Tensor manifold-based lightweight neural network fault diagnosis method and system

The invention discloses a lightweight neural network fault diagnosis method and system based on tensor manifold, and the method comprises the steps: taking a fault diagnosis image as a vector, reconstructing the fault diagnosis image into a first multi-order tensor under a constraint condition, and representing the first multi-order tensor in a tensor manifold in a tensor string decomposition form; under the condition that a tensor rank is not changed, performing multiple iterations on the first multi-order tensor by using a gradient descent method of a tensor manifold to obtain a first shrinkage multi-order tensor after data dimension reduction; reconstructing a weight matrix in the fault diagnosis neural network into a second multi-order tensor under a constraint condition, and representing the second multi-order tensor in a tensor manifold in a tensor string decomposition form to obtain a new neural network taking a factor tensor as a weight; training the new neural network by taking the first shrinkage mu