Railway vehicle condition monitoring apparatus

A railway vehicle condition monitoring apparatus, comprising:

• a detection device including a sensor to detect vehicle information represented by a wheel load of a wheel included in a railway vehicle running on a railroad track;
• and
• a determination device including a classifier to which the vehicle information detected by the detection device is input and which is configured to output a vehicle condition such as a presence or absence of an abnormality of the railway vehicle, wherein
• the railway vehicle includes a pair of front and rear bogies which each have two pairs of right and left wheels in front and rear sides;
• vehicle information which is detected by the detection device is a primary spring abnormality index of the front bogie and a primary spring abnormality index of the rear bogie of the railway vehicle which are represented by Expression (1) and Expression (2) below, respectively, and a secondary spring abnormality index of the railway vehicle which is represented by Expression (3) below;
• a vehicle condition which is output from the classifier is a presence or absence of an abnormality and a class of an abnormality of the railway vehicle;
• the classifier is generated by using machine learning which is performed by using training data of a railway vehicle of which the vehicle condition is known, the training data being the vehicle information and the vehicle condition which is known, the machine learning being performed so that when the vehicle information of the training data is input to the classifier, the classifier outputs the vehicle condition of the training data;
• and
• the classifier is generated by using machine learning which is performed by using, as the vehicle information of the training data, the primary spring abnormality index of the front bogie, the primary spring abnormality index of the rear bogie and the secondary spring abnormality index of the railway vehicle which are actually detected by the detection device with respect to a normal railway vehicle, and the primary spring abnormality index of the front bogie, the primary spring abnormality index of the rear bogie and the secondary spring abnormality index of the railway vehicle with resect to an abnormal railway vehicle which are calculated by using numerical simulation based on the primary spring abnormality index of the front bogie, the primary spring abnormality index of the rear bogie and the secondary spring abnormality index of the railway vehicle with respect to the normal railway vehicle, and using, as the vehicle condition of the training data, a known presence or absence of an abnormality and a known class of an abnormality with respect to the normal railway vehicle and the abnormal railway vehicle, the machine learning being performed so that when vehicle information of the normal railway vehicle and the abnormal railway vehicle is input to the classifier, the classifier outputs a vehicle condition of the normal railway vehicle and the abnormal railway vehicle,
• primary spring abnormality index of front bogie=(P1+P4)−(P2+P3) (1)
• primary spring abnormality index of rear bogie=(P5+P8)−(P6+P7) (2)
• secondary spring abnormality index of railway vehicle=(P1+P3+P6+P8)−(P2+P4+P5+P7) (3)
• where P1 represents a wheel load of a wheel located in a front right side of the front bogie, P2 represents a wheel load of a wheel located in a front left side of the front bogie, P3 represents a wheel load of a wheel located in a rear right side of the front bogie, P4 represents a wheel load of a wheel located in a rear left side of the front bogie P5 represents a wheel load of a wheel located in a front right side of the rear bogie, P6 represents a wheel load of a wheel located in a front left side of the rear bogie, P7 represents a wheel load of a wheel located in a rear right side of the rear bogie, and P8 represents a wheel load of a wheel located in a rear left side of the rear bogie.