To optimize machine allocation and task dispatching in smart manufacturing factories, this paper proposes a manufacturing resource scheduling framework based on reinforcement learning (RL). The framework formulates the entire scheduling process as a multi-stage sequential decision problem, and further obtains the scheduling order by the combination of deep convolutional neural network (CNN) and improved deep Q-network (DQN). Specifically, with respect to the representation of the Markov decision process (MDP), the feature matrix is considered as the state space and a set of heuristic dispatching rules are denoted as the action space. In addition, the deep CNN is employed to approximate the state-action values, and the double dueling deep Q-network with prioritized experience replay and noisy network (D3QPN2) is adopted to determine the appropriate action according to the current state. In the experiments, compared with the traditional heuristic method, the proposed method is able to learn high-quality scheduling policy and achieve shorter makespan on the standard public datasets.

Time series data has attached extensive attention as multi-domain data, but it is difficult to analyze due to its high dimension and few labels. Self-supervised representation learning provides an effective way for processing such data. Considering the frequency domain features of the time series data itself and the contextual feature in the classification task, this paper proposes an unsupervised Long Short-Term Memory (LSTM) and contrastive transformer-based time series representation model using contrastive learning. Firstly, transforming data with frequency domain-based augmentation increases the ability to represent features in the frequency domain. Secondly, the encoder module with three layers of LSTM and convolution maps the augmented data to the latent space and calculates the temporal loss with a contrastive transformer module and contextual loss. Finally, after self-supervised training, the representation vector of the original data can be got from the pre-trained encoder. Our model achieves satisfied performances on Human Activity Recognition (HAR) and sleepEDF real-life datasets.