Search

• Four different types of melt pool signals are integrated and fused to enhance the signal-to-defect correlation. • A data partitioning method is proposed to reduce registration errors and facilitate the defect-related feature extraction. • The feature-level fusion obtains the optimal classification performance compared with the other two fusion strategies. • The proposed data fusion method outperforms the individual signal-based methods for quality classification. In situ monitoring technologies for additive manufacturing are severely affected by signal-to-defect spatial registration, and a single monitoring solution that captures a specific process signature cannot identify defects with sufficient accuracy in industrial applications. In this paper, a data partitioning method was proposed to reduce registration errors and facilitate the extraction of defect-related representative features from data segments. Then, three multi-signal fusion methods based on convolutional neural networks were established from feature-level fusion, decision-level fusion, and data-level fusion, which enhanced the signal-to-defect correlation by means of complementary sensor information. The results showed that the feature-level fusion obtained the optimal classification performance compared with the other two fusion strategies. Among them, the classification accuracies of 100-, 150-, 200-, 300-, 400-, and 500-μm defects reached 74.41%, 89.84%, 90.82%, 93.95%, 97.85%, and 98.83%, respectively. The proposed data fusion approach outperformed the individual signal-based models for quality classification. [ABSTRACT FROM AUTHOR]

• A novel methodology integrating percussion and CNN model is established to detect concrete voids. • The CNN model trained by both CWT and MFCC show satisfying categorizing performances. • The methodology proves its good noise-immune capacity in processing percussion sound signals. • The superiority of the proposed method is well demonstrated by comparison with ML methods. As a fundamental construction material, concrete commonly encounter void defects, which could significantly impair the integrity of concrete members and seriously weaken their load-bearing capacity. Worse still, void defects typically exist inside concrete members, making it difficult to detect them in real time. Therefore, conducting regular inspections for internal void defects of concrete structures is of great importance to maintain structural health. This paper investigates an integrated method to detect concrete voids based on percussion approach and a convolutional neural network (CNN). Firstly, percussion-induced sound signals were collected from 18 concrete specimens with different precast void defects. Each specimen was tapped 100 times, for a total of 1,800 percussion-induced acoustic signals. Subsequently, continuous wavelet transform (CWT) and Mel frequency cepstral coefficient (MFCC) were introduced to convert the collected acoustic signals into two types of images. The converted images were divided into testing and training datasets according to the ratio of 1:3, and 600 images were input to the CNN as the test dataset and 1200 images as the training dataset. The CNN was then trained and verified by the joint input of CWT and MFCC images to perform concrete voids classification. The result shows that the proposed method presents a satisfying classification accuracy of 100 %. Next, percussion tests under 5 levels of noisy environments were conducted, and the results prove the remarkable anti-noise performance of the methodology by achieving 88.12 % accuracy even at the strongest signal-to-noise ratio of 0 dB. Finally, the superiority of the proposed method was further demonstrated through the accuracy comparison with machine learning (ML) methods, with the accuracy consistently outperforming over 25 %. In conclusion, the proposed method has a great potential for future in-situ applications in concrete void detection. [ABSTRACT FROM AUTHOR]

• A high accuracy fault diagnosis scheme for harmonic reducer practical application scenarios is proposed. • Dynamic rules for harmonic reducer in dynamic operating conditions are quantifiably expressed. • Behavioral model is constructed based on the mapping relationship between external excitation and monitoring signals. • The interpretability of Hidden Markov Model in fault diagnosis application is improved. Harmonic reducer is a critical and vulnerable component of industrial robots. Its dynamic rules are difficult to express because harmonic reducers always operate under continuously varying operating conditions in practical application scenarios. How to achieve fault diagnosis under dynamic operating parameters is a challenge. This paper proposes a fault diagnosis scheme for harmonic reducer under practical operating conditions. Hidden Markov Model is used to depict the dynamic rules, novel features state transition probability and observation probability are extracted to construct the mapping relationship between external excitation and monitoring signals. A CNN framework is employed for fault recognition based on fault impacts on the mapping relationship. The results of the verification experiment show that the scheme can achieve high accuracy fault diagnosis in dynamic operating conditions by fully utilizing Hidden Markov Model. The mapping relationship can serve as a behavioral model to support digital-twin modeling and health management for industrial robots. [ABSTRACT FROM AUTHOR]

Fault diagnosis plays a crucial role in ensuring industrial safety and enhancing social benefits. However, identifying fault paths accurately becomes challenging due to process coupling and complex flow. This paper proposes a data space-based approach for fault diagnosis and path recognition. It involves constructing a multi-tiered, multi-causal data space by extracting causal relationships from data. An attention-based convolutional neural network efficiently captures the causal relationships. Pruning and expert knowledge contribute to an elaborate structure. The fault propagation direction is determined using LSTM-least-square linear (OLS) method, aiding rapid problem location. Experimental verification on float glass production and the Tennessee Eastman demonstrates remarkable results, supporting improved reliability and efficiency in industrial processes. • A fault diagnosis and propagation path identification method based on data space is proposed, which can accurately identify the fault propagation path and root cause. • Proposing the construction of a multi-level, multi-causal data space to precisely identify direct causal relationships and eliminate indirect causal links. Multi-level causal diagrams offer enhanced interpretability for diagnosing fundamental reasons in practical production processes. • Introducing a data-driven fault propagation path identification method that efficiently eliminates redundant connections. It determines the propagation direction by assessing the strength of causal relationships between upstream causal variables and downstream affected variables. • The experiment of anomaly location tracing is carried out using the actual production process of float glass, and the effectiveness of the method is verified. [ABSTRACT FROM AUTHOR]

• Multi-source monitoring for quantitative diagnosis of CFRP fatigue cracks. • Quantitative method has combined the global guided wave and local FBG strain. • A heterogeneous feature fusion strategy of "image square" is proposed. • The accuracy of quantitative crack diagnosis models of 1D and 2D CNN are compared. In this paper, a quantitative method for fatigue crack diagnosis of composites is proposed, which combines guided wave and strain multi-source heterogeneous feature fusion with deep learning. Two types of sensors, i.e. piezoelectric sensor and fiber Bragg grating sensor, are used to measure the global and local response of composites under fatigue loading. The one-dimensional signal of the guide wave is encoded into an image by the Gram angular field transform method. The deep features of the image are further extracted by the AlexNet deep learning model. Furthermore, a heterogeneous feature fusion based on image square array of the guided wave and strain is presented. Finally, the CNN models with different input features are constructed. The results show that the quantitative diagnosis of the hole-edge crack based on the heterogeneous fusion of the guided wave and FBG is more accurate and reliable than that based on the single feature of guided wave. [ABSTRACT FROM AUTHOR]