diff --git a/ebos/eclnewtonmethod.hh b/ebos/eclnewtonmethod.hh
new file mode 100644
index 000000000..fff1577c7
--- /dev/null
+++ b/ebos/eclnewtonmethod.hh
@@ -0,0 +1,232 @@
+// -*- mode: C++; tab-width: 4; indent-tabs-mode: nil; c-basic-offset: 4 -*-
+// vi: set et ts=4 sw=4 sts=4:
+/*
+ This file is part of the Open Porous Media project (OPM).
+
+ OPM is free software: you can redistribute it and/or modify
+ it under the terms of the GNU General Public License as published by
+ the Free Software Foundation, either version 2 of the License, or
+ (at your option) any later version.
+
+ OPM is distributed in the hope that it will be useful,
+ but WITHOUT ANY WARRANTY; without even the implied warranty of
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
+ GNU General Public License for more details.
+
+ You should have received a copy of the GNU General Public License
+ along with OPM. If not, see .
+
+ Consult the COPYING file in the top-level source directory of this
+ module for the precise wording of the license and the list of
+ copyright holders.
+*/
+/*!
+ * \file
+ *
+ * \copydoc Ewoms::EclNewtonMethod
+ */
+#ifndef EWOMS_ECL_NEWTON_METHOD_HH
+#define EWOMS_ECL_NEWTON_METHOD_HH
+
+#include
+#include
+
+#include
+
+BEGIN_PROPERTIES
+
+NEW_PROP_TAG(EclNewtonSumTolerance);
+NEW_PROP_TAG(EclNewtonStrictIterations);
+NEW_PROP_TAG(EclNewtonRelaxedVolumeFraction);
+NEW_PROP_TAG(EclNewtonSumToleranceExponent);
+NEW_PROP_TAG(EclNewtonRelaxedTolerance);
+
+END_PROPERTIES
+
+namespace Ewoms {
+
+/*!
+ * \brief A newton solver which is ebos specific.
+ */
+template
+class EclNewtonMethod : public BlackOilNewtonMethod
+{
+ typedef BlackOilNewtonMethod ParentType;
+ typedef typename GET_PROP_TYPE(TypeTag, DiscNewtonMethod) DiscNewtonMethod;
+
+ typedef typename GET_PROP_TYPE(TypeTag, Simulator) Simulator;
+ typedef typename GET_PROP_TYPE(TypeTag, FluidSystem) FluidSystem;
+ typedef typename GET_PROP_TYPE(TypeTag, SolutionVector) SolutionVector;
+ typedef typename GET_PROP_TYPE(TypeTag, GlobalEqVector) GlobalEqVector;
+ typedef typename GET_PROP_TYPE(TypeTag, PrimaryVariables) PrimaryVariables;
+ typedef typename GET_PROP_TYPE(TypeTag, EqVector) EqVector;
+ typedef typename GET_PROP_TYPE(TypeTag, Indices) Indices;
+ typedef typename GET_PROP_TYPE(TypeTag, Scalar) Scalar;
+ typedef typename GET_PROP_TYPE(TypeTag, Linearizer) Linearizer;
+ typedef typename GET_PROP_TYPE(TypeTag, ElementContext) ElementContext;
+
+ static const unsigned numEq = GET_PROP_VALUE(TypeTag, NumEq);
+
+ static constexpr int contiSolventEqIdx = Indices::contiSolventEqIdx;
+ static constexpr int contiPolymerEqIdx = Indices::contiPolymerEqIdx;
+ static constexpr int contiEnergyEqIdx = Indices::contiEnergyEqIdx;
+
+ friend NewtonMethod;
+ friend DiscNewtonMethod;
+ friend ParentType;
+
+public:
+ EclNewtonMethod(Simulator& simulator) : ParentType(simulator)
+ {
+ errorPvFraction_ = 1.0;
+ relaxedMaxPvFraction_ = EWOMS_GET_PARAM(TypeTag, Scalar, EclNewtonRelaxedVolumeFraction);
+
+ sumTolerance_ = 0.0; // this gets determined in the error calculation proceedure
+ relaxedTolerance_ = EWOMS_GET_PARAM(TypeTag, Scalar, EclNewtonRelaxedTolerance);
+
+ numStrictIterations_ = EWOMS_GET_PARAM(TypeTag, int, EclNewtonStrictIterations);
+ }
+
+ /*!
+ * \brief Register all run-time parameters for the Newton method.
+ */
+ static void registerParameters()
+ {
+ ParentType::registerParameters();
+
+ EWOMS_REGISTER_PARAM(TypeTag, Scalar, EclNewtonSumTolerance,
+ "The maximum error tolerated by the Newton"
+ "method for considering a solution to be "
+ "converged");
+ EWOMS_REGISTER_PARAM(TypeTag, int, EclNewtonStrictIterations,
+ "The number of Newton iterations where the"
+ " volumetric error is considered.");
+ EWOMS_REGISTER_PARAM(TypeTag, Scalar, EclNewtonRelaxedVolumeFraction,
+ "The fraction of the pore volume of the reservoir "
+ "where the volumetric error may be voilated during "
+ "strict Newton iterations.");
+ EWOMS_REGISTER_PARAM(TypeTag, Scalar, EclNewtonSumToleranceExponent,
+ "The the exponent used to scale the sum tolerance by "
+ "the total pore volume of the reservoir.");
+ EWOMS_REGISTER_PARAM(TypeTag, Scalar, EclNewtonRelaxedTolerance,
+ "The maximum error which the volumetric residual "
+ "may exhibit if it is in a 'relaxed' "
+ "region during a strict iteration.");
+ }
+
+ /*!
+ * \brief Returns true if the error of the solution is below the
+ * tolerance.
+ */
+ bool converged() const
+ {
+ if (errