Ventilation system adjustment strategy optimization method and system for amc concentration

Abstract

The application relates to the technical field of ventilation system control, and particularly discloses a ventilation system adjustment strategy optimization method and system for AMC concentration, which comprises the following steps: receiving operation data fed back by a ventilation device based on a monitor built in the ventilation device, analyzing the operation data, and generating a state diagram; displaying the state diagram, receiving a control instruction sent by a worker, and training a neural network model according to the control instruction; and generating a control instruction of the ventilation device based on the trained neural network model when the error rate of the neural network model is less than a preset threshold. According to the application, the working state of the ventilation device is visually displayed based on the monitor built in the ventilation device, a state diagram is obtained, a control instruction input by a worker is received, and a neural network model of the state diagram to the control instruction is trained at the same time. When the error rate of the neural network model is high enough, an automatic control process can be realized.

Description

Technical Field

[0001] The present invention relates to the technical field of ventilation system control, and in particular to an AMC concentration-oriented ventilation system adjustment strategy optimization method and system. Background Art

[0002] AMC pollutants (airborne molecular contaminants) refer to pollutants that exist in gaseous form in the air. They primarily affect the semiconductor manufacturing and microelectronics industries. Cleanrooms are typically the site of demand for AMC pollutant concentrations. Ventilation equipment is installed in cleanrooms to extract AMC pollutants from the cleanroom and then filter and exchange filtered air into the cleanroom. Due to the extremely small size of AMC pollutants, the requirements for ventilation equipment are very high, and the cost of operating ventilation equipment is also relatively high. Therefore, real-time control is required based on actual conditions.

[0003] The existing control method is to use the existing AMC monitor to obtain the AMC pollutant concentration at different points in the clean room, and then manually control it. This control method is feasible, but it is not very targeted and the control process of the ventilation equipment is not precise enough. How to improve the control precision of the ventilation equipment, optimize the operation process of the ventilation system, and increase its lifespan to improve the air quality of the clean room in disguise is the technical problem that the technical solution of the present invention aims to solve. Summary of the Invention

[0004] The object of the present invention is to provide a ventilation system adjustment strategy optimization method and system for AMC concentration to solve the problems raised in the above background technology.

[0005] To achieve the above object, the present invention provides the following technical solutions:

[0006] A ventilation system adjustment strategy optimization method and system for AMC concentration, the method comprising:

[0007] Obtaining a scene graph of the ventilation scene, and collecting statistics on the distribution of ventilation equipment based on the scene graph;

[0008] receiving operation data fed back by the ventilation equipment based on a monitor built into the ventilation equipment, analyzing the operation data, and generating a state diagram;

[0009] Display a state diagram, receive control instructions sent by a staff member, and train a neural network model according to the control instructions;

[0010] When the error rate of the neural network model is less than a preset threshold, a control instruction for the ventilation equipment is generated based on the trained neural network model;

[0011] The generation frequency of the state diagram is adjusted according to the error rate of the neural network model, and the generation frequency is used to update a display update period of the state diagram.

[0012] As a further scheme of the present application, the step of obtaining a scene graph of the ventilation scene comprises:

[0013] A three-dimensional map of the ventilation scene is obtained as the scene graph.

[0014] The origin position, three-dimensional model and installation information of the built-in monitor of the ventilation equipment are queried; the installation information of the monitor comprises the origin position and three-dimensional model of the monitor.

[0015] The origin positions and three-dimensional models of all the ventilation equipment are counted based on the scene graph, and a device model corresponding to the ventilation equipment is created.

[0016] The origin positions and three-dimensional models of the monitors are counted based on the device model, and a monitor model corresponding to the monitor is created.

[0017] As a further scheme of the present application, the step of receiving the running data of the ventilation equipment by the monitor built in the ventilation equipment, analyzing the running data and generating a state diagram comprises:

[0018] A connection channel with the monitor built in the ventilation equipment is established, and the running data is obtained based on the connection channel.

[0019] The running data is analyzed to determine the risk radius of each monitor model.

[0020] The device model is expanded according to the risk radius to obtain a risk area.

[0021] After all the device models are expanded, a state diagram is obtained.

[0022] The running data is collected and comprises a monitor label and a time label.

