From engineering diagrams to engineering models: Visual recognition and applications

We present a computational recognition approach to convert network-like, image-based engineering diagrams into engineering models with which computations of interests, such as CAD modeling, simulation, information retrieval and semantic-aware editing, are enabled. The proposed approach is designed to work on diagrams produced using computer-aided drawing tools or hand sketches, and does not rely on temporal information for recognition. Our approach leverages a Convolutional Neural Network (CNN) as a trainable engineering symbol recognizer. The CNN is capable of learning the visual features of the defined symbol categories from a few user-supplied prototypical diagrams and a set of synthetically generated training samples. When deployed, the trained CNN is applied either to the entire input diagram using a multi-scale sliding window or, where applicable, to each isolated pixel cluster obtained through Connected Component Analysis (CCA). Then the connectivity between the detected symbols are analyzed to obtain an attributed graph representing the engineering model conveyed by the diagram. We evaluate the performance of the approach with benchmark datasets and demonstrate its utility in different application scenarios, including the construction and simulation of control system or mechanical vibratory system models from hand-sketched or camera-captured images, content-based image retrieval for resonant circuits and sematic-aware image editing for floor plans.     

Sliding mode control of a bioreactor

In this paper, sliding mode control (SMC) of a bioreactor is considered and is compared with PID control. The magnitude of the error in SMC is found to be lower than that in PID control. Moreover, the magnitudes of cells and nutrients were very close to the selected reference values in SMC, whereas they were quite different in PID control. Overall, SMC was more robust against disturbances and had better performance than PID control.     

Effects of chromium (VI) on the activities of ureolytic mixed culture

BACKGROUND: Ureolytic microbial carbonate precipitation (MCP) is a novel process for removing excess calcium from industrial effluents. This process is based on the hydrolization of urea to provide a suitable medium for the precipitation of Ca as CaCO₃. RESULT: A toxicity identification evaluation was conducted on synthetic wastewater simulating wastewater from paper recycling to determine the inhibition or toxicity of chromium(VI) ions on the activities of a ureolytic mixed culture (UMC) with respect to the removal rate of COD and removal of Ca²⁺ in batch reactors. The 50% inhibiting concentration (IC₅₀) and the 25% inhibiting concentration (IC₂₅) values of Cr(VI) on UMC were determined as 40 and 18 mg L^?1, respectively, for an exposure time of 1 day. The inhibitory effects of Cr(VI) were lower on UMC after longer exposure times of 3, 4 and 5 days. COD removal rate, ammonium production and Ca removal of samples with 512 mg Cr(VI) L^?1 were significantly inhibited. CONCLUSION The present study has shown that the effects of different metallic species, organic toxicants and other environmental factors should be taken into account when removing problematic Ca and other possible elements from the environment using UMC. Copyright © 2008 Society of Chemical Industry     

A Tutorial on Multiplierless Design of FIR Filters: Algorithms and Architectures

Finite impulse response (FIR) filtering is a ubiquitous operation in digital signal processing systems and is generally implemented in full custom circuits due to high-speed and low-power design requirements. The complexity of an FIR filter is dominated by the multiplication of a large number of filter coefficients by the filter input or its time-shifted versions. Over the years, many high-level synthesis algorithms and filter architectures have been introduced in order to design FIR filters efficiently. This article reviews how constant multiplications can be designed using shifts and adders/subtractors that are maximally shared through a high-level synthesis algorithm based on some optimization criteria. It also presents different forms of FIR filters, namely, direct, transposed, and hybrid and shows how constant multiplications in each filter form can be realized under a shift-adds architecture. More importantly, it explores the impact of the multiplierless realization of each filter form on area, delay, and power dissipation of both custom (ASIC) and reconfigurable (FPGA) circuits by carrying out experiments with different bitwidths of filter input, design libraries, reconfigurable target devices, and optimization criteria in high-level synthesis algorithms.     

Characterization of products evolved from supercritical water gasification of xylose (principal sugar in hemicellulose)

The catalyst decomposition of xylose (the principal sugar in hemicellulose) was examined in supercritical water for temperature from 400 to 600°C. Experiments were performed in the absence and presence of three main types of catalysts [alkali catalysts (K₂CO₃ and KOH) and metal impregnated activated carbons (Ni/AC) and (Ru/AC)] with a reaction time of 1h. Gasification yield reaches maximum values by using K₂CO₃ and KOH at the highest temperature of 600°C. The highest H₂ yield and the highest CH₄ yield were obtained by using Ru/AC and Ni/AC, respectively. Acetic acid and 5-methyl furfural were determined as the main aqueous products and reached maximum value by using Ru/AC.     

Mode I fracture toughness determination with straight notched disk bending method

A new method called the straight notched disk bending method is developed for mode I fracture toughness determination using rock cores. Disk specimens of andesite and marble having a single straight edge notch were subjected to three-point bending loads. Dimensionless stress intensity factor estimations and fracture toughness tests were conducted for different notch lengths, span lengths, thicknesses and diameters of the cylindrical rock specimens. Stress intensity factors were computed by three-dimensional finite element modeling and the results were presented for a wide range of specimen geometrical parameters. Results of experiments were compared to the results of well-known mode I fracture toughness testing methods. For specimens having thickness equal to the radius, mode I fracture toughness was lower and close to the results obtained by semi-circular bending method. When thickness was increased and doubled, mode I fracture toughness increased and approached to the value found by the suggested cracked chevron notched Brazilian disk method. Advantages of the new method included easy specimen preparation and testing procedure, stiffer specimen geometry, smaller fracture process zone, and flexibility of the specimen geometry for the investigation of the size effect behavior.     

What determines static friction and controls the transition to sliding?

We studied the shear response of a confined lubricant layer on approach of the transition to sliding with a surface force apparatus modified for oscillatory shear. In a given experiment, we found that the transition to sliding occurred always around the same deformation amplitude although the shear stress needed to initiate sliding varied up to a factor of two depending on sample history. This suggests the concept of deformation-controlled switching from rest to sliding. The elastic spring-constant, in the stick state, weakened with increasing deformation amplitude. This decrease can be described by a power law when plotted versus the distance to a critical deformation amplitude. The build-up of solid-like behavior after sliding stopped was also gradual and was consistent with a logarithmic time dependence. We suggest a model relating the gradual decrease of stiffness to weakening of the boundary layer, specifically to destruction of some elastic links between molecules or between molecules and the solid surfaces. Static friction (the force that must be overcome at the onset of kinetic motion) is proportional to the number of such links formed during the time of stick.     

Distributed call rerouting in multiclass broadband networks

Broadband integrated services digital networks (B-ISDN) must handle multiclass traffic with diverse quality of service requirements. We consider a multiclass routing model in which routes are calculated in a distributed fashion by the call originating nodes. Within this general context, we address the problem of rerouting a set of previously routed calls to avoid a failed link. Under the approach we propose, a single node executes an aggregate, global rerouting of all affected calls and then converts the set of aggregate routes into an allocation of bandwidth on each link to call origination nodes for the purpose of rerouting. The bandwidth allocation is distributed to each origination node, which in turn then calculates routes for the individual calls. The problem faced by each call origination node is a variant of the socalled bandwidth packing problem. We develop and analyze an approximate algorithm for solving this problem in the specific context that arises in our setting.The research was supported by IBM Corporation and by NSF Grant No. CDR-8803012.On personal leave from IBM RTP.