RCAIDE.Framework.Optimization.Packages.trmm.Trust_Region_Optimization.Trust_Region_Optimization#
- class Trust_Region_Optimization(*args, **kwarg)[source]#
Bases:
DataA trust region optimization
Assumptions: Only SNOPT is implemented
Source: None
- __init__(*args, **kwarg)#
Methods
accuracy_ratio(f_center, f_hi, f_corr, ...)Compute the trust region accuracy ratio.
calculate_constraint_violation(gval, lb, ub)Calculates the constraint violation using a 2-norm of the violated constraint values.
calculate_correction(f, df, g, dg, tr)Calculates additive correction factors.
evaluate_constraints(x[, problem, ...])Evaluates the RCAIDE nexus problem provides an objective value based on constraint violation.
evaluate_corrected_model(x[, problem, ...])Evaluates the RCAIDE nexus problem and applies corrections to the results.
evaluate_model(problem, x[, der_flag])Evaluates the RCAIDE nexus problem.
initialize_SLSQP_constraints(con, problem, ...)optimize(problem[, print_output])Optimizes the problem
scale_vals(inp, con, ini, bnd, scl)Scales inputs, constraints, and their bounds.
unpack_constraints_slsqp(x, con_ind, sign, ...)update_tr_size(rho, tr, accepted)Updates the trust region size based on the accuracy ratio and if it has been accepted.
- optimize(problem, print_output=False)[source]#
Optimizes the problem
Assumptions: Currently only works with SNOPT
Source: “A trust-region framework for managing the use of approximation models in optimization,” Alexandrov et. al., 1998 Details do not follow exactly.
Inputs: problem. <Nexus class> (also passed into other functions)
- optimization_problem.
inputs Numpy array matching standard RCAIDE optimization setup objective Numpy array matching standard RCAIDE optimization setup constraints Numpy array matching standard RCAIDE optimization setup
fidelity_level [-]
print_output <boolean> Determines if output is printed during the optimization run
Outputs: (fOpt_corr,xOpt_corr,str):
fOpt_corr <float> xOpt_corr <numpy array> str Varies depending on the result of the optimization
Properties Used: self.
trust_region_max_iterations [-] fidelity_levels [-] evaluation_order List of the fidelity level order evaluate_model(..) calculate_correction(..) calculate_constraint_violation(..) optimizer <string> Determines what optimizer is used evaluate_corrected_model(..) optimizer_max_iterations [-] optimizer_convergence_tolerance [-] optimizer_constraint_tolerance [-] optimizer_function_precision [-] optimizer_verify_level Int determining if SNOPT will verify that the minimum is level accuracy_ratio(..) update_tr_size(..) convergance_tolerance [-]
- evaluate_model(problem, x, der_flag=True)[source]#
Evaluates the RCAIDE nexus problem. This is often a mission evaluation.
Assumptions: None
Source: N/A
Inputs: problem. <Nexus class>
objective(..) all_constraints(..) finite difference(..)
x <numpy array> der_flag <boolean> Determines if finite differencing is done
Outputs: f - function value <float> df - derivative of f <numpy array> g - constraint value <numpy array> (only returned if der_flag is True) dg - jacobian of g <numpy array> (only returned if der_flag is True)
Properties Used: self.difference_interval [-]
- evaluate_corrected_model(x, problem=None, corrections=None, tr=None, return_cons=False)[source]#
Evaluates the RCAIDE nexus problem and applies corrections to the results.
Assumptions: None
Source: N/A
Inputs: problem. <Nexus class>
objective(..) all_constraints(..)
corrections <tuple> Contains correction factors tr.center <array>
Outputs: obj function objective cons list of contraint values fail indicates if the evaluation was successful
Properties Used: None
- evaluate_constraints(x, problem=None, corrections=None, tr=None, lb=None, ub=None)[source]#
Evaluates the RCAIDE nexus problem provides an objective value based on constraint violation. Correction factors are applied to the evaluation results.
Assumptions: None
Source: N/A
Inputs: problem. <Nexus class>
objective(..) all_constraints(..)
corrections <tuple> Contains correction factors tr.center <numpy array> lb <numpy array> lower bounds on the constraints up <numpy array> upper bounds on the constraints
Outputs: obj_cons objective based on constraint violation cons list of contraint values fail indicates if the evaluation was successful
Properties Used: self.calculate_constraint_violation(..)
- calculate_constraint_violation(gval, lb, ub)[source]#
Calculates the constraint violation using a 2-norm of the violated constraint values.
Assumptions: None
Source: N/A
Inputs: gval <numpy array> constraint values lb <numpy array> lower bounds on the constraints up <numpy array> upper bounds on the constraints
Outputs: constraint violation [-]
Properties Used: None
- calculate_correction(f, df, g, dg, tr)[source]#
Calculates additive correction factors.
Assumptions: None
Source: N/A
Inputs: f - function value <float> df - derivative of f <numpy array> g - constraint value <numpy array> (only returned if der_flag is True) dg - jacobian of g <numpy array> (only returned if der_flag is True)
Outputs: corr <tuple> correction factors
Properties Used: None
- scale_vals(inp, con, ini, bnd, scl)[source]#
Scales inputs, constraints, and their bounds.
Assumptions: None
Source: N/A
Inputs: (all RCAIDE format specific numpy arrays) inp Design variables con Constraint limits ini Initial values bnd Variable bounds scl Scaling factors
Outputs: x <numpy array> Scaled design variables scaled_constraints <numpy array> x_low_bound <numpy array> x_up_bound <numpy array> con_up_edge <numpy array> con_low_edge <numpy array> name <list of strings> List of variable names
Properties Used: None
- accuracy_ratio(f_center, f_hi, f_corr, g_viol_center, g_viol_hi, g_viol_corr, tr)[source]#
Compute the trust region accuracy ratio.
Assumptions: None
Source: N/A
Inputs: f_center Objective value at the center of the trust region f_hi High-fidelity objective value at the expected optimum f_corr Corrected low-fidelity objective value at the expected optimum g_viol_center Constraint violation at the center of the trust region g_viol_hi High-fidelity constraint violation at the expected optimum g_viol_corr Corrected low-fidelity constraint violation at the expected optimum tr.evaluation_function(..)
Outputs: rho [-] accuracy ratio
Properties Used: self.trust_region_function_precision [-]
- update_tr_size(rho, tr, accepted)[source]#
Updates the trust region size based on the accuracy ratio and if it has been accepted.
Assumptions: None
Source: N/A
Inputs: rho [-] accuracy ratio tr.
size [-] contraction_factor [-] contract_threshold [-] expand_threshold [-] expansion_factor [-]
Outputs: tr_action [-] number indicating the type of action done by the trust region
Properties Used: None