[0023] As a further scheme of the present application, the step of analyzing the running data and determining the risk radius of each monitor model comprises:

[0024] The device model of the ventilation equipment and the monitor model therein are queried, and a three-dimensional matrix is created according to the relative positions of the monitor models; the row and column positions in the three-dimensional matrix correspond to the monitors one by one.

[0025] For the running data at the same time, the row and column positions are queried according to the monitor label in the running data, the running data is inserted into the corresponding row and column positions, and a three-dimensional matrix at the time is obtained.

[0026] Arrange the three-dimensional matrix in chronological order, extract the row and column positions in the arranged three-dimensional matrix, obtain the data at each moment at the row and column positions, and obtain an array with the row and column positions as labels;

[0027] Analyze the array of the extracted row and column positions to determine the risk radius of the monitor model corresponding to the row and column positions;

[0028] Query the monitor type and modify the risk radius based on the type;

[0029] The type of monitor includes at least a monitor for collecting AMC concentration. When the type of monitor is a monitor for collecting AMC concentration, the correction coefficient thereof takes the maximum value.

[0030] As a further solution of the present invention, the step of analyzing the array of extracted row and column positions to determine the risk radius of the monitor model corresponding to the row and column positions includes:

[0031] Performing discrete Fourier transform on the array to obtain a spectrum graph;

[0032] Calculate the amplitude of each frequency in the spectrum graph and determine the risk radius based on the amplitude of each frequency;

[0033] Obtain the maximum amplitude of the frequency and adjust the extraction period of the row and column positions according to the maximum amplitude;

[0034] The process of determining the risk radius is as follows:

[0035] Where, is the risk radius, is the preset correction factor, For the The amplitude corresponding to each frequency point; is the total number of frequency points;

[0036] The process of determining the extraction period is:

[0037] Where, is the row and column position The extraction cycle at is the preset standard cycle, is the preset correction factor.

[0038] As a further solution of the present invention: the step of adjusting the generation frequency of the state diagram according to the error rate of the neural network model, wherein the generation frequency is used to update the display update period of the state diagram, includes:

[0039] When receiving a control instruction sent by a staff member, the control instruction sent by the staff member is regarded as a real instruction;

[0040] The state diagram is input into the trained neural network model, and a control instruction is output as a virtual instruction.

[0041] The similarity between the virtual instruction and the real instruction is calculated, and a display update period of the state diagram is determined according to the similarity; the display update period is directly proportional to the similarity.

[0042] The technical scheme of the present application also provides an air ventilation system adjustment strategy optimization system for an AMC concentration, which comprises:

[0043] A distribution information statistical module is configured to acquire a scene graph of an air ventilation scene and to statistically acquire distribution information of air ventilation equipment according to the scene graph.

[0044] A state diagram generation module is configured to receive operation data fed back by the air ventilation equipment based on a monitor built in the air ventilation equipment, to analyze the operation data, and to generate a state diagram.

[0045] A model training module is configured to display the state diagram, to receive a control instruction sent by a worker, and to train a neural network model according to the control instruction.

[0046] A control instruction generation module is configured to generate a control instruction of the air ventilation equipment based on the trained neural network model when an error rate of the neural network model is less than a preset threshold.

[0047] A recursive update module is configured to adjust a generation frequency of the state diagram according to the error rate of the neural network model, and the generation frequency is used to update a display update period of the state diagram.

[0048] As a further scheme of the present application, the distribution information statistical module comprises:

[0049] A scene graph acquisition unit is configured to acquire a three-dimensional map of an air ventilation scene as a scene graph.

[0050] An information query unit is configured to query an original position, a three-dimensional model and installation information of a built-in monitor of the air ventilation equipment; the installation information of the monitor comprises an original position and a three-dimensional model of the monitor.

[0051] An equipment model construction unit is configured to statistically acquire the original position and the three-dimensional model of all the air ventilation equipment based on the scene graph, and to create an equipment model corresponding to the air ventilation equipment.

[0052] A monitor model construction unit is configured to statistically acquire the original position and the three-dimensional model of the monitor based on the equipment model, and to create a monitor model corresponding to the monitor.

