Intelligent fault diagnosis method for domain separation adaptive one-dimensional convolutional neural network
A convolutional neural network and fault diagnosis technology, applied in the field of fault diagnosis, can solve the problems of reduced diagnostic accuracy and low robustness
- Summary
- Abstract
- Description
- Claims
- Application Information
AI Technical Summary
Problems solved by technology
Method used
Image
Examples
Embodiment 1
[0062] refer to figure 1 , provides a schematic diagram of the overall structure of a domain separation adaptive one-dimensional convolutional neural network intelligent fault diagnosis method, as shown in figure 1 , a domain separation adaptive one-dimensional convolutional neural network intelligent fault diagnosis method includes obtaining mechanical vibration signals, and constructing a sample set and a label set; establishing a model loss function and constructing a fault diagnosis model; training and confirming the model; wherein, the vibration Signals are divided into source domain signals and target domain signals.
[0063] This method compensates the mismatch by adjusting the model parameters or input features, making the field of mechanical fault diagnosis self-adaptive, and simultaneously extracts its domain discriminative features and domain invariant features for fault diagnosis, and solves the problem that the training data and test data in fault diagnosis come f...
Embodiment 2
[0077] refer to figure 2 , This embodiment is different from the first embodiment in that: the steps of constructing a fault diagnosis model and establishing a model loss function in the above embodiment include: establishing a joint loss function for fault diagnosis; constructing a fault diagnosis model; and inputting a sample set.
[0078] Specifically, the steps of constructing a fault diagnosis model and establishing a model loss function include:
[0079] S21: Establish a joint loss function for fault diagnosis;
[0080] Among them, the formula of the joint loss function L is as follows:
[0081]
[0082] Among them, L rec Represents the reconstruction loss function, L diff Denotes the minimization difference loss function, L adv Denotes the adversarial loss function, L task Represents the classification loss function, α, β and γ are the weights of the control loss items, Indicates the parameters, specifically, θ c is the shared CNN encoder E c parameter; θ ...
Embodiment 3
[0098] refer to Figure 4 with Figure 5 , what this embodiment is different from above embodiment is: the step of training and confirming model comprises:
[0099] Initialize the model; optimize the confirmation model; model prediction. Specifically, the steps of training and confirming the model include:
[0100] S31: Initialize the model, wherein, the initialization model uses the source domain data D s According to the classification loss function L of formula (2) task to initialize θ c and θ y ;
[0101] S32: Optimizing and confirming the model;
[0102] It should be noted that the input of the DS-1DCNN fault diagnosis model is a labeled source domain sample {X s , Y s} and unlabeled target domain samples {X t}, the goal of model optimization is to obtain the loss function L that minimizes formula (1), which uses BP-based stochastic gradient descent method (SGD) to update parameters As shown in formula (6-11):
[0103]
[0104]
[0105]
[0106]
...
PUM
Login to View More Abstract
Description
Claims
Application Information
Login to View More