US 12,147,916 B1
Multi-classification and prediction method for electrical signal of aircraft small sample based on naive Bayes
Ke Li, Beijing (CN); Donghao Wang, Beijing (CN); Pengjiao Li, Beijing (CN); Haopeng Wu, Beijing (CN); Yan Kan, Beijing (CN); Liping Pang, Beijing (CN); and Shunkun Yang, Beijing (CN)
Assigned to Beihang University, Beijing (CN)
Filed by Beihang University, Beijing (CN)
Filed on Apr. 9, 2024, as Appl. No. 18/630,728.
Claims priority of application No. 202310687009.3 (CN), filed on Jun. 9, 2023.
Int. Cl. G06N 7/01 (2023.01); G06F 18/2415 (2023.01)
CPC G06N 7/01 (2023.01) [G06F 18/24155 (2023.01)] 3 Claims
OG exemplary drawing
 
1. A method for classifying and identifying an aircraft small sample signal based on naive Bayes, comprising:
(A3) determining a signal source (104);
if the signal source is historical data, entering a historical data reading step (105), performing signal clustering analysis on the data (106), effectively assisting expert annotation work (107), and building the historical data and annotation corresponding to the historical data into an expert database (108); or
if the signal source is real-time data, entering a real-time data reading step (109);
(A4) inputting both the historical data and the real-time data into a naive Bayes classification module (110), and calculating a conditional probability by using a Bayes algorithm;
(A5) sending a calculation result to a probability comparison algorithm classifier (111), and obtaining a classification result of an aircraft signal (112); and
(A6) outputting a real-time fault diagnosis result (113),
wherein
the naive Bayes classification module is a conditional probability model, and
adopts a maximum a posteriori (MAP) probability decision, that is, a classification error probability takes a minimum value, and a corresponding classifier is a classification formula defined as follows:

OG Complex Work Unit Math
wherein
C represents a category variable, c represents a specific category, F represents a characteristic variable of an electrical signal, f represents a specific feature, n represents a summation upper limit, p(C=c) represents a probability of a c category, p(Fi=fi|C=c) represents a conditional probability, and the classification result of the aircraft signal is that when it is assumed that each feature is independent, a maximum probability is selected after the conditional probability is calculated; and
operations of the naive Bayes classification module comprise:
(B1) building an electrical signal of a known category based on a historical aircraft signal and a real-time aircraft signal, and training a sample (202);
(B2) performing feature extraction based on a built naive Bayes model classifier (203);
(B3) obtaining the classification result of the aircraft signal (112) by using the probability comparison algorithm (111), which comprises:
(B31) separately calculating a probability for each category (205);
(B32) calculating a conditional probability for each feature (206); and
(B33) obtaining a classification result of a sample signal (208) through statistic acquisition by using the probability comparison algorithm (207); and
(B4) pre-determining a next status by using a time series prediction method after electrical signal data is classified and determined by using the naive Bayes algorithm (306), which comprises:
(B41) inputting observed data acquired based on a time series to perform model identification (302);
(B42) determining whether the model is applicable by calculating an autocorrelation coefficient and a partial correlation coefficient (303);
(B43) estimating a model parameter (304);
(B44) building a model prediction function (305); and
(B45) outputting a prediction result in a future period of time (306).