[en] GRESS is a FORTRAN precompiler and run time library which automatically processes computer models and adds derivative-taking capabilities to the normal calculated results. The GRESS system is under development at ORNL and work is proceeding in upgrading test versions of the precompiler to allow processing by GRESS of most of the ANSI X3.9-1978 FORTRAN syntax. Heretofore, a major limitation has been the inability of GRESS to process arrays with more than two dimensions. Consequently, a significant amount of human effort was required to convert three- and higher-dimensional arrays into two-dimensional arrays for processing with GRESS. This report describes the addition of ten operation codes to the GRESS precompiler and run time library to allow application to FORTRAN programs with three-dimensional arrays. The new operation codes were tested on a sample problem. A comparison of CPU time is made for sample problems run on a VAX 8600 computer with one-, two-, and three-dimensional arrays. It is recommended that GRESS be further modified to accommodate higher-dimensional arrays up to the ANSI X3.9-1978 limit of seven dimensions

[en] WAPPA-C is a waste package performance assessment code that predicts the temporal and spatial extent of the loss of containment capability of a given waste package design. This code was enhanced by the addition of the capability to calculate the sensitivity of model results to any parameter. The GRESS automated procedure was used to add this capability in only two man-months of effort. The verification analysis of the enhanced code, WAPPAG, showed that the sensitivities calculated using GRESS were accurate to within the precision of perturbation results against which the sensitivities were compared. Sensitivities of all summary table values to eight diverse data values were verified

[en] New computer graphic developments at Oak Ridge National Ridge National Laboratory (ORNL) are being used to provide visualization of criticality safety models and calculational results as well as tools for criticality safety analysis input preparation. The purpose of this paper is to present the status of current development efforts to continue to enhance the SCALE (Standardized Computer Analyses for Licensing Evaluations) computer software system. Applications for criticality safety analysis in the areas of 3-D model visualization, input preparation and execution via a graphical user interface (GUI), and two-dimensional (2-D) plotting of results are discussed

[en] Criticality safety analyses often require detailed modeling of complex geometries. Effective visualization tools can enhance checking the accuracy of these models. This report describes the KENO3D visualization tool developed at the Oak Ridge National Laboratory (ORNL) to provide visualization of KENO V.a and KENO-VI criticality safety models. The development of KENO3D is part of the current efforts to enhance the SCALE (Standardized Computer Analyses for Licensing Evaluations) computer software system

[en] In recent years, two automated systems have been developed that add derivative taking capability to existing FORTRAN codes. The derivative information is used to calculate sensitivities of model results to input data. The two systems, GRESS and ADGEN, solve the forward and adjoint sensitivity problems, respectively. Both systems are based on a precompiler that automates the calculation of derivatives of the real variable on the left-hand-side (LHS) of each equation with respect to the real variables on the right-hand-side (RHS) of each equation. This paper describes the features of the extended arithmetic processor (EXAP) precompiler that makes the automation of differentiation capability practical

[en] A sensitivity analysis of EQ3, a computer code which has been proposed to be used as one link in the overall performance assessment of a national high-level waste repository, has been performed. EQ3 is a geochemical modeling code used to calculate the speciation of a water and its saturation state with respect to mineral phases. The model chosen for the sensitivity analysis is one which is used as a test problem in the documentation of the EQ3 code. Sensitivities are calculated using both the CHAIN and ADGEN options of the GRESS code compiled under G-float FORTRAN on the VAX/VMS and verified by perturbation runs. The analyses were performed with a preliminary Version 1.0 of GRESS which contains several new algorithms that significantly improve the application of ADGEN. Use of ADGEN automates the implementation of the well-known adjoint technique for the efficient calculation of sensitivities of a given response to all the input data. Application of ADGEN to EQ3 results in the calculation of sensitivities of a particular response to 31,000 input parameters in a run time of only 27 times that of the original model. Moreover, calculation of the sensitivities for each additional response increases this factor by only 2.5 percent. This compares very favorably with a running-time factor of 31,000 if direct perturbation runs were used instead. 6 refs., 8 tabs

[en] GRESS is an efficient automated software system for calculating sensitivities of model results to input data in existing FORTRAN 77 computer models. The GRESS ADGEN option calculates first derivatives of selected model results with respect to an almost unlimited number of data by applying the adjoint method; however, data storage has been a major limiting factor. The data storage requirement for an ADGEN application was more than 322 Mbytes for a code that executes in 1 min on a VAX 8600 computer. The purpose of this paper is to describe two new algorithms that dramatically reduce the data storage requirement and execution time for application of the ADGEN option

[en] This paper presents the development of a FORTRAN compiler and an associated supporting software library called ADGEN. ADGEN reads FORTRAN models as input and produces and enhanced version of the input model. The enhanced version reproduces the original model calculations but also has the capability to calculate derivatives of model results of interest with respect to any and all of the model data and input parameters. The method for calculating the derivatives and sensitivities is the adjoint method. Partial derivatives are calculated analytically using computer calculus and saved as elements of an adjoint matrix on direct assess storage. The total derivatives are calculated by solving an appropriate adjoint equation. ADGEN is applied to a major computer model of interest to the Low-Level Waste Community, the PRESTO-II model. PRESTO-II sample problem results reveal that ADGEN correctly calculates derivatives of response of interest with respect to 300 parameters. The execution time to create the adjoint matrix is a factor of 45 times the execution time of the reference sample problem. Once this matrix is determined, the derivatives with respect to 3000 parameters are calculated in a factor of 6.8 that of the reference model for each response of interest. For a single 3000 for determining these derivatives by parameter perturbations. The automation of the implementation of the adjoint technique for calculating derivatives and sensitivities eliminates the costly and manpower-intensive task of direct hand-implementation by reprogramming and thus makes the powerful adjoint technique more amenable for use in sensitivity analysis of existing models. 20 refs., 1 fig., 5 tabs

[en] The primary objective of this manual is to provide a description of the Gradient Enhanced Software System (GRESS) and to explain how to use GRESS to enhance FORTRAN 77 models for gradient calculation. The use of the Extended Arithmetic Processor (EXAP) as the precompiler for GRESS is presented. A complete description of how to enhance a source code for forward propagation of derivatives using the calculus chain rule is provided. On option, EXAP can be used to generate derivatives and store them on a direct access device for subsequent solution of the numerical adjoint equations. Programming information is also provided to aid in the installation and maintenance of the software

[en] A system that can automatically enhance computer codes with a sensitivity calculation capability is presented. With this new system, named ADGEN, rapid and cost-effective calculation of sensitivities can be performed in any FORTRAN code for all input data or parameters. The resulting sensitivities can be used in performance assessment studies related to licensing or interactions with the public to systematically and quantitatively prove the relative importance of each of the system parameters in calculating the final performance results. A general procedure calling for the systematic use of sensitivities in assessment studies is presented. The procedure can be used in modeling and model validation studies to avoid over modeling, in site characterization planning to avoid over collection of data, and in performance assessments to determine the uncertainties on the final calculated results. The added capability to formally perform the inverse problem, i.e., to determine the input data or parameters on which to focus to determine the input data or parameters on which to focus additional research or analysis effort in order to improve the uncertainty of the final results, is also discussed. 7 references, 2 figures