Search

Superheat is an important indicator to reflect the current production efficiency of aluminium electrolytic cells. Due to the difficulty of superheat online real-time measurement， this paper proposes a superheat identification method based on residual convolution self-attention neural network (RCSANN). As the production data in aluminium electrolysis process is time series data and featured with multi-source heterogeneous characteristics， the isomorphic representation method is designed for heterogeneous data. On this basis， the RCSANN superheat model is proposed to extract the global and local features of the isomorphic time series data. Aiming at the problem of few labels and uneven category distribution of superheat data， the unsupervised pre-training method based on auto-encoder and the weighted cross-entropy loss function are used to improve the performance of the superheat identification task. The validity of the proposed method is verified by simulation and comparison experiments on the benchmark dataset. Then， experiments are carried out on the dataset of superheat in aluminium electrolysis with only a few unbalanced labels. The results show that not only the accuracy of superheat identification is improved compared with other existing models， but also the generalization ability can be guaranteed under few training labeled-samples. [ABSTRACT FROM AUTHOR]

In general adversarial attacks oriented to sample sets, the general perturbation design that causes most sample to output errors is the key to the research. This paper takes the typical convolutional neural networks as the research object, summarizes the existing general perturbation generation algorithms, and proposes a general perturbation generation algorithm that combines batch random gradient ascent and spherical projection search. In each iteration of the algorithm, a small batch of samples are extracted from the sample set, and the general perturbation is calculated by using the random gradient rising strategy which reduces the value of the loss function. The general perturbation is then projected to the high-dimensional spherical surface with a radius of ε, so as to reduce the search space of general disturbances. The algorithm also introduces a regularization technique to improve the generation quality of general disturbances. Experimental results show that compared with the baseline algorithm, the attack success rate is significantly increased, and the solution efficiency of general perturbation is improved by about 30 times. [ABSTRACT FROM AUTHOR]