Minced pork jowl meat, also called the sticking-piece, is commonly used to be adulterated in minced pork, which influences the overall product quality and safety. In this study, hyperspectral imaging (HSI) methodology was proposed to identify and visualize this kind of meat adulteration. A total of 176 hyperspectral images were acquired from adulterated meat samples in the range of 0%–100% (w/w) at 10% increments using a visible and near-infrared (400–1000 nm) HSI system in reflectance mode. Mean spectra were extracted from the regions of interests (ROIs) and represented each sample accordingly. The performance comparison of established partial least square regression (PLSR) models showed that spectra pretreated by standard normal variate (SNV) performed best with R_p² = 0.9549 and residual predictive deviation (RPD) = 4.54. Furthermore, functional wavelengths related to adulteration identification were individually selected using methods of principal component (PC) loadings, two-dimensional correlation spectroscopy (2D-COS), and regression coefficients (RC). After that, the multispectral RC-PLSR model exhibited the most satisfactory results in prediction set that R_p² was 0.9063, RPD was 2.30, and the limit of detection (LOD) was 6.50%. Spatial distribution was visualized based on the preferred model, and adulteration levels were clearly discernible. Lastly, the visualization was further verified that prediction results well matched the known distribution in samples. Overall, HSI was tested to be a promising methodology for detecting and visualizing minced jowl meat in pork.

The counting task, which plays a fundamental role in numerous applications (e.g., crowd counting, traffic statistics), aims to predict the number of objects with various densities. Existing object counting tasks are designed for a single object class. However, it is inevitable to encounter newly coming data with new classes in our real world. We name this scenario as evolving object counting . In this paper, we build the first evolving object counting dataset and propose a unified object counting network as the first attempt to address this task. The proposed network consists of two key components: a class-agnostic mask module and a class-incremental module. The class-agnostic mask module learns generic object occupation prior by predicting a class-agnostic binary mask (e.g., 1 denotes there exists an object at the considering position in an image and 0 otherwise). The class-incremental module is used to handle new classes and provides discriminative class guidance for density map prediction. The combined outputs of the class-agnostic mask module and image feature extractor are used to predict the final density map. When new classes arrive, we first add new neural nodes to the last regression and classification layers of the class-incremental module. Then, instead of retraining the model from scratch, we utilize knowledge distillation to help the model retain and consolidate what it has previously learned. We also employ a support sample bank to store a small number of typical training samples for each class, which are used to prevent the model from forgetting key information from old data. With this design, our model can efficiently and effectively adapt to new classes while maintaining good performance on already-seen data without large-scale retraining. Extensive experiments on the collected dataset demonstrate favorable performance. The dataset and code will be available at: https://github.com/Tanyjiang/EOCO.

Three-dimensional numerical simulations are carried out to investigate the effects of film-hole arrangement and blowing ratio on the squealer tip leakage flow field and tip film-cooling performance. Six film-hole arrangements with 13 holes are designed in the current study for comparison. In type-A and type-B, the film-cooling holes are arranged in a single row, located at the middle camber line or close to the suction-side squealer. The four modified film-hole arrangements are realized by placing two rows of total 7 film-cooling holes at the leading edge (type-C, type-D, type-E and type-F) and remaining the rest film-cooling holes in a row at the middle chord zone. The results show that the leakage flow entering the tip gap from the leading edge of suction side, the leading edge of pressure side and the middle chord and trailing edge of pressure side behaves different flow feature inside the tip cavity, inducing complicated swirling flow filed. The modified film-hole arrangements yield more reasonable film coverage on the tip surface by comparing with the single row film-hole arrangement under relatively high blowing ratios. In addition, the modified film-hole arrangements also show different rules on the film-cooling effectiveness distributions over some specific surfaces, such as tip cavity bottom surface and squealer top surface, as well as PS squealer inner surface and SS squealer inner surface. Among the presented four modified film-hole arrangements, type-D and type-F gain the most favorable film-cooling improvement.

This paper concentrates on the study of adaptive control for the synchronization of fractional-order neural networks. Instead of classical adaptive control updating method, an intermittent-active updating strategy is proposed to adaptively tune the control gain in a fractional-order fashion. Moreover, quantization is brought into the control design to take into account the restricted bandwidth in signal transmission. Note that the suggested controller is basic yet effective in terms of the fractional-order system. The main theorem is established with the method of reduction to absurdity as well as Lyapunov stability theorem. Finally, simulation calculation is conducted to validate the effectiveness of our proposed method.

Efficient models for remote sensing object counting are urgently required for applications in scenarios with limited computing resources, such as drones or embedded systems. A straightforward yet powerful technique to achieve this is knowledge distillation, which steers the learning of student networks by leveraging the experience of already-trained teacher networks. However, it faces a pair of challenges: Firstly, due to its two-stage training nature, a longer training period is essential, especially as the training samples increase. Secondly, despite the proficiency of teacher networks in transmitting assimilated knowledge, they tend to overlook the latent insights gained during their learning process. To address these challenges, we introduce an online distillation learning method for remote sensing object counting. It builds an end-to-end training framework that seamlessly integrates two distinct networks into a unified one. It comprises a shared shallow module, a teacher branch, and a student branch. The shared module serving as the foundation for both branches is dedicated to learning some primitive information. The teacher branch utilizes prior knowledge to reduce the difficulty of learning and guides the student branch in online learning. In parallel, the student branch achieves parameter reduction and rapid inference capabilities by means of channel reduction. This design empowers the student branch not only to receive privileged insights from the teacher branch but also to tap into the latent reservoir of knowledge held by the teacher branch during the learning process. Moreover, we propose a relation-in-relation distillation method that allows the student branch to effectively comprehend the evolution of the relationship of intra-layer teacher features among different inter-layer features. Extensive experiments on two challenging datasets demonstrate the effectiveness of our method, which achieves comparable performance to state-of-the-art methods despite using far fewer parameters.

Purpose: This paper aims to investigate the effects of inhomogeneities on the rolling contact fatigue (RCF) life in elastohydrodynamically lubricated (EHL) point contacts.Design/methodology/approach: A numerical model for predicting the RCF life of inhomogeneous materials in EHL contacts was established by combining the EHL model and the inclusion model through the eigen-displacement and then connecting to the RCF life model through the subsurface stresses. Effects of the type, size, location and orientation of a single inhomogeneity and the distribution of multiple inhomogeneities on the RCF life were investigated.Findings: The RCF life of a half-space containing manganese sulfide (MnS) inhomogeneity or the mixed inhomogeneity of aluminium oxide (Al₂O₃) and calcium oxide (CaO) was longer than that for the case of Al₂O₃inhomogeneity. For a single ellipsoidal MnS inhomogeneity, increases of its semi-axis length and decreases of its horizontal distance between the inhomogeneity and the contact center shortened the RCF life. Furthermore, the relationship between the depth of a single MnS inhomogeneity and the RCF life was found. For the half-space containing multiple inhomogeneitites, the RCF life decreased remarkably compared with the homogeneous half-space and showed discreteness.Originality/value: This paper implements the prediction of the RCF life of inhomogeneous materials under EHL condition.