OpenMDAO issues an errant DerivativesWarning if design variables only impact linear constraints.

[robfalck]

Description

In _TotalJacInfo, method check_total_jac only considers the nonlinear jacobian. This is the jacobian used for coloring, as we don't want to consider linear responses for coloring.

However, this check will issue a warning if a design variable only impacts linear constraints, because it appears to have a column of zeros in the corresponding nonlinear total jacobian.

Example

Executing the following test in test_pyoptsparse_driver demonstrates the issue.

    def test_total_jac_warning_linear_cons(self):

        prob = om.Problem()
        model = prob.model

        model.add_subsystem('p1', om.IndepVarComp('x', 50.0), promotes=['*'])
        model.add_subsystem('p2', om.IndepVarComp('y', 50.0), promotes=['*'])
        model.add_subsystem('comp', Paraboloid(), promotes=['*'])
        model.add_subsystem('con', om.ExecComp('c = x - y + k'), promotes=['*'])

        prob.set_solver_print(level=0)

        prob.driver = pyOptSparseDriver()
        prob.driver.options['optimizer'] = OPTIMIZER
        if OPTIMIZER == 'SLSQP':
            prob.driver.opt_settings['ACC'] = 1e-9
        prob.driver.options['print_results'] = False

        model.add_design_var('x', lower=-50.0, upper=50.0)
        model.add_design_var('y', lower=-50.0, upper=50.0)
        model.add_design_var('k', lower=-1, upper=1)
        model.add_objective('f_xy')
        model.add_constraint('c', lower=15.0, linear=True)

        prob.setup()

        with assert_no_warning(category=om.DerivativesWarning):
            prob.run_driver()

OpenMDAO Version

3.31.2-dev

Relevant environment information

No response

[naylor-b]

Should we check for DVs that only have an impact on linear constraints and just not include them at all in the nonlinear jacobian in order to avoid zero columns?