an output, such as being non-linear. This term can be assigned a high

weight and HASSAL will attempt to not assign it as the output.

5.3.3 Preliminary Function-Variable Output Set Assignment

 After the index outputs are assigned the variable types are

assigned as outputs to the function types. This is accomplished by a

call to the subroutine "I3FV0S." This subroutine also makes use of the

subroutine HASSAL. In this case, however, the aim in assigning outputs

is to enhance the possibility of convergence of the solution procedure

derived. This is done by maximizing the product of the first deriv-

atives of the function with respect to their outputs (Edie, 1970). This

cannot be done rigorously in the case of indexed equations. The pro-

cedure adopted is to use approximate variable values to estimate the

derivatives for an expanded Jacobian matrix whose size is dictated by

the index blocking factor. Output products for each function and

variable are then calculated and inserted into the FVIN. This is

necessary to insure that the outputs are assigned by function and

variable type. The maximum product is achieved by having HASSAL rini-

mize the sum of the negative of the logarithms of the calculated

weights.

5.3.4 Minimum Weighted Tearing

 After the set of equations is completely output set assigned the

torn, or recycle, variables must be chosen. These are chosen by

applying the tearing algorithm of Pho and Lapidus (1973). The algorithm

allows variables to be assigned weights and minimizes the sum of the

weights of the torn variables. The weighted tearing is performed by

the subroutine "ALWET." The weights assigned ideally should reflect

the difficulty in converging the various variables, with difficult to