[0053] As a further scheme of the present application, the state diagram generation module comprises:

[0054] An operation data acquisition unit, configured to establish a connection channel with a monitor built into the ventilation equipment and acquire operation data based on the connection channel;

[0055] an operating data analysis unit, configured to analyze the operating data and determine a risk radius of each monitor model;

[0056] A device model expansion unit, configured to expand the device model according to the risk radius to obtain a risk area;

[0057] A state diagram output unit is used to obtain a state diagram after all device models are expanded;

[0058] The operation data is collected data containing a monitor tag and a time tag.

[0059] As a further solution of the present invention: the recursive update module includes:

[0060] a real instruction determining unit, configured to, upon receiving a control instruction sent by a staff member, regard the control instruction sent by the staff member as a real instruction;

[0061] a virtual instruction generation unit, configured to input the state diagram into the trained neural network model and output control instructions as virtual instructions;

[0062] The cycle update unit is used to calculate the similarity between the virtual instruction and the real instruction, and determine the display update cycle of the state diagram according to the similarity; the display update cycle is proportional to the similarity.

[0063] Compared with the prior art, the present invention has the following beneficial effects:

[0064] The present invention visualizes the working status of the ventilation equipment based on the monitor built into the ventilation equipment, obtains a state diagram, receives control instructions input by the staff, and at the same time trains a neural network model from the state diagram to the control instructions. When the error rate of the neural network model is high enough, an automated control process can be realized, thereby improving the real-time performance of the control process. BRIEF DESCRIPTION OF THE DRAWINGS

[0065] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention.

[0066] Figure 1 The overall flow chart of the ventilation system adjustment strategy optimization method for AMC concentration is shown.

[0067] Figure 2The structural diagram of the ventilation system adjustment strategy optimization system for AMC concentration is shown. DETAILED DESCRIPTION

[0068] In order to make the technical problems, technical solutions and beneficial effects to be solved by the present invention more clearly understood, the present invention is further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0069] Figure 1 The figure is a flowchart of a ventilation system adjustment strategy optimization method for AMC concentration. In an embodiment of the present invention, a ventilation system adjustment strategy optimization method for AMC concentration is provided. The method includes:

[0070] Step S100: obtaining a scene graph of a ventilation scene, and collecting statistics on the distribution of ventilation equipment according to the scene graph;

[0071] The ventilation scene involved in the AMC concentration monitoring process is generally a clean room. The three-dimensional BIM model of the clean room is the scene diagram. The ventilation equipment will transfer the gas containing AMC pollutants in the room to the outside, filter the outside gas, and then replace it with the clean room; obtain the scene diagram of the ventilation scene, and count the distribution information of the ventilation equipment based on the scene diagram. The ventilation equipment is a device that exchanges air with the outside world, which contains filtering components, ventilation components and monitoring components. The distribution information is the location of the ventilation equipment.

[0072] Step S200: receiving operation data fed back by the ventilation equipment based on a monitor built into the ventilation equipment, analyzing the operation data, and generating a state diagram;

[0073] The ventilation equipment in the technical solution of the present invention is a highly intelligent device, which is equipped with multiple monitors. The monitors include at least sensors and controllers with data transmission functions. The data generated by these monitors during the operation of the ventilation equipment is the operation data. By analyzing the operation data, the operation status of the ventilation equipment can be obtained, and the operation status is inserted into the scene graph to obtain a state graph; wherein, the operation status can be some display parameters. For example, when the operation data exceeds the preset standard threshold, a red layer is inserted near the corresponding monitor, and the red layer is displayed flashing, indicating that the position is in an abnormal state.

[0074] Step S300: displaying a state diagram, receiving control instructions sent by a staff member, and training a neural network model according to the control instructions;

[0075] After the state diagram is generated, the state diagram is displayed. Since the state diagram itself is a three-dimensional diagram, it is somewhat similar to a three-dimensional model. It can be directly displayed with the help of existing three-dimensional display software. The display process is essentially an interactive process, which is used to inform the staff of the current working status of each ventilation equipment. After the staff knows the working status of each ventilation equipment, they will input control instructions to control the working process of the ventilation equipment; about the process from the state diagram to the control instruction, the most accurate one is the control instruction input by the staff. On this basis, the present invention regards the state diagram as a feature, that is, an independent variable, and the control instruction as an output, that is, a dependent variable. Every time a control instruction is obtained, a sample from the state diagram to the control instruction is constructed, and all samples are counted to obtain a sample set. When the number of samples is large enough, the neural network model is trained based on the sample set, so that no artificial parameters are required, and an intelligent generation process of the control instruction is obtained.

