融合小波变换与多模态大模型的液压泵仿真信号一致性评估OA

The effectiveness of digital twin systems relies on the dynamic consistency between virtual models and physical entities.As the core power component of hydraulic systems,hydraulic pumps exhibit strongly non-stationary operating signals,such as pressure pulsations and vibration shocks.Traditional consistency assessment methods based on mean squared error or frequency-domain statistics struggle to effectively capture structural dynamic deviations like transient impacts and parameter drifts.To address this,a consistency assessment method integrating continuous wavelet transform and a multimodal large model is proposed.This method involves collecting simulation and measured signals from the hydraulic pump,constructing residual and real noise signals,and converting them into time-frequency images via continuous wavelet transform to highlight dynamic features.Subsequently,the image encoder of a domain-adapted multimodal large model is utilized to extract deep semantic features,and consistency is quantified using feature cosine similarity.Experimental results demonstrate that the proposed method possesses a significant ability to discriminate differences in non-stationary dynamic responses,outperforming traditional evaluation metrics.It can accurately identify model structural deviations under working conditions such as internal leakage and bearing wear,providing reliable technical support for the verification,optimization,and engineering application of hydraulic pump digital twin models.