Experimental studies have been carried out to compare the performance of two dimensional straight and curved diffusers of same area ratio and effective divergence angle in the Reynolds number range of7.8 X 105 to 1.29 X 105. Free stream turbulence effects have.also been studied at the increased turbulence level to 3.4per cent. The results indicate that straight diffuser pressure recovery is slightly higher as compared to the curved diffusers. However, stream turbulence, which improves the pressure recovery in both cases, has been observed to have greater effect in case of curved diffuser. Boundary layer velocity profiles on the diffuser surfaces have also been presented at various streamwise stations. It is observed that the growth of inner surface boundary layer has a major effects on losses in case of a curved diffuser. Nomenclature

Scalar and vector data arising from the numerical simulation of flows and transport equations encode a wealth of information that needs to be extracted and displayed so as to gain insight into the underlying patterns. In this respect, it is an increasingly important task to process, represent, visualize, and analyze flow data. This paradigm maps an array of discrete data values either to geometric elements followed by graphics rendering or directly to pixels. Coupled with this pipeline of numerical flow visualization is an emerging preprocessing stage of great promise that unleashes the power of artificial intelligence, particularly machine learning, to perform feature detection and pattern recognition. In light of the importance of computer graphics and machine learning to data analysis, the editors conceived the idea of a special issue, with the core purpose of extracting and exploiting information by means of numerical flow visualization to help with scientific discovery.Among the submissions received, five have been selected to appear in this special issue. Liu provides a multifaceted survey of representative advances in numerical flow visualization over the past three decades, with emphasis on geometry-based and texture-based methods. Also introduced in this context is the author's own research for representing, depicting, and exploring flow data. The discussions revolve around streamline integration, interactive streamline placement, automatic streamline placement, streamline clustering, pathlines, flow texture synthesis, feature mining, and parallel visualization. The paper by Wu et al. presents a fast algorithm that can generate dense high-resolution images of a 2D steady flow field by combining a geometry-based method (evenly-spaced streamlines) and texture mapping. The result bears resemblance to that produced by a well-known texture-based approach called line integral convolution, while enhancing image contrast and accelerating image synthesis.Han et al. apply machine learning to interactive exploration of unsteady vector data with pathlines. A small set of particle traces is generated from the original/full spatiotemporal resolution via parallel computing in situ as a computational fluid dynamics (CFD) simulation runs. In the form of a flow map, these integral curves offer an alternative representation of the time-varying vector field and are hence taken as a feature-oriented training dataset. This dataset is digested by a deep neural network for accurately inferring the spatiotemporal trajectory of any new particle that the user may designate. Maharjan and Zaspel propose data-driven filters for augmenting the preprocessing modules of ParaView (a visualization package) with the pattern recognition capabilities of PyTorch (a deep-learning framework). The effectiveness is demonstrated by several examples in feature segmentation and classification. Employing a CFD framework, Nair et al. discuss the connection between shock patterns and turbulent mixing in a burner of a supersonic combustion ramjet. Their...