[0076] It is worth mentioning that the types of control instructions are generally limited. For control instructions, staff can first quantify them and number each control instruction. At this time, they can first train the state diagram to the numbered model, and then introduce a numbered to the control instruction database query and reading model.

[0077] Step S400: When the error rate of the neural network model is less than a preset threshold, a control instruction for the ventilation equipment is generated based on the trained neural network model;

[0078] When the error rate of the neural network model is small enough, the neural network model can be put into use. At this time, the working process of the monitor is controlled by the program, and the neural network model is also controlled by the program. The application scenario of the entire ventilation equipment will be in an unmanned automatic operation state.

[0079] Step S500: adjusting the generation frequency of the state diagram according to the error rate of the neural network model, wherein the generation frequency is used to update the display update period of the state diagram.

[0080] On the basis that the error rate of the neural network model is relatively small and it is put into use, the present invention does not completely eliminate the control instruction input process of the staff, but gradually reduces the proportion of the control instruction input process of the staff. Specifically, the display frequency of the state diagram is adjusted according to the error rate of the neural network model. The smaller the error rate, the smaller the display frequency. Correspondingly, the frequency of the staff inputting control instructions is also smaller. Its actual meaning is that the time interval between two adjacent input control instructions becomes longer.

[0081] Regarding step S100, the step of obtaining a scene graph of the ventilation scene and collecting statistics on the distribution information of the ventilation equipment according to the scene graph includes:

[0082] Obtain a three-dimensional map of the ventilation scene as a scene graph;

[0083] Querying the origin position, three-dimensional model and installation information of the built-in monitor of the ventilation equipment; the installation information of the monitor includes the origin position and three-dimensional model of the monitor;

[0084] Based on the scene graph, the origin positions and 3D models of all ventilation equipment are counted, and equipment models corresponding to the ventilation equipment are created;

[0085] The origin position of the monitor and its three-dimensional model are statistically analyzed based on the device model, and a monitor model corresponding to the monitor is created.

[0086] In an example of the technical solution of the present invention, the generation and application process of the scene graph is explained. The BIM model of the clean room is obtained and simplified. For example, only the lines are retained. The resulting three-dimensional graph is called a three-dimensional map, which is used as the scene graph. On this basis, the origin position, three-dimensional model and installation information of the built-in monitor of the ventilation equipment are queried. The installation information of the monitor includes the origin position and three-dimensional model of the monitor. These are all known data and can be directly queried and read. It is worth mentioning that regarding the concept of the origin position, since both the ventilation equipment and the monitor are objects with volume, it is difficult to obtain their position. Therefore, the present invention simplifies it to the origin position, that is, the reference point among these objects, such as the center point of a device, which will be determined when modeling these objects.

[0087] Then, based on the scene graph, the data of all ventilation equipment are counted, and a scene graph including the equipment model is constructed. Then, based on each equipment model, the data of the monitor in the equipment model is counted, and the monitor model is inserted into the equipment model. This is a step-by-step process and is not complicated to implement.

[0088] Regarding step S200, the steps of receiving the operating data fed back by the ventilation equipment based on the monitor built into the ventilation equipment, analyzing the operating data, and generating a state diagram include:

[0089] Establishing a connection channel with a monitor built into the ventilation equipment and acquiring operation data based on the connection channel;

[0090] Analyzing the operating data to determine the risk radius of each monitor model;

[0091] Expanding the equipment model according to the risk radius to obtain a risk area;

[0092] When all device models are expanded, the state diagram is obtained;

[0093] The operation data is collected data containing a monitor tag and a time tag.

[0094] In an example of the technical solution of the present application, the generation process of the state diagram is described. For any ventilation equipment, the monitor therein is inquired, a connection channel with the monitor is established, data transmitted by the monitor is acquired based on the connection channel, which is called operation data. The operation data is identified to determine the working state of each monitor and whether the working state is stable, which is indicated by the risk radius. The device model is expanded according to the risk radius to obtain an expanded device model. When all the device models are expanded, the state diagram is obtained.

