish a two-level diagnostic model that separately diagnoses the type and degree of disturbance. When disturbances occur during actual operation, the model can diagnose the type and degree of disturbance in real time, thereby further guiding maintenance personnel to perform targeted maintenance, reduce the impact of disturbances, and thus continuously ensure the system temperature control accuracy.

[0051] The present invention achieves the following technical effects compared to the prior art:

[0052] 1. This invention can leverage the advantages of machine learning technology in handling nonlinear multi-input problems. It uses machine learning clustering and regression algorithms to classify and diagnose the type and degree of disturbance in the constant temperature air conditioning air supply system based on monitoring data, thereby improving diagnostic accuracy.

[0053] 2. The training dataset for the two-level diagnostic model of the present invention is obtained by CFD numerical simulation after being corrected and verified by measured data, thereby filling the gap in the lack of perturbation conditions in the measured data. The constructed dataset can summarize various perturbation conditions and improve the completeness of the dataset.

[0054] 3. This invention can be combined with the monitoring system of constant temperature air conditioner, and can realize disturbance diagnosis by relying solely on temperature sensors, which reduces investment costs and enables real-time diagnosis, helping maintenance personnel to make timely adjustments and maintenance.

[0055] The preferred embodiments of the present invention have been described in detail above. It should be understood that those skilled in the art can make numerous modifications and variations based on the concept of the present invention without creative effort. Therefore, all technical solutions that can be obtained by those skilled in the art based on the concept of the present invention through logical analysis, reasoning, or limited experimentation on the basis of existing technology should be within the scope of protection defined by the claims.

Claims

1. A two-stage detection method for air supply disturbance in a constant temperature air conditioner based on machine learning, characterized in that, The method includes the following steps: S1. Construct a dataset for a constant temperature air conditioning air supply system. The dataset includes multiple disturbance data, which are the time-series temperature, disturbance type label, and disturbance degree of each measuring point under a single disturbance in different operating conditions. S2. Using the time-series temperature as input and the perturbation type label as the expected output data, train the perturbation type diagnosis network to obtain the first-level perturbation type diagnosis model. S3. Using the time-series temperature and perturbation type label as input and the perturbation level as the expected output data, train the perturbation level diagnosis network to obtain the second-level perturbation level diagnosis model. S4. Obtain the actual disturbance time-series temperature data, input the actual disturbance time-series temperature data into the first-level disturbance type diagnostic model to obtain the actual disturbance time-series temperature data with actual disturbance type, input the actual disturbance time-series temperature data with actual disturbance type into the second-level disturbance degree diagnostic model to obtain the actual disturbance type and actual disturbance degree corresponding to the actual disturbance time-series temperature data.

2. The two-stage detection method for air supply disturbance of a constant temperature air conditioner based on machine learning according to claim 1, characterized in that, The specific steps for constructing the dataset of the constant temperature air conditioning supply system are as follows: A simulated constant temperature air conditioning air supply system is constructed, which includes a constant temperature room, an air supply system, an internal equipment heat source, and a temperature monitoring system. The types and degrees of disturbance in the simulated constant temperature air conditioning supply system are adjusted under various operating conditions. The time-series temperature at each measuring point under different types and degrees of disturbance is obtained. The actual data is used to adjust the time-series temperature at each measuring point under different types and degrees of disturbance. Finally, the data sets of the constant temperature air conditioning supply system are integrated.

3. The two-stage detection method for air supply disturbance of a constant temperature air conditioner based on machine learning according to claim 2, characterized in that, The simulated constant temperature air conditioning air supply system is built based on fluid dynamics simulation software.

4. The two-stage detection method for air supply disturbance of a constant temperature air conditioner based on machine learning according to claim 2, characterized in that, Each measuring point is located inside the constant temperature chamber.

5. The two-stage detection method for air supply disturbance of a constant temperature air conditioner based on machine learning according to claim 2, characterized in that, The temperature monitoring system is used to obtain the time-series temperature at each measuring point under different disturbance types and disturbance degrees.

6. The two-stage detection method for air supply disturbance of a constant temperature air conditioner based on machine learning according to claim 5, characterized in that, The temperature monitoring system is a temperature sensor array.

7. A two-stage detection method for air supply disturbance of a constant temperature air conditioner based on machine learning according to claim 2, characterized in that, The air supply system, internal equipment heat source, and temperature monitoring system are located inside the constant temperature room.

8. The two-stage detection method for air supply disturbance of a constant temperature air conditioner based on machine learning according to claim 1, characterized in that, The disturbance type labels include supply air temperature fluctuation, supply air volume fluctuation, and cable heat dissipation fluctuation.

9. A two-stage detection method for air supply disturbance in a constant temperature air conditioner based on machine learning, as described in claim 1, is characterized in that... The perturbation type diagnostic network uses the K-means algorithm for machine learning clustering analysis to classify perturbation types and is trained using the backpropagation algorithm.

10. A two-stage detection method for air supply disturbance in a constant temperature air conditioner based on machine learning, as described in claim 1, characterized in that, The perturbation level diagnosis network is a BP neural network, trained using the backpropagation algorithm.

Images