Intelligent acoustic emission-based pipeline leak detection technology plays a critical role in Internet of Things structural health monitoring for offshore platforms. However, traditional deep networks possess large parameter counts and high computational complexity, making them infeasible for deployment on resource-constrained edge nodes, while lightweight methods universally adopt single-scale feature extraction and cannot simultaneously capture short-duration burst and long-range attenuation characteristics of acoustic emission signals, resulting in insufficient discriminative capability for adjacent valves. To address this, this article proposes Tiny Leak Sense Net (TinyLSN), a novel, lightweight leak localization framework specifically designed for Internet of Things edge nodes. TinyLSN achieves an optimal balance between computational efficiency and detection performance through three innovative components we designed, including the inverted residual block (IRB), multiscale dilated perception module (MSDPM), and large kernel feed-forward network (LK-FFN), which, respectively, enhance cross-channel interactions, capture multiscale temporal features, and extract global attenuation patterns. On our self-constructed experimental dataset simulating real offshore platform operational pipeline leakage, TinyLSN achieved detection accuracy of 97.11%–97.45% and an extremely low false positive rate of 0.27%–0.32%, significantly outperforming lightweight baseline methods. Validation on publicly available benchmark datasets further confirmed its generalization capability. When deployed on the STM32H7B3I-DK microcontroller, TinyLSN requires only 267.16-KiB flash memory and achieves 3.417-ms inference latency. Furthermore, TinyLSN maintains over 90% accuracy under strong noise and achieves 94.83% accuracy with only 10% training samples, fully validating its reliability in harsh industrial environments and providing an efficient and feasible solution for offshore platform Internet of Things edge intelligence.

Weed classification is a fundamental perception task for agricultural robots and an essential enabler of precision and sustainable farming. Existing solutions often rely on high-power edge computing platforms, which limit long-term autonomous operation in Internet of Things (IoT) environments. Meanwhile, the computational complexity of high-accuracy deep learning (DL) models hinders their deployment on resource-constrained microcontrollers (MCUs), a critical component of IoT edge nodes. To address these challenges, this article proposes a tiny machine learning (TinyML)-enabled energy-efficient IoT framework for on-device weed classification, integrating a novel 3-D alignment knowledge distillation (TDA-KD) strategy with a lightweight multilayer dilated-convolution student network. The framework enhances knowledge transfer by jointly aligning: 1) individual predictions; 2) inter-sample correlations; and 3) class semantics, further strengthened through a multitemperature calibration mechanism. Experimental results on the DeepWeeds and 4-Weeds datasets demonstrate that the proposed student model achieves over 95% classification accuracy with only 240 K parameters and 87.51 MFLOPs. The model is successfully deployed on an OpenMV H7 Plus board with an STM32H7 MCU, requiring just 105.68 KB Flash memory and achieving an inference time of 378.3 ms with 510.7 mJ energy consumption per sample. A runtime analysis on the Vitirover horticultural robot shows that, compared with a Jetson Nano-based implementation, the proposed IoT pipeline extends operational time by approximately 30.5%. These results highlight the feasibility of deploying high-accuracy weed classification directly on ultralow-power IoT devices, thereby significantly enhancing the autonomy, energy efficiency, and scalability of agricultural robots.

In the context of industrial bearing prognostics, conventional deep learning methods have demonstrated remarkable prediction capabilities but typically provide only deterministic Remaining Useful Life (RUL) estimates without quantifying prediction uncertainty. This significant limitation hinders reliable decision-making in critical maintenance scenarios where understanding prediction confidence is essential. While statistical approaches attempt to model degradation uncertainty, they often struggle with the complex degradation patterns exhibited by modern industrial bearings. This paper presents UABP-BTNSCL, a novel framework integrating supervised contrastive learning with Bayesian deep learning to address these challenges.A key innovation is the unique integration of a Deep Residual Shrinkage Network (DRSN) enhanced Temporal Convolutional Network backbone with a dual uncertainty quantification mechanism that simultaneously captures aleatoric uncertainty through Negative Log-Likelihood optimization and epistemic uncertainty via Monte Carlo Dropout. A key innovation lies in our SupCon-RUL strategy, which discretizes the degradation progression into meaningful stages to structure the feature space, significantly enhancing uncertainty estimation reliability. Extensive validation on PHM2012 and XJTU-SY benchmarks demonstrates the framework’s effectiveness, achieving superior prediction accuracy with average RMSE values of 0.086 and 0.077 respectively, while providing well-calibrated uncertainty bounds (PICP: 0.97, MPIW: 0.40). Ablation studies confirm that each component—DRSN enhancement, SupCon-RUL feature structuring, and Bayesian uncertainty estimation—contributes synergistically to the framework’s overall performance in both point prediction accuracy and uncertainty quantification quality.

With the continuous growth of global population and increasing pressure on food security, the transformation toward precise and intelligent agricultural production has become an inevitable trend. In this context, accurate identification of field weeds is crucial for improving crop yields and reducing agricultural inputs. However, agricultural Internet of Things (IoT) edge devices are generally subject to strict constraints in terms of power consumption, storage, and real-time performance. Existing lightweight convolutional neural networks often struggle to simultaneously achieve high accuracy and low resource consumption for fine-grained weed identification tasks. To address this challenge, this paper proposes a hardware aware lightweight convolutional neural network named Weed-Aware Ghost Enhanced Network (WAGENet) for microcontroller deployment. The network synergistically integrates Ghost low-cost feature generation, Mobile Inverted Bottleneck Convolution (MBConv) for deep semantic extraction, Squeeze and Excitation (SE) and Coordinate Attention (CA) dual attention mechanisms for channel space joint calibration, and Atrous Spatial Pyramid Pooling (ASPP) for multi-scale context fusion. It constructs a progressive feature abstraction system from shallow textures to high-level semantics. On the public DeepWeeds dataset, WAGENet achieves 95.71% classification accuracy and 93.80% F1 score with only 0.163 M parameters and 2.43 × 10⁸ multiply accumulate operations (MACC), attaining a parameter efficiency of 587.19%/M and significantly outperforming existing mainstream lightweight models. The model has been successfully deployed on the STM32H7B3I microcontroller development board, achieving a single inference latency of 94.63 ms, an internal Flash footprint of only 686.95 KiB, and a single inference energy consumption of 41.45 mJ. Experimental results demonstrate that WAGENet achieves a trade off among accuracy, latency, and energy consumption under strict resource constraints, providing a reproducible microcontroller deployment paradigm for battery powered field robots, drones, and other agricultural IoT edge devices.