Search

Variational Auto-Encoders (VAEs) have shown great potential in the unsupervised learning of data distributions. An VAE trained on normal images is expected to only be able to reconstruct normal images, allowing the localization of anomalous pixels in an image via manipulating information within the VAE ELBO loss. The ELBO consists of KL divergence loss (image-wise) and reconstruction loss (pixel-wise). It is natural and straightforward to use the later as the predictor. However, usually local anomaly added to a normal image can deteriorate the whole reconstructed image, causing segmentation using only naive pixel errors not accurate. Energy based projection was proposed to increase the reconstruction accuracy of normal regions/pixels, which achieved the state-of-the-art localization accuracy on simple natural images. Another possible predictors are ELBO and its components gradients with respect to each pixels. Previous work claimed that KL gradient is a robust predictor. In this paper, we argue that the energy based projection in medical imaging is not as useful as on natural images. Moreover, we observe that the robustness of KL gradient predictor totally depends on the setting of the VAE and dataset. We also explored the effect of the weight of KL loss within beta-VAE and predictor ensemble in anomaly localization., Comment: arXiv admin note: substantial text overlap with arXiv:2002.03734 by other authors

There are two main purposes for our research on wake vortex: one is to improve operation safety, the other is to reduce the wake spacing and ensure the runway capacity and operation efficiency under the premise of ensuring safety. By comparing and analyzing the technology of wake spacing reduction, we have studied aerial classification technology, synchronous offset approach technology, high glide slope approach technology, wake vortex warning system, aircraft wake vortex interval system and so on. Research shows that by proposing advanced concepts or improving the ATM program, namely the wake interval reduction technology, we can avoid the influence of the wake generated by the front aircraft, ensure the operation safety and increase the airport capacity by 3.7%-25%. The key technology of the wake interval reduction technology is the wake vortex prediction model and the safety analysis of the wake encounter. However, the higher the accuracy of the model is, the more complex the computation is. The research on wake detection technology will be the future development direction.

In the paper, the production of wood vinegar from Chinese fir sawdust (FS), cotton stalk (CS) and bamboo sawdust (BS) by carbonization process was addressed. The wood vinegar yield was investigated and the organic compounds contained were determined by gas chromatography and mass spectrometry. It was found that the refined wood vinegar yield of FS increased firstly and then decreased with increasing carbonization temperature and the highest yield reached about 25% in 350–450 °C. The relative contents of acids and ketones from FS decreased and that of phenols increased with increasing temperature. The relative contents of acids and phenols in the wood vinegars produced from the samples were in the order of BS > CS > FS and that of ketones reversed. KCl solution treatment caused a decrease in the relative contents of the phenols and ketones, but an increase in that of the acids in FS wood vinegar.

The Chinese lignite upgrading process with high temperature nitrogen as a drying agent using a horizontal fixed-bed reactor was addressed and a kinetic model for the single particle lignite drying process was developed based on the experimental results. The results show that the lignite drying process consists of two stages, one is drying rate increasing stage and the other is drying rate decreasing stage. The lignite drying process can be promoted by initial nitrogen temperature increasing and lignite particle size decreasing. The effective diffusivity coefficient ( D e ) of the moisture from the lignite particle during the drying process with different initial temperature was obtained theoretically by calculation. The activated energy of the moisture emission from the lignite with different particle size (10–25 mm) was also obtained.

Modeling uncertainties and measurement noise in physical servo systems severely deteriorate their tracking performance. In this paper, an output feedback control strategy which integrates a discontinuous robust control with extended state observer via feed-forward cancellation technique is proposed for high precision motion control of a dc motor. The designed extended state observer can estimate the unmeasured system state and the modeling uncertainties simultaneously. Since the vast majority of the modeling uncertainties can be estimated by the extended state observer and then compensated in the controller design, the task of robust control can be greatly alleviated which prevents the discontinuous output feedback controller from chattering. In addition, the proposed controller theoretically achieves an asymptotic output tracking performance even in the presence of modeling uncertainties, which is rather significant for high precision control of motion systems. Extensive comparative simulation results are provided to verify the high performance nature of the proposed control strategy.