Reducing time and memory requirements in topology optimization of transient problems

Feature-rank-code-based classifiers have been proposed recently in order to reduce the complexity for electronic nose system (ENS). The performance of these classifiers is degraded when the discriminatory information of gases lie in the actual feature values not in the ranks of features. To overcome the problem, a gas identification system based on a simple distance measure in combination with different type of features is proposed. In order to improve the computational cost of the existing ENS, a novel combination of optimum subset of features is proposed. To achieve the aim of low computational cost, a large feature vector is recursively reduced to a small number of features without compromising the classification accuracy of the system. Discrete binary particle swarm optimization, a metaheuristic for global search is used in a recursive setup to select the optimum subset of features for classification. Euclidian distance is used as a similarity measure for identification of different industrial gases. The proposed system is tested for the identification of 13 different industrial gases, namely, C3H8, Cl2, CO, CO2, SO2, NO2, NH3, C2H4O, C3H6O, C2H4, C2H6O, C7H8, and CH4. These gases are contained in three different data sets; two among these data sets are acquired experimentally in a laboratory setup, while one of these data sets is taken from the University of California at Irvine machine learning repository. Results reveal that the computational cost and the memory requirement of an ENS can be significantly reduced by combining different type of features. An average classification accuracy of 98.17% is achieved by the proposed system with an average 69.04% reduction in memory requirement using only three to five features.

This paper extends the nonsmooth Relaxed Variational Approach (RVA) to topology optimization, proposed by the authors in a preceding work, to the solution of thermal optimization problems. First, the RVA topology optimization method is briefly discussed and, then, it is applied to a set of representative problems in which the thermal compliance, the deviation of the heat flux from a given field and the average temperature are minimized. For each optimization problem, the relaxed topological derivative and the corresponding adjoint equations are presented. This set of expressions are then discretized in the context of the finite element method and used in the optimization algorithm to update the characteristic function. Finally, some representative (3D) thermal topology optimization examples are presented to asses the performance of the proposed method and the Relaxed Variational Approach solutions are compared with the ones obtained with the level set method in terms of the cost function, the topology design and the computational cost.

While intensity-modulated proton therapy (IMPT) has great potential to improve the therapeutic efficacy of radiotherapy, IMPT optimization can be computationally demanding, particularly for large and complex tumors. Here we propose a dose calculation strategy to accelerate IMPT optimization while reducing memory requirements. By using two adjustable threshold parameters, our method separates dose contributions from proton beamlets into major and minor components for each dose voxel. The optimization proceeds with two levels of iterations: in inner iterations, doses are updated in correspondence with changes in beamlet intensities from only the major contributions while keeping the portions from the minor contributions constant; in outer iterations, doses are recalculated exactly by considering both major and minor contributions. Since the number of elements in the influence matrix for major contributions is relatively small, each inner iteration proceeds quickly. Each outer iteration requires a longer computation time, but only a few such iterations are needed. Our study shows that the proposed strategy leads to nearly identical dose distributions as those optimized with the full influence matrix, but reducing computing time by at least a factor of 3 and internal memory requirements by a factor of 10 or more. In addition, we show that the proposed approach could enhance other optimization-related applications such as optimizing beam angles. By using an advanced lung cancer case that would demand large computing resources by conventional optimization approach, we show how our method may potentially help improve IMPT treatment planning in real clinical situations.

Introduction A very efficient formulation for the inverse design of two-dimensional turbomachinery blades in turbulent transonic flow conditions is presented. The proposed optimization technique employs a highly accurate space discretization scheme for the accurate computation of the flow solution and of the objective function. The sensitivity derivatives are evaluated very efficiently by using a discrete adjoint formulation. The use of an auxiliary dissipative flow solver in the adjoint equations, rather than the accurate basic flow solver, allows a robust evaluation of the sensitivity derivatives, even in presence of shocks and of numerical noise. Moreover, if the variable prescribed on the blade surface is the pressure, as usual, a simplified inviscid formulation of the adjoint problem can be used, even in the real case of turbulent flows. The entire procedure is termed progressive optimization, because both the flow computation and the iterative solution of the adjoint problem, based on the addition of an artificial time-dependent term in the adjoint equations, are brought to convergence while approaching the optimum. The idea of the simultaneous convergence of all iterative solutions suggests also the use of progressively finer grids for the progressive computation of the flow field; the cost of the adjoint problem is strongly reduced by solving the adjoint equations always at the coarsest level. The proposed approach has been tested on the inverse design of a two-dimensional turbine blade in turbulent transonic flow conditions, both with fixed stagger angle and with fixed outlet angle of the camberline. The numerical results demonstrate that the methodology is robust and highly efficient, with a converged design optimization produced in no more than the amount of computational work needed to perform from 0.7 to 1.3 converged flow analyses on the finest grid. 'Associate Professor. ^Professor and Department Head, AIAA senior member. Copyright © 2000 by L. A. Catalano and A. Dadone. Published by the American Institute of Aeronautics and Astronautics, Inc. with permission. Until few years ago, the optimization of the aerodynamic efficiency of turbomachinery bladings has been mostly approached by subsequent improvements based on the experience of the designers, each one of them followed by expensive and timeconsuming wind tunnel tests. For such a reason, the recent development of automatic tools based on the numerical solution of the flow equations and on an optimization process for the shape changes has greatly attracted the interest of turbomachinery designers. Due to the complex flow patterns about turbomachinery bladings, these numerical methods have been first developed and applied to the inverse and the direct design of airfoils; some of them have been later extended to the optimization of turbine and compressor blades. Nevertheless, many of the inverse design methods for turbomachinery bladings currently used are based on the use of a transpiration model', which cannot be extended to the direct design of the blade. For such a reason, the authors find very attractive to extend more general metodologies, already developed and tested for the optimization of airfoils or wings, to the design of turbomachinery bladings. In the application of these methods to transonic flows, typical of high-load bladings, a major difficulty commonly arises: the objective function is generally noisy or non-smooth, so that the convergence of standard optimization techniques may be compromised or even prevented, as shown in Ref. 3 for the case of a transonic nozzle. In such situations a standard finite-difference approach for design sensitivities or an alternative route based on the solution of a set of sensitivity equations' may result in an optimization process hung up at one of the many local minima artificially created by the shock shifting. The adjoint procedure proposed by Jameson et al overcomes this problem by effectively smoothing the objective function, which however may change the result of the optimization process, as outlined in Ref. 7. An alternative smoothing procedure which operates only on the design sensitivities and not on the objective function itself is proposed in Ref. 8, where it is (c)2000 American Institute of Aeronautics & Astronautics or Published with Permission of Author(s) and/or Author(s)' Sponsoring Organization. shown to be effective and efficient in designing transonic diffusers and airfoils by using two-dimensional inviscid flow equations. A second, very important challenge is the development of an optimization technique able to converge to the optimum in a computational work corresponding to few flow analyses. Indeed, most existing methodologies are based on a sequence of global iterations, each one of them requiring a computation of the converged flow field, a solution for the sensitivity derivatives and an appropriate update of the design variables. The repeated solution of the flow equations typical of such a serial approach results in a high computational cost of the entire design procedure. The efficiency of the smoothing procedure developed in Ref. 8 has been recently improved by introducing a progressive optimization strategy''', whereby the optimization process is based on partially converged flow solutions with the aim to converge the flow solution while converging the design problems. In particular, in Ref. 11 this efficient approach is applied to the inverse design of turbine blades in inviscid transonic flow conditions. In Refs. 9 — 1 2 , the adjoint equations are solved using a direct solver, appropriate for two-dimensional flow problems, but not effective for three-dimensional flow problems. The extension to three dimensions has been considered in Ref. 13, where the progressive procedure suggested in Ref. 9 is modified by adding a time derivative in the adjoint equations, and by solving them iteratively. in a progressive manner quite similar to the one employed for the flow equations. The smoothing procedure and the use of partially converged flow solutions as well as of partially converged adjoint equation solutions in the progressive optimization implies that the optimization process is based on approximate values of the gradient of the objective function. Another example of approximate design sensitivity has been discussed by Matsuzawa and Hafez, where adjoint equations based on an inviscid flow formulation have been used successfully to determine approximate gradients of the objective function for airfoil inverse design in laminar flow conditions. The aim of the present paper is to apply all of the aforementioned strategies that contribute to improve the robustness and the efficiency of the optimization process to the inverse design of two-dimensional cascades in turbulent flow conditions. The final goal is to validate the robustness and efficiency of the outlined procedure also for this important class of design problems and, more importantly, to perform a first important step towards the development of an automatic tool for the direct optimization of turbomachinery bladings in real flow conditions. The adjoint equations will be derived using an inviscid solver involving a traditional centered scheme with added numerical viscosity. The inviscid formulation will be used in order to preserve the simplicity of the adjoint equations. Moreover, an additional time derivative will be introduced in order to solve the time dependent adjoint equations progressively in time. Finally, the flow equations will be solved on a progressively finer mesh, while the adjoint equations will be always solved on the coarsest mesh in order to reduce the cost of the computation of the gradient of the objective function to a few percents of the total cost of the optimization procedure. The following sections will be devoted to the description of the main ingredients of the proposed approach, that will be tested on the inverse design of a two-dimensional turbine blade in turbulent transonic flow conditions, both with fixed stagger angle and with fixed outlet angle of the camberline.

