Porosity is a common defect in the Laser Powder Bed Fusion (LPBF) process, significantly limiting its potential in mass customization applications. Existing studies have shown that porosity can be effectively detected by analyzing the thermal history of the melt pool. However, the relationship between thermal features extracted from thermograms and porosity is highly nonlinear, making defect prediction challenging. This paper proposes a Convolutional Neural Network (CNN)-integrated Transformer (CiT) model for defect prediction and anomaly localization in LPBF. The CiT model enhances predictive accuracy by combining global and local feature extraction while leveraging global average pooling instead of the classification (CLS) token after removing the Transformer decoder. Additionally, a lightweight multi-head self-attention mechanism is designed to optimize the model structure, effectively reducing the number of parameters while maintaining accuracy. Furthermore, an anomaly localization method based on Score-CAM is introduced to identify potential causes of porosity formation, enabling defect traceability in LPBF. The proposed CiT model is evaluated against four state-of-the-art algorithms, demonstrating its superior performance in defect prediction and anomaly localization.

The significance of rainfall detection is generally acknowledged, and the linked opportunistic approach to giving it new momentum has also been extensively demonstrated. This study conducted a rainfall monitoring experiment using two stationary smartphones in a light indoor setting, receiving downlink signals from a long-term evolution (LTE) base station for two months. The analysis reveals that the received signals are not stable during dry periods, but the reference signal receiving power (RSRP) and reference signal strength indicator (RSSI) parameters can produce a certain degree of degradation during rainfall. To address the concern about frequent switching of connection to the base station resulting in different fluctuation levels, the standardized standard deviations of the four signaling parameters for different time windows were extracted as features to build a dry-rainy classification model. Furthermore, a rain rate class identification model is also established based on the extraction of the specific rain-induced attenuation and standardized standard deviation. The experiment mentioned has shown promising results in rainfall monitoring based on widely available opportunistic signal sources from wireless terminals.

Mass individualized prototyping (MIP) is a kind of advanced and high-value-added manufacturing service. In the MIP context, the service providers usually receive massive individualized prototyping orders, and they should keep a stable state in the presence of continuous significant stresses or disruptions to maximize profit. This article proposed a graph convolutional neural network-based deep reinforcement learning (GCNN-DRL) method to achieve the resilient production control of MIP (RPC-MIP). The proposed method combines the excellent feature extraction ability of graph convolutional neural networks with the autonomous decision-making ability of deep reinforcement learning. First, a three-dimensional disjunctive graph is defined to model the RPC-MIP, and two dimensionality-reduction rules are proposed to reduce the dimensionality of the disjunctive graph. By extracting the features of the reduced-dimensional disjunctive graph through a graph isomorphic network, the convergence of the model is improved. Second, a two-stage control decision strategy is proposed in the DRL process to avoid poor solution quality in the large-scale searching space of the RPC-MIP. As a result, the high generalization capability and efficiency of the proposed GCNN-DRL method are obtained, which is verified by experiments. It could withstand system performance in the presence of continuous significant stresses of workpiece replenishment and also make fast rearrangement of dispatching decisions to achieve rapid recovery after disruptions happen in different production scenarios and system scales, thereby improving the system’s resilience.