Industry 4.0 technologies have been gaining significant momentum in recent years. Despite widespread adoption, many companies struggle with the successful implementation of these technologies. This study aims to identify the critical success factors for implementing Industry 4.0 technologies and to examine the effects of various factor combinations. Using the technology–organization–environment framework and the practice-based view, this paper proposes a comprehensive research model. This study employs a hybrid approach combining structural equation modeling (SEM) and fuzzy-set qualitative comparative analysis (fsQCA) to analyze survey data collected from 202 Chinese manufacturing firms. The SEM results indicate that top management support, technology competence, supplier support, and pilot projects are significantly associated with successful implementation. However, the fsQCA results reveal that individual factors alone are insufficient for success; rather, it is the combinations of these factors that drive successful implementation. Specifically, three key combinations lead to success: (1) top management support and technology competence; (2) top management support and supplier support; and (3) technology competence, supplier support, and pilot projects. By recognizing these combinations, manufacturing firms can develop more effective implementation strategies for Industry 4.0 technologies.

Rotation-invariant recognition of shapes is a common challenge in computer vision. Recent approaches have significantly improved the accuracy of rotation-invariant recognition by encoding the rotational invariance of shapes as hand-crafted image features and introducing deep neural networks. However, the methods based on pixels have too much redundant information, and the critical geometric information is prone to early leakage, resulting in weak rotation-invariant recognition of fine-grained shapes. In this paper, we reconsider the shape recognition problem from the perspective of contour points rather than pixels. We propose an anti-noise rotation-invariant convolution module based on contour geometric aware for fine-grained shape recognition. The module divides the shape contour into multiple local geometric regions (LGA), where we implement finer-grained rotation-invariant coding in terms of point topological relations. We provide a deep network composed of five such cascaded modules for classification and retrieval experiments. The results show that our method exhibits excellent performance in rotation-invariant recognition of fine-grained shapes. In addition, we demonstrate that our method is robust to contour noise and the rotation centers.

The landslide incidence around petroleum pipelines usually leads to severe pipeline damage, causing serious environmental pollution, economic losses, and even casualties. The commonly used methods for monitoring petroleum pipelines include remote sensing and ground monitoring, which can be further applied in predicting landslide hazards. However, current landslide prediction strategies are generally limited because the provided predictive indicators are restricted, and the prediction needs to be more timely. These limitations have caused significant difficulties in the practical application of landslide prediction. Based on the actual operating conditions and landslide incidence records of a section of the pipeline, we propose a machine learning-based landslide prewarning system for buried pipelines to achieve real-time landslide monitoring and rapid warning. Landslide conditions and pipeline operation records are collected using a ground monitoring system around the target pipeline section. They are integrated to generate machine learning datasets extended using the finite element method (FEM). After comparing multiple machine learning algorithms, the XGBoost model is ultimately adopted for the prediction system. The system is calibrated and verified by comparing the prediction results with landslide data. In the verification test, the landslide warning message is acquired about 10 h before the landslide occurrence, and the error in predicting the position of the landslide is about 20 m in the case of the 466 m target pipeline.