The invention discloses a steel rail corrugation identification method based on a multi-task learning neural network. The method comprises the following steps: data acquisition; data category division and preprocessing: performing one-to-one correspondence on the vehicle-mounted data and the actual steel rail corrugation data through a linear interpolation method, and further dividing the vehicle-mounted data and the actual steel rail corrugation data into a training data set and a test data set; calculating a vibration-noise frequency spectrum of a 1/3 octave and taking the vibration-noise frequency spectrum as the input of the model; training a multi-task learning steel rail corrugation detection neural network model; and inputting test set data into the trained model, and outputting a corrugation identification result. According to the method, multi-task detection can be realized, namely, the wavelength and wave depth information of the steel rail corrugation can be identified at the same time, so that the category and severity of the steel rail corrugation are judged, railway operation and maintenance personnel are helped to formulate more targeted maintenance and maintenance measures, the method has higher detection efficiency and portability, and the detection manpower and material resource cost is reduced.

    本发明公开了一种基于多任务学习神经网络的钢轨波磨识别方法,包括如下步骤:数据采集;数据类别划分和预处理:通过线性插值的方法将车载数据和实际的钢轨波磨数据一一对应,并进一步划分为训练数据集和测试数据集;计算1/3倍频程的振动‑噪声频谱并作为模型的输入;训练多任务学习钢轨波磨检测神经网络模型;将测试集数据输入到训练好的模型中,输出波磨识别结果。本发明方法能够实现多任务检测,即同时识别钢轨波磨的波长和波深信息,从而判断钢轨波磨的类别和严重程度,帮助铁路运维人员制定更有针对性的养护维修措施,本发明方法具有更高的检测效率和便携性,降低了检测人力物力成本。


    Zugriff

    Download


    Exportieren, teilen und zitieren



    Titel :

    Rail corrugation identification method based on multi-task learning neural network


    Weitere Titelangaben:

    一种基于多任务学习神经网络的钢轨波磨识别方法


    Beteiligte:
    HE QING (Autor:in) / LI CHENZHONG (Autor:in) / WAN ZHUANG (Autor:in) / LI PEIHAI (Autor:in) / YANG FEI (Autor:in) / TAN SHEHUI (Autor:in) / WANG PING (Autor:in) / WANG YIFENG (Autor:in) / SUN HUAKUN (Autor:in)

    Erscheinungsdatum :

    2023-03-07


    Medientyp :

    Patent


    Format :

    Elektronische Ressource


    Sprache :

    Chinesisch


    Klassifikation :

    IPC:    G06F ELECTRIC DIGITAL DATA PROCESSING , Elektrische digitale Datenverarbeitung / B61K Andere Hilfseinrichtungen für Eisenbahnen , OTHER AUXILIARY EQUIPMENT FOR RAILWAYS / G01H MEASUREMENT OF MECHANICAL VIBRATIONS OR ULTRASONIC, SONIC OR INFRASONIC WAVES , Messen von mechanischen Schwingungen oder Ultraschall-, Schall- oder Infraschallwellen / G01P MEASURING LINEAR OR ANGULAR SPEED, ACCELERATION, DECELERATION OR SHOCK , Messen der Linear- oder Winkelgeschwindigkeit, der Beschleunigung, der Verzögerung oder des Stoßes / G06N COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS , Rechnersysteme, basierend auf spezifischen Rechenmodellen / G10L Analyse oder Synthese von Sprache , SPEECH ANALYSIS OR SYNTHESIS



    Rail corrugation

    Morgan, H.H. | Engineering Index Backfile | 1941


    Rail corrugation

    Busse, A. | Engineering Index Backfile | 1914


    Rail corrugation

    Busse, A. | Engineering Index Backfile | 1910


    Rail corrugation

    Aman, F.T. | Engineering Index Backfile | 1906


    Rail corrugation

    Holt, Robert B. | Engineering Index Backfile | 1923