[0095] It should be noted that the monitor contains an AMC concentration monitoring device. This monitor is a device that directly reflects the AMC concentration, and it is the most important. Other monitors, such as current monitors, reflect the working parameters of the ventilation equipment, and are slightly less important. When determining the risk radius, a correction coefficient is determined according to the type of the monitor to correct the risk radius.

[0096] The above technical solution essentially limits the display method. Whether there is a risk is indicated by the increase or decrease of the device model, which is more understandable than the color-based display process. There is no need to remember the color representing the state in advance, and the phenomenon that color-blind people cannot identify is also avoided.

[0097] It should be noted that the operation data is acquisition data containing a monitor label and a time label. The monitor label and the time label indicate which position in the device model and which time the operation data corresponds to. Correspondingly, the obtained state diagram also contains a time label.

[0098] Further, the step of analyzing the operation data to determine the risk radius of each monitor model comprises:

[0099] The device model of the ventilation equipment and the monitor model therein are inquired, and a three-dimensional matrix is created according to the relative positions of the monitor models. The row and column positions in the three-dimensional matrix correspond to the monitors one by one.

[0100] For the operation data at the same time, the row and column positions are inquired according to the monitor label in the operation data, the operation data is inserted into the corresponding row and column positions, and a three-dimensional matrix at the time is obtained.

[0101] The three-dimensional matrix is arranged according to the time sequence, the row and column positions are extracted from the arranged three-dimensional matrix, the data at each time at the row and column positions is acquired, and an array with the row and column positions as labels is obtained.

[0102] The array of the extracted row and column positions is analyzed to determine the risk radius of the monitor model corresponding to the row and column positions.

[0103] The type of the monitor is inquired, and the risk radius is corrected according to the type.

[0104] The type of monitor includes at least a monitor for collecting AMC concentration. When the type of monitor is a monitor for collecting AMC concentration, the correction coefficient thereof takes the maximum value.

[0105] In an example of the technical solution of the present invention, the process of generating the risk radius is explained. The equipment model of the ventilation equipment and the monitor model therein are queried, and a three-dimensional matrix is ​​created according to the relative position of the monitor model. This process requires three preset directions, such as east, south and upward. Starting from a corner point at the top of the equipment model, the monitors are read in sequence along the above directions. East, south and upward correspond to right, forward and downward in the matrix respectively (they can be set freely, and once set, they need to be unified). The first monitor is taken as the (1, 1, 1) position, and the monitors are read in sequence to establish the relationship between the monitor and the row and column positions in the three-dimensional matrix, thereby obtaining the three-dimensional matrix.

[0106] When the operation data is obtained, the operation data obtained by each monitor at the same time is inserted into the three-dimensional matrix to obtain a three-dimensional matrix containing data, which is called the three-dimensional matrix at that moment.

[0107] Arrange the three-dimensional matrix in chronological order. After the arrangement is completed, the data at each moment can be easily queried at each row and column position. Extract some row and column positions in the three-dimensional matrix, and query the data at each moment at the row and column positions in the arranged three-dimensional matrix (generally a time range is set, starting from the current moment and working backwards for a preset duration), to obtain the array at the row and column positions.

[0108] Finally, the array is analyzed to analyze the working status of the monitor, and then the monitor model is updated. After the monitors corresponding to the extracted row and column positions are updated, the updated device model is obtained.

[0109] Specifically, the step of analyzing the array of extracted row and column positions to determine the risk radius of the monitor model corresponding to the row and column positions includes:

[0110] Performing discrete Fourier transform on the array to obtain a spectrum graph;

[0111] Calculate the amplitude of each frequency in the spectrum graph and determine the risk radius based on the amplitude of each frequency;

[0112] Obtain the maximum amplitude of the frequency and adjust the extraction period of the row and column positions according to the maximum amplitude;

[0113] In an example of the technical solution of the present invention, the recognition process of the array is explained. The data in the array is the data of a monitor (corresponding to a row and column position) at each moment, which is itself discrete data. Discrete Fourier transform is performed on the discrete data to obtain a spectrum graph. The risk radius can be determined by analyzing the spectrum graph. At the same time, the extraction period of each row and column position can also be adjusted according to the spectrum graph, so that the extraction frequency of each row and column position is different.

