sensors

Early-stage engineering design often neglects sensor integration, treating signal pathways and data flows as secondary considerations rather than core design dimensions. Method: This paper proposes a “sensor-native” Digital Systems Design (DSD) paradigm that intrinsically embeds sensing capabilities at the conceptual design stage. DSD holistically integrates multimodal sensor selection and placement optimization, embedded signal conditioning, edge–cloud collaborative data architecture, and real-time digital twin simulation. Unlike conventional “retrofitted sensing,” DSD treats the signal chain and data flow as first-class design variables, enabling end-to-end data闭环 across design, operation, and maintenance. Contribution/Results: Experiments demonstrate that DSD improves early-design predictability by over 40%, supports dynamic iterative optimization and predictive maintenance, and has been validated in mechanical systems and intelligent equipment education as well as prototype development.

Complex engineering systems exhibit multiple fault modes, yet distinguishing degradation causes remains challenging under label-scarce conditions; additionally, sensor responses exhibit modality heterogeneity, leading to biased remaining useful life (RUL) predictions. To address these issues, we propose an unsupervised joint framework that simultaneously performs fault-mode clustering and modality-adaptive sensor selection, integrating sparse feature learning, multi-sensor temporal fusion, and modality-conditional RUL modeling. Without requiring fault labels, the method automatically partitions fault categories and identifies the most discriminative subset of sensors, thereby closing the “fault identification–RUL prediction” loop. Evaluated on dual-modal synthetic data and the NASA C-MAPSS turbofan engine dataset, our approach achieves an 18.7% improvement in fault-mode identification accuracy and a 23.5% reduction in RUL prediction error, significantly outperforming existing unsupervised methods.

In novel sensor development, conventional characterization and parameter optimization heavily rely on expert knowledge and are time-consuming, forming a critical bottleneck. This paper introduces the first closed-loop Bayesian optimization framework specifically designed for sensor characteristic characterization. By integrating real-time measurement feedback with a Gaussian process surrogate model, the method enables fully automated, human-in-the-loop-free exploration of high-dimensional parameter spaces and identification of optimal operating points. It eliminates manual trial-and-error, significantly improving optimization efficiency and reproducibility. Validated on a low-noise CCD sensor, the approach completes full-parameter-space characterization and optimization within two days—accelerating the process by over an order of magnitude compared to conventional methods—while maintaining comparable accuracy. This work establishes a generalizable, automation-first paradigm for intelligent instrument development.

Event-driven sensors (e.g., event-based sensors, EBS) exhibit heterogeneous and rapidly evolving hardware output formats, causing severe data fragmentation that impedes cross-device interoperability and system integration. To address this, we propose Event-Format—the first standardized framework for event-stream data—designed to be modality-agnostic and forward-compatible. Its layered, flexible structure uniformly represents mainstream event modalities (e.g., polarity, timestamp, pixel coordinates), while built-in extension mechanisms accommodate emerging sensor architectures. The standard explicitly specifies serialization protocols, metadata schemas, and validation rules. We implement an industrial-grade lightweight parser and validate Event-Format across robotics perception and ultra-low-power vision applications, demonstrating robust cross-platform compatibility and extensibility. Event-Format establishes a foundational infrastructure for seamless interchange and sharing of heterogeneous multi-source event data in industrial, commercial, and defense domains. (149 words)

Sensor placement optimization for structural health monitoring of semiconductor probe cards remains challenging. Method: This paper proposes a physics-informed Transformer-based deep learning framework. Leveraging frequency response function data generated via finite element simulation, the model jointly encodes dynamic response features and physical constraints by integrating convolutional layers with self-attention mechanisms. Physics-guided data augmentation and perception-aware statistical enhancement are introduced to improve generalizability, while attention-weight visualization identifies critical sensor locations. Contribution/Results: The method achieves 99.83% accuracy in health-state classification and 99.73% recall in crack detection. Robustness is validated through triple 10-fold stratified cross-validation. The approach delivers an interpretable, cost-effective sensor deployment paradigm for active maintenance systems, advancing predictive maintenance in semiconductor manufacturing.

This work addresses the high instruction overhead and software complexity inherent in traditional soft processors, which rely on explicit instructions to read sensors and generate PWM signals within control loops. The authors propose an innovative architecture that maps high-frequency peripheral inputs directly to general-purpose registers and fully offloads PWM generation to dedicated hardware, thereby enabling zero-instruction sensor access and continuous actuation without software intervention. Implemented on a 32-bit, five-stage pipelined MIPS-style RISC soft core, the design incorporates direct peripheral-to-register-file write ports, a single-cycle multiplier, and hardware-based PWM logic. Experimental results demonstrate a reduction in the control loop cycle count from 91 to 43, eliminating five critical instructions; under a 20 ms control frame, the system achieves a real-time margin of 7,300–15,000×, substantially simplifying software and enhancing computational efficiency.

This work addresses the challenges of high early-stage uncertainty in manufacturing monitoring system development—leading to redundant modeling and substantial training costs—and the limited transferability of filtering pipelines in cross-domain image segmentation tasks. To tackle these issues, the authors propose a problem-centric design paradigm that constructs an abstract system model to continuously accumulate and retrieve historical segmentation tasks along with their associated filtering pipelines, enabling solution reuse and incremental optimization. The approach integrates similarity-based problem retrieval, abstract modeling, pipeline reuse, and a retrieval-augmented evolutionary learning mechanism. Experimental results demonstrate that the method significantly reduces training costs and late-stage revision risks, provides the first systematic validation of filtering pipeline transferability across similar segmentation tasks, and achieves a favorable balance among complexity, technical requirements, and reliability under lightweight model constraints.

This work addresses the challenges of low diagnostic accuracy and inadequate uncertainty quantification in monitoring complex nonlinear industrial systems by proposing an uncertainty-aware framework that synergistically integrates data-driven and physics-based models. The approach employs a lightweight physics-informed residual design enhanced with temporal features at the feature level, while leveraging model-level ensembling and conformal prediction to effectively fuse sensor measurements, time-lagged features, and physical residuals. Evaluated on the continuous stirred-tank reactor (CSTR) benchmark, the proposed method achieves a 2.9% improvement in diagnostic accuracy over the best-performing baseline and produces smaller, well-calibrated prediction sets, thereby significantly enhancing both decision reliability and the fidelity of uncertainty quantification.