| CPC G16H 40/60 (2018.01) [H04W 64/003 (2013.01); H04W 76/14 (2018.02)] | 7 Claims |
|
1. A communication system for a respiratory protection device, comprising:
pairing transmission modules, each comprising a first pairing unit and a second pairing unit which are paired and connected within a specified location area and are respectively installed on different components of the respiratory protection device; and
a main control module connected with the plurality of pairing transmission modules and comprising a communication unit, a control unit and a storage unit; wherein
the communication unit is in communication connection with the pairing transmission modules for information transmission;
the control unit collects and processes identity information of the first pairing unit and/or the second pairing unit and pairing information of the first pairing unit and the second pairing unit, the identity information at least comprising component information of the respiratory protection device;
the storage unit stores collection and processing results of the identity information and the pairing information;
the main control module also comprises a data processing unit which processes and analyzes data stored in the storage unit and predicts a usage status of the respiratory protection device according to analysis results;
the data processing unit comprises:
a data processor for processing the data obtained from the storage unit, the processing at least comprising data cleaning, conversion, normalization and feature engineering;
data analysis algorithms which deeply analyze and mine the processed data to find trends and correlations in the data and obtain analysis results; and
a prediction model which is established according to the analysis results and used for predicting a future usage status of the respiratory protection device;
the prediction model is an LSTM model, comprising:
an input layer for receiving the analysis results;
an LSTM layer for learning the long-term dependence of time series data in the analysis results and generating an internal representation;
an output layer for receiving the internal representation from the LSMT layer and generating final prediction results; and
a sliding window which divides the time series data from the analysis results into different windows and allows for movement on a time axis in a sliding manner to generate a series of subsequence data and input the same into the LSTM layer; and
the relationship between a window size and a sliding step is defined as follows:
S=W×(α+β)/2
where S represents the sliding step, W represents the window size, α and β denote a periodic factor and a trend factor respectively, which are adjusted based on the trends identified in the analysis results of the data analysis algorithms; the unit of S depends on the unit of W, with W measured in time units, and α and β are pure numerical values without units, ranging from 0 to 1, representing the degree of influence of the corresponding factors on the sliding step; and the periodic factor α takes into account periodic variations in the data, and the trend factor β takes into account trend variations in the data.
|