[0114] The process of determining the risk radius is as follows:

[0115] Where, is the risk radius, is the preset correction factor, For the The amplitude corresponding to each frequency point; is the total number of frequency points.

[0116] The risk radius actually considers whether there are frequency components with higher amplitudes in the spectrum. If so, it means that it has periodicity. At this time, the stability is higher and the risk radius is smaller. The parameter for judging whether there are frequency components with higher amplitudes is If the amplitude of this item is much higher than that of other frequency components, the value will be very large. In fact, the variance can also be used, but the present application provides a more advanced calculation method, which is to compound one of the items with an exponential function. The exponential function changes very quickly and can better highlight the presence of a frequency component with a higher amplitude.

[0117] The process of determining the extraction period is:

[0118] Where, is the row and column position The extraction cycle at is the preset standard cycle, is the preset correction factor.

[0119] For the extraction cycle, each row and column position has an extraction cycle. The smaller the risk radius corresponding to the row and column position, the more stable the corresponding position is. At this time, the extraction cycle is larger.

[0120] Regarding step S500, the step of adjusting the generation frequency of the state diagram according to the error rate of the neural network model, wherein the generation frequency is used to update the display update period of the state diagram, includes:

[0121] When receiving a control instruction sent by a staff member, the control instruction sent by the staff member is regarded as a real instruction;

[0122] The state diagram is input into the trained neural network model and the control instructions are output as virtual instructions;

[0123] The similarity between the virtual instruction and the real instruction is calculated, and a display update period of the state diagram is determined according to the similarity; the display update period is proportional to the similarity.

[0124] In an example of the technical solution of the present invention, the process of the staff inputting control instructions is specifically limited. Whenever a control instruction sent by the staff is received, it is regarded as a trigger signal, and the control instruction sent by the staff is regarded as a real instruction. At the same time, the neural network model does not stop working. The neural network model also works. It will also output a control instruction based on the state diagram, called a virtual instruction, and calculate the similarity between the virtual instruction and the real instruction. The display period of the state diagram is determined based on the similarity. The higher the similarity, the higher the accuracy of the neural network model. At this time, the staff's intervention can be less, which is reflected in the specific parameters, that is, the display period of the state diagram.

[0125] Figure 2 The structure diagram of the ventilation system adjustment strategy optimization system for AMC concentration is shown. In a preferred embodiment of the technical solution of the present invention, a ventilation system adjustment strategy optimization system for AMC concentration is also provided. The system 10 includes:

[0126] A distribution information statistics module 11 is used to obtain a scene graph of the ventilation scene and to collect statistics on the distribution information of the ventilation equipment according to the scene graph;

[0127] A state diagram generating module 12 is configured to receive operating data fed back by the ventilation equipment based on a monitor built into the ventilation equipment, analyze the operating data, and generate a state diagram;

[0128] The model training module 13 is used to display the state diagram, receive control instructions sent by the staff, and train the neural network model according to the control instructions;

[0129] A control instruction generating module 14 is configured to generate a control instruction for the ventilation device based on the trained neural network model when the error rate of the neural network model is less than a preset threshold;

[0130] The recursive updating module 15 is used to adjust the generation frequency of the state diagram according to the error rate of the neural network model, and the generation frequency is used to update the display update period of the state diagram.

[0131] Furthermore, the distribution information statistics module 11 includes:

[0132] A scene graph acquisition unit, used to acquire a three-dimensional map of the ventilation scene as a scene graph;

[0133] An information query unit, configured to query the origin position, three-dimensional model, and installation information of a built-in monitor of the ventilation equipment; the installation information of the monitor includes the origin position and three-dimensional model of the monitor;

[0134] A device model building unit is used to count the origin positions and three-dimensional models of all ventilation devices based on the scene graph and create a device model corresponding to the ventilation devices;

[0135] The monitor model building unit is used to calculate the origin position of the monitor and its three-dimensional model based on the equipment model, and create a monitor model corresponding to the monitor.

