Abstract: With the development of structural inspection technology, signal feature extraction and structural health diagnosis based on intensive multi-sensor monitoring has become a research hotspot. Wave is a common external excitation for marine structures, and the dynamic response of pile foundation under this excitation has the characteristics of multi-type signal mixing, and signal decomposition is a necessary analysis process. However, the commonly used decomposition methods mainly focus on the independent analysis of individual sensors, ignoring the channel correlation information and the consistency of the same-order intrinsic mode function (IMF) spectrum, which leads to the poor comparability of the signal characteristics of different nodes and affects the damage detection effect. In order to solve the above problems, this study introduces the noise assisted multidimensional empirical mode decomposition (NA-MEMD) into the multi-sensor feature extraction of pile foundation, and combines the mutual information between the IMFs of each channel, and proposes an optimal noise assisted multidimensional empirical mode decomposition (ONA-MEMD) to achieve optimal and consistent decomposition of the signals under the intensive monitoring of multi-sensors, and further uses the energy factor for the damage detection of pile foundation. Through simulated signals and high-pile wharf pile foundation model experiments, comparisons indicate that the ONA-MEMD method effectively reduces mode aliasing and over-decomposition, maintains the consistency of IMF features, and better detects the local damage of the pile foundation. The method provides new ideas and methods for feature extraction and damage detection of multi-sensor monitoring signals for complex marine structures.