Search

CN-121982777-A - Intelligent behavior real-time identification and alarm system for clean room

CN121982777ACN 121982777 ACN121982777 ACN 121982777ACN-121982777-A

Abstract

The invention discloses an intelligent real-time identification and alarm system for a clean room, which relates to the field of computer systems and comprises a sensing module, a calibration module and a calibration module, wherein the sensing module is used for synchronously sensing human body action related signals and environment related parameters in the clean room in real time, the calibration module is used for screening signals and parameters acquired by the sensing module, removing interference signals and calibrating characteristic dimensions to construct a standardized action characteristic data set to be identified, interference information and calibration characteristic dimensions are effectively removed by synchronously capturing human body related action signals and environment parameters in real time, the data standardization level is improved, and the accurate association mapping of abnormal actions and environment damage risks is carried out by combining accurate analysis of time sequence characteristics and dynamic environment characteristics and timely correction of action deviation degree and environment fluctuation coefficients.

Inventors

  • ZHAO WENLONG

Assignees

  • 快嗨云(江苏)信息科技有限公司

Dates

Publication Date
20260505
Application Date
20260130

Claims (10)

  1. 1. The intelligent behavior real-time identification and alarm system for the clean room is characterized by comprising the following components: The sensing module is used for synchronously sensing human body action related signals and environment related parameters in the clean room in real time; The calibration module is used for screening the signals and parameters acquired by the sensing module, removing interference signals and calibrating characteristic dimensions so as to construct a standardized action characteristic data set to be identified; the recognition module is used for receiving the action characteristic data set to be recognized to construct a neural network recognition model, and establishing association mapping between action characteristics and sterile environment damage risks through characteristic learning and pattern matching; the alarm module is used for receiving the abnormal behavior identification result and triggering a corresponding acousto-optic alarm signal according to a preset risk level; The linkage module is used for responding to the alarm trigger signal and controlling the starting and parameter adjustment of the built-in associated purifying equipment in the clean room; And the management module is used for storing the abnormal behavior identification record, the alarm log and the equipment linkage data.
  2. 2. The intelligent real-time identification and alarm system for clean room according to claim 1, wherein the human body motion related signals in the sensing module comprise limb joint angle signals, limb motion rate signals and contact vibration signals of the limbs and facilities in the clean room, and the environment related parameters comprise airflow velocity, suspended particle concentration and pressure difference parameters in the clean room; And the sensing module synchronously aligns each frame of human body action related signals with environment related parameters of corresponding time periods through time stamps, and constructs a time-space related sensing data set of action-environment.
  3. 3. The intelligent real-time identification and alarm system for clean room according to claim 1, wherein the calibration module operates the phase to judge the signals which are output by the sensing module and have the deviation from the motion characteristics of the front and rear time periods exceeding the preset threshold value as interference signals in the time sequence continuity based on the human motion; when the feature dimension is calibrated, different types of action related signals and environment related parameters are mapped to a unified feature description space, and a standardized action feature data set to be identified is constructed through normalization processing of the signal dimension.
  4. 4. The intelligent real-time identification and alarm system for clean rooms according to claim 1, wherein the neural network identification model is a double-branch convolutional neural network model which fuses action time sequence characteristic branches and environment parameter dynamic characteristic branches; Performing association mapping on the action characteristics and the sterile environment damage risk through risk association degree calculation; Wherein, the risk association degree ; Wherein: A and b are characteristic weights, a and b are positive numbers, and the sum of the characteristic weights is 1; Is the degree of deviation of the motion characteristics; Is the fluctuation coefficient of the environmental parameter; correcting factors for action duration; is a real-time motion feature vector; The motion characteristic vector is a preset standard sterile operation motion characteristic vector; Associating parameter values for a real-time environment; is a reference parameter value of the sterile environment of the clean room.
  5. 5. The intelligent real-time identification and alarm system for clean room of claim 4, wherein the structure and processing procedure of the dual-branch convolutional neural network model comprises: The motion time sequence feature branch takes a motion related signal time sequence in a standardized motion feature data set to be identified, which is output by the calibration module, as input, extracts local time sequence related features of motion through at least 3 layers of stacked 1D convolution layers in sequence, and then performs dimension reduction on the extracted local time sequence related features through a maximum pooling layer to obtain motion time sequence feature vectors with uniform dimensions; The environmental parameter dynamic characteristic branch takes an environmental correlation parameter dynamic sequence in a standardized action characteristic data set to be identified, which is output by the calibration module, as input, sequentially extracts multi-scale dynamic change characteristics of environmental parameters through at least 2 layers of 1D convolution layers with different expansion coefficients, and then performs distribution adjustment on the multi-scale dynamic change characteristics through a batch normalization layer to obtain an environmental dynamic characteristic vector with dimensions matched with action time sequence characteristic vectors; the feature fusion processing introduces an attention weight matrix, and performs weighted splicing on the motion time sequence feature vector and the environment dynamic feature vector to obtain a joint feature vector; and inputting the combined feature vector into a full-connection layer for feature mapping, and calculating the deviation degree of the action feature and the fluctuation coefficient of the environmental parameter by using the output feature result.
  6. 6. The intelligent real-time identification and alarm system for cleaning room according to claim 4, wherein said action duration correction factor Is obeyed by the following values: ; Wherein t is the duration of the current action, and k is a preset duration correction coefficient.
  7. 7. The system for real-time identification and alarm of intelligent behaviors in a clean room according to claim 1, wherein the alarm module is characterized in that the preset risk level is divided into preset intervals based on the risk association degree, and the preset intervals at least comprise a low risk interval, a medium risk interval and a high risk interval; Different risk levels correspond to different audible and visual alarm combinations.
  8. 8. The intelligent real-time identification and alarm system for a clean room according to claim 1, wherein the linkage module responds to the alarm trigger signal stage and invokes a preset purification equipment adjustment strategy according to the risk level output by the alarm module: when the risk level is low risk, starting local airflow circulation equipment in the clean room, and adjusting the air speed of the equipment to a preset low gear; When the risk level is a middle risk, starting clean-room global purification equipment, and adjusting equipment purification power to a preset middle gear; When the risk level is high risk, starting an emergency cleaning module of the clean room, and synchronously adjusting the operation parameters of the related cleaning equipment to a preset emergency interval; And in the operation stage of the linkage module, environment-related parameters after the operation of the purifying equipment are synchronously acquired and fed back to the sensing module.
  9. 9. The intelligent real-time identification and alarm system for clean rooms according to claim 1, wherein the management module is operative to bind and store abnormal behavior identification records, alarm logs and equipment linkage data in corresponding time periods with a time axis as an index; the management module is internally provided with association analysis logic of abnormal behavior-risk-linkage effect: and counting the average risk association degree, the alarm response time length and the environment recovery efficiency of the linkage equipment corresponding to different types of abnormal behaviors, and outputting a structural analysis report.
  10. 10. The intelligent real-time identification and alarm system for the clean room according to claim 1, wherein the sensing module is interactively connected with the calibration module and the identification module through a wireless network, the identification module is interactively connected with the alarm module through the wireless network, and the alarm module is interactively connected with the linkage module and the management module through the wireless network.

Description

Intelligent behavior real-time identification and alarm system for clean room Technical Field The invention relates to the technical field of computer systems, in particular to a clean room intelligent behavior real-time identification and alarm system. Background The clean room reduces pollutants such as particles and microorganisms in the air by means of air filtration, temperature and humidity regulation, pressure control and the like, maintains a closed space with a specific cleanliness level, and is widely applied to scenes with strict requirements on environmental cleanliness, such as electronics, medicines, precision manufacturing and the like. The invention patent application with the application number 202411697863.9 discloses a correction method, a correction device and a correction system for detecting humidity in a sterile isolation environment, and aims to solve the problem that a coated humidity sensor only can measure the humidity of water in a cavity of an isolator and neglects the humidity of hydrogen peroxide in a sterile isolation area, so that the difference between the display humidity of the coated humidity sensor and the actual humidity in the sterile isolation area in the isolator is large. However, for the internal environment of the aseptic chamber, when some specific actions occur, the user in the aseptic chamber can destroy the internal aseptic environment conditions of the aseptic chamber, such as sneeze, splash of water, etc., if the alarm is not monitored in time, the internal aseptic environment of the aseptic chamber is difficult to maintain. Therefore, we propose a clean room intelligent behavior real-time identification and alarm system. Disclosure of Invention Aiming at the defects existing in the prior art, the invention provides a clean room intelligent behavior real-time identification and alarm system which can effectively solve the problems in the prior art. In order to achieve the above object, the present invention is achieved by the following technical scheme; the invention discloses a real-time identification and alarm system for intelligent behaviors of a clean room, which comprises the following components: The system comprises a sensing module, a calibration module, an identification module, an alarm module, a linkage module, a management module and a control module, wherein the sensing module is used for synchronously sensing human body action related signals and environment related parameters in a clean room in real time, the calibration module is used for screening signals and parameters acquired by the sensing module, removing interference signals and calibrating characteristic dimensions to construct a standardized action characteristic data set to be identified; The sensing module is interactively connected with the calibration module and the identification module through a wireless network, the identification module is interactively connected with the alarm module through the wireless network, and the alarm module is interactively connected with the linkage module and the management module through the wireless network. Furthermore, in the sensing module, the human body motion related signals comprise a limb joint angle signal, a limb motion rate signal and a contact vibration signal of a limb and facilities in the clean room, and the environment related parameters comprise airflow velocity, suspended particle concentration and pressure difference parameters in the clean room; And the sensing module synchronously aligns each frame of human body action related signals with environment related parameters of corresponding time periods through time stamps, and constructs a time-space related sensing data set of action-environment. Furthermore, the operation stage of the calibration module judges signals, which are output by the sensing module and have the characteristic deviation exceeding a preset threshold value with the action of the front and back time periods, in the time sequence continuity based on the human action as interference signals and eliminates the interference signals; when the feature dimension is calibrated, different types of action related signals and environment related parameters are mapped to a unified feature description space, and a standardized action feature data set to be identified is constructed through normalization processing of the signal dimension. Furthermore, the neural network identification model is a double-branch convolutional neural network model which fuses action time sequence characteristic branches and environment parameter dynamic characteristic branches; Performing association mapping on the action characteristics and the sterile environment damage risk through risk association degree calculation; Wherein, the risk association degree ; Wherein: A and b are characteristic weights, a and b are positive numbers, and the sum of the characteristic weights is 1; Is the degree of deviation of the motion characteristi