[0136] Specifically, the state diagram generating module 12 includes:

[0137] An operation data acquisition unit, configured to establish a connection channel with a monitor built into the ventilation equipment and acquire operation data based on the connection channel;

[0138] an operating data analysis unit, configured to analyze the operating data and determine a risk radius of each monitor model;

[0139] A device model expansion unit, configured to expand the device model according to the risk radius to obtain a risk area;

[0140] A state diagram output unit is used to obtain a state diagram after all device models are expanded;

[0141] The operation data is collected data containing a monitor tag and a time tag.

[0142] Furthermore, the recursive update module 15 includes:

[0143] a real instruction determining unit, configured to, upon receiving a control instruction sent by a staff member, regard the control instruction sent by the staff member as a real instruction;

[0144] a virtual instruction generation unit, configured to input the state diagram into the trained neural network model and output control instructions as virtual instructions;

[0145] The cycle update unit is used to calculate the similarity between the virtual instruction and the real instruction, and determine the display update cycle of the state diagram according to the similarity; the display update cycle is proportional to the similarity.

[0146] The above are only preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Any equivalent structure or equivalent process transformation made using the contents of the present invention description and drawings, or directly or indirectly applied in other related technical fields, are also included in the scope of protection of the present invention.

Claims

1. A ventilation system adjustment strategy optimization method for AMC concentration, characterized by: The method comprises: Obtaining a scene graph of the ventilation scene, and collecting statistics on the distribution of ventilation equipment based on the scene graph; Receive operation data fed back by the ventilation equipment based on a monitor built into the ventilation equipment, analyze the operation data, and generate a state diagram; Displaying a state diagram, receiving control instructions sent by a staff member, and training a neural network model based on the state diagram and the control instructions; When the error rate of the neural network model is less than a preset threshold, a control instruction for the ventilation equipment is generated based on the trained neural network model; The generation frequency of the state diagram is adjusted according to the error rate of the neural network model, and the generation frequency is used to update the display update period of the state diagram.

2. The ventilation system adjustment strategy optimization method for AMC concentration according to claim 1 is characterized in that: The step of obtaining a scene graph of the ventilation scene and collecting statistics on the distribution information of the ventilation equipment according to the scene graph includes: Obtain a three-dimensional map of the ventilation scene as a scene graph; Querying the origin position, three-dimensional model and installation information of the built-in monitor of the ventilation equipment; the installation information of the monitor includes the origin position and three-dimensional model of the monitor; Based on the scene graph, the origin positions and 3D models of all ventilation equipment are counted, and equipment models corresponding to the ventilation equipment are created; The origin position of the monitor and its three-dimensional model are statistically analyzed based on the device model, and a monitor model corresponding to the monitor is created.

3. The ventilation system adjustment strategy optimization method for AMC concentration according to claim 1 is characterized in that: The step of receiving the operating data fed back by the ventilation equipment based on the monitor built into the ventilation equipment, analyzing the operating data, and generating the state diagram includes: Establishing a connection channel with a monitor built into the ventilation equipment and acquiring operation data based on the connection channel; Analyzing the operating data to determine the risk radius of each monitor model; Expanding the equipment model according to the risk radius to obtain a risk area; When all device models are expanded, the state diagram is obtained; The operation data is collected data containing a monitor tag and a time tag.

4. The ventilation system adjustment strategy optimization method for AMC concentration according to claim 3 is characterized in that: The step of analyzing the operating data to determine the risk radius of each monitor model includes: Query the equipment model of the ventilation equipment and the monitor model therein, and create a three-dimensional matrix based on the relative positions of the monitor models; the row and column positions in the three-dimensional matrix correspond one-to-one with the monitors; For the operation data at the same time, query the row and column positions according to the monitor tags in the operation data, insert the operation data into the corresponding row and column positions, and obtain the three-dimensional matrix at that time; Arrange the three-dimensional matrix in chronological order, extract the row and column positions in the arranged three-dimensional matrix, obtain the data at each moment at the row and column positions, and obtain an array with the row and column positions as labels; Analyze the array of the extracted row and column positions to determine the risk radius of the monitor model corresponding to the row and column positions; Query the monitor type and modify the risk radius based on the type; The type of monitor includes at least a monitor for collecting AMC concentration. When the type of monitor is a monitor for collecting AMC concentration, the correction coefficient thereof takes the maximum value.

