Turnout switch machine fault diagnosis method and device based on ensemble learning

• Main Authors: JIN ZHENZHEN, LIU QIYANG, CHEN YANJUN, HE DEQIANG, LAO ZHENPENG, HE YILING, LI YULIN, FU YANG, LI XIANWANG
• Format: Patent
• Language: Chinese; English
• Published: 17.10.2023
• Subjects: CALCULATING; COMPUTING; COUNTING; ELECTRIC DIGITAL DATA PROCESSING; ENSURING THE SAFETY OF RAILWAY TRAFFIC; GUIDING RAILWAY TRAFFIC; MEASURING; MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES; PERFORMING OPERATIONS; PHYSICS; RAILWAYS; TESTING; TRANSPORTING

The invention discloses a turnout switch machine fault diagnosis method and device based on ensemble learning. The method comprises the following steps: collecting action current signals of a turnout switch machine; calculating a time domain feature and a multi-scale permutation entropy feature of each phase of current signal of the switch machine, and constructing a feature set of each state; calculating correlation factors between the feature sets to obtain the weight of each category feature, and adaptively selecting sensitive features according to a dynamic threshold to divide a training set and a test set; constructing a turnout switch machine fault diagnosis IFL-LightGBM model, and enabling the model to be more concentrated on difficult-to-classify samples by improving a loss function; and training and optimizing parameters of the IFL-LightGBM model by using the training set, and performing fault diagnosis on data of the turnout switch machine to be diagnosed to obtain a fault classification result of t