In this research manuscript, a new protocol is proposed for predicting the available space in the cloud and verifying the security of stored data. The protocol is utilized for learning the available data, and based on this learning, the available storage space is identified, after which the cloud service providers allow for data storage. The Integrity verification separates the private and the public data, which avoids privacy issues. The integration of the private data is done with the help of cloud service providers with respect to the third-party auditing (TPA). Earlier, public key cryptography and bilinear map technologies have been combined by the researchers, but the computation time and costs were high. To secure the integrity of the data storage, the client should execute several computations. Therefore, this research suggests a reliable and effective method called position-aware Merkle tree (PMT), which is implemented for ensuring data integrity. The proposed system uses a PMT that enables the TPA to perform multiple auditing tasks with high efficiency, less computational cost and computation time. Simulation results clearly shows that the developed PMT method consumed 0.00459 milliseconds of computation time, which is limited when compared to the existing models.

Mathematical optimization theories are employed for the design of structures in structural optimization. Structural optimization is being widely utilized for practical problems due to well-developed commercial software systems. Three representative structural optimization systems such as Genesis, MSC Nastran and OptiStruct are investigated and evaluated by solving various test examples in different scales. The design capabilities of three software systems are explored and the performances of the systems are compared. The performance of structural optimization depends on the quality of the optimum solution and the computational time, and these aspects are compared from an application viewpoint. For a fair comparison, the same formulations are utilized, and the same optimization methods are employed for each example. Also, the same system environment is prepared, and the same optimization parameters are used. Additionally, various design options of each software system are tested for the best performance. Linear static response size, shape, topology, topometry and topography optimizations are applied to the examples and the results are compared. No system seems to be the best in all cases and each system has advantages and disadvantages depending on the application. In general, Genesis is excellent in computational time while OptiStruct gives excellent optimum solutions, in size, topometry and topology optimizations. Meanwhile, MSC Nastran presents good solutions in shape and topography optimizations.

Results of the application of an adaptive finite element (FE) based solution using the FETK library of M. Holst to Density Functional Theory (DFT) approximation to the electronic structure of atoms and molecules are reported. The severe problem associated with the rapid variation of the electronic wave functions in the near singular regions of the atomic centers is treated by implementing completely unstructured simplex meshes that resolve these features around atomic nuclei. This concentrates the computational work in the regions in which the shortest length scales are necessary and provides for low resolution in regions for which there is no electron density. The accuracy of the solutions significantly improved when adaptive mesh refinement was applied, and it was found that the essential difficulties of the Kohn-Sham eigenvalues equation were the result of the singular behavior of the atomic potentials. Even though the matrix representations of the discrete Hamiltonian operator in the adaptive finite element basis are always sparse with a linear complexity in the number of discretization points, the overall memory and computational requirements for the solver implemented were found to be quite high. The number of mesh vertices per atom as a function of the atomic number Z andmore » the required accuracy e (in atomic units) was esitmated to be v (e;Z) = 122:37 * Z2:2346 /1:1173 , and the number of floating point operations per minimization step for a system of NA atoms was found to be 0(N3A*v(e,Z0) (e.g. Z=26, e=0.0015 au, and NA=100, the memory requirement and computational cost would be ~0.2 terabytes and ~25 petaflops). It was found that the high cost of the method could be reduced somewhat by using a geometric based refinement strategy to fix the error near the singularities.« less

In this paper, we propose a polymorphic configuration architecture, that can be tailored to efficiently support reconfiguration needs of the applications at runtime. Today, CGRAs host multiple applications, running simultaneously on a single platform. Novel CGRAs allow each application to exploit late binding and time sharing for enhancing the power and area efficiency. These features require frequent reconfigurations, making reconfiguration time a bottleneck for time critical applications. Existing solutions to this problem either employ powerful configuration architectures or hide configuration latency (using configuration caching). However, both these methods incur significant costs when designed for worst-case reconfiguration needs. As an alternative to worst-case dedicated configuration mechanism, we exploit reconfiguration to provide each application its private configuration environment (PCE). PCE relies on a morphable configuration infrastructure, a distributed memory sub-system, and a set of PCE controllers. The PCE controllers customize the morphable configuration infrastructure and reserve portion of the a distributed memory sub-system, to act as a context memory for each application, separately. Thereby, each application enjoys its own configuration environment which is optimal in terms of configuration speed, memory requirements and energy. Simulation results using representative applications (WLAN and Matrix Multiplication) showed that PCE offers up to 58% reduction in memory requirements, compared to dedicated, worst case configuration architecture. Synthesis results show that the morphable reconfiguration architecture incurs negligible overheads ( 3% area and 4% power compared of a single processing element).

The error recovery problem in wireless sensor networks is studied from a generic resource‐constrained energy‐optimization perspective. To characterize the features of error recovery schemes that suit the majority of applications, an energy model is developed and inferences are drawn based on a suitable performance metric. For applications that require error control coding, an efficient scheme is proposed based on an interesting observation related to shortened Reed–Solomon (RS) codes for packet reliability. It is shown that multiple instances (γ) of RS codes defined on a smaller alphabet combined with interleaving results in smaller resource usage, while the performance exceeds the benefits of a shortened RS code defined over a larger alphabet. In particular, the proposed scheme can have an error correction capability of up to γ times larger than that of the conventional RS scheme without changing the rate of the code with much lower power, timing and memory requirements. Implementation results show that such a scheme is 43% more power efficient compared with the RS scheme with the same code rate. Besides, such an approach results in 46% faster computations and 53% reduction in memory requirements. Copyright © 2008 John Wiley & Sons, Ltd.

This paper presents work on using hierarchical long term memory to reduce the memory requirements of nearest sequence memory (NSM) learning, a previously published, instance-based reinforcement learning algorithm. A hierarchical memory representation reduces the memory requirements by allowing traces to share common sub-sequences. We present moderated mechanisms for estimating discounted future rewards and for dealing with hidden state using hierarchical memory. We also present an experimental analysis of how the sub-sequence length affects the memory compression achieved and show that the reduced memory requirements do not effect the speed of learning. Finally, we analyse and discuss the persistence of the sub-sequences independent of specific trace instances.

ABSTRACTAn alternative parameter‐free adaptive approach for the grouping of the basis function patterns in the multilevel fast multipole method is presented, yielding significant memory savings compared to the traditional Octree grouping for most discretizations, particularly when using higher‐order basis functions. Results from both a uniformly and nonuniformly meshed scatterer are presented, showing how the technique is worthwhile even for regular meshes, and demonstrating that there is no loss of accuracy in spite of the large reduction in memory requirements and the relatively low computational cost. © 2014 Wiley Periodicals, Inc. Microwave Opt Technol Lett 56:2451–2456, 2014

This paper presents a topology optimization method of compliant grippers considering stress constraints. The proportional topology optimization (PTO) algorithm is applied to the design of compliant grippers, and it is improved by introducing weight factors into the objective function and adding stress conditions on the basis of constraints. In the current gradient modeling of topology optimization, the global maximum stress is measured by P-norm function, and its sensitivity analysis of stress constraints is derived by adjoint equations. It is worth noting that more rigorous gradient calculations are employed in stress problems and their computation brings an additional computational burden. By contrast, the non-gradient method using PTO algorithm allocates design variables to the element proportionally according to the values of stress. It can eliminate difficulties in the analytical derivation and calculation of gradient, and improve the calculation efficiency. Subsequently, performances of compliant grippers generated by these two methods are compared through finite element analysis. Finally, the optimized compliant gripper prototype is manufactured by three-dimensional (3D) printing using flexible thermoplastic urethane. Experimental results indicate that the non-gradient method is effective, and the optimized compliant gripper has excellent characteristics of low stress and high output performance.

In this work we present key elements of DeepChip, a framework that bridges recent trends in machine learning with applicable forward propagation on resource-constrained devices. Main objective of this work is to reduce compute and memory requirements by removing redundancy from neural networks. DeepChip features a flexible quantizer to reduce the bit width of activations to 8-bit fixed-point and weights to an asymmetric ternary representation. In combination with novel algorithms and data compression we leverage reduced precision and sparsity for efficient forward propagation on a wide range of processor architectures. We validate our approach on a set of different convolutional neural networks and datasets: ConvNet on SVHN, ResNet-44 on CIFAR10 and AlexNet on ImageNet. Compared to single-precision floating point, memory requirements can be compressed by a factor of 43, 22 and 10 and computations accelerated by a factor of 5.2, 2.8 and 2.0 on a mobile processor without a loss in classification accuracy. DeepChip allows trading accuracy for efficiency, and for instance tolerating about 2% loss in classification accuracy further reduces memory requirements by a factor of 88, 29 and 13, and speeds up computations by a factor of 6.0, 4.3 and 5.0. Code related to this paper is available at: https://github.com/UniHD-CEG/ECML2018.

The field of genomics has seen a glorious explosion of high-quality data, with tremendous strides having been made in genomic sequencing instruments and computational genomics applications meant to make sense of the data. A common use case for genomics data is to answer the question if a specific genetic signature is correlated with some disease manifestations. Support Vector Machine (SVM) is a widely used classifier in computational literature. Previous studies have shown success in using these SVMs for the above use case of genomics data. However, SVMs suffer from a widely-recognized scalability problem in both memory use and computational time. It is as yet an unanswered question if training such classifiers can scale to the massive sizes that characterize many of the genomics data sets. We answer that question here for a specific dataset, in order to decipher whether some regulatory module of a particular combinatorial epigenetic “pattern” will regulate the expression of a gene. However, the specifics of the dataset is likely of less relevance to the claims of our work. We take a proposed theoretical technique for efficient training of SVM, namely Cascade SVM, create our classifier called EP-SVM, and empirically evaluate how it scales to the large genomics dataset. We implement Cascade SVM on the Apache Spark platform and open source this implementation1. Through our evaluation, we bring out the computational cost on each application process, the way of distributing the overall workload among multiple processes, which can potentially execute on different cores or different machines, and the cost of data transfer to different cores or different machines. We believe we are the first to shed light on the computational and network costs of training an SVM on a multi-dimensional genomics dataset. We also evaluate the accuracy of the classifier result as a function of the parameters of the SVM model.

Bolted joints are widely used in composite aircraft structures, for their assembly. The appropriate bolted joint configuration (hole/bolt diameter, pitch, etc.) is carefully selected during the detail design phase, where high fidelity numerical models are required with substantial computational cost and time. This work presents a design criterion, which allows the selection of the bolted joint configuration during the preliminary design phase with less computational time. The developed design criterion is based on a fully parametric finite element (FE) model, built in ANSYS V19 (Canonsburg, PA, USA), of a bolted joint with progressive damage modelling (PDM) capabilities, so that the failure of the joint can be predicted. From the numerical analyses, the bearing load and the load that bypasses the hole are calculated, up to failure, for a variety of joint configurations and loading conditions. The results of each analysis are used for plotting the failure envelope for the investigated bolted-joint configuration. Consequently, a design criterion is generated for the bolted joint. The availability of these failure envelopes, as design criterion, permit the appropriate selection of the bolted-joint configuration in an earlier design phase saving valuable time and computational cost.