5. The ventilation system adjustment strategy optimization method for AMC concentration according to claim 4 is characterized in that: The step of analyzing the array of extracted row and column positions to determine the risk radius of the monitor model corresponding to the row and column positions includes: Performing discrete Fourier transform on the array to obtain a spectrum graph; Calculate the amplitude of each frequency in the spectrum graph and determine the risk radius based on the amplitude of each frequency; Obtain the maximum amplitude of the frequency and adjust the extraction period of the row and column positions according to the maximum amplitude; The process of determining the risk radius is as follows: Where, is the risk radius, is the preset correction factor, For the The amplitude corresponding to each frequency point; is the total number of frequency points; The process of determining the extraction period is: Where, is the row and column position The extraction cycle at is the preset standard cycle, is the preset correction factor.

6. The ventilation system adjustment strategy optimization method for AMC concentration according to claim 1, characterized in that: The step of adjusting the generation frequency of the state diagram according to the error rate of the neural network model, wherein the generation frequency is used to update the display update period of the state diagram, comprises: When receiving a control instruction sent by a staff member, the control instruction sent by the staff member is regarded as a real instruction; The state diagram is input into the trained neural network model and the control instructions are output as virtual instructions; The similarity between the virtual instruction and the real instruction is calculated, and a display update period of the state diagram is determined according to the similarity; the display update period is proportional to the similarity.

7. A ventilation system adjustment strategy optimization system for AMC concentration, characterized by: The system comprises: A distribution information statistics module is used to obtain a scene graph of the ventilation scene and to count the distribution information of the ventilation equipment according to the scene graph; A state diagram generating module, configured to receive operating data fed back by the ventilation equipment based on a monitor built into the ventilation equipment, analyze the operating data, and generate a state diagram; A model training module is used to display the state diagram, receive control instructions sent by the staff, and train the neural network model according to the state diagram and the control instructions; A control instruction generation module is used to generate control instructions for the ventilation equipment based on the trained neural network model when the error rate of the neural network model is less than a preset threshold; A recursive update module is used to adjust the generation frequency of the state diagram according to the error rate of the neural network model, and the generation frequency is used to update the display update period of the state diagram.

8. The ventilation system adjustment strategy optimization system for AMC concentration according to claim 7, characterized in that: The distribution information statistics module includes: A scene graph acquisition unit, used to acquire a three-dimensional map of the ventilation scene as a scene graph; An information query unit, configured to query the origin position, three-dimensional model, and installation information of a built-in monitor of the ventilation equipment; the installation information of the monitor includes the origin position and three-dimensional model of the monitor; A device model building unit is used to count the origin positions and three-dimensional models of all ventilation devices based on the scene graph and create a device model corresponding to the ventilation devices; The monitor model building unit is used to calculate the origin position of the monitor and its three-dimensional model based on the equipment model, and create a monitor model corresponding to the monitor.

9. The ventilation system adjustment strategy optimization system for AMC concentration according to claim 7, characterized in that: The state diagram generation module includes: An operation data acquisition unit, configured to establish a connection channel with a monitor built into the ventilation equipment and acquire operation data based on the connection channel; an operating data analysis unit, configured to analyze the operating data and determine a risk radius of each monitor model; A device model expansion unit, configured to expand the device model according to the risk radius to obtain a risk area; A state diagram output unit is used to obtain a state diagram after all device models are expanded; The operation data is collected data containing a monitor tag and a time tag.

10. The ventilation system adjustment strategy optimization system for AMC concentration according to claim 7, characterized in that: The recursive update module includes: a real instruction determining unit, configured to, upon receiving a control instruction sent by a staff member, regard the control instruction sent by the staff member as a real instruction; a virtual instruction generation unit, configured to input the state diagram into the trained neural network model and output control instructions as virtual instructions; The cycle update unit is used to calculate the similarity between the virtual instruction and the real instruction, and determine the display update cycle of the state diagram according to the similarity; the display update cycle is proportional to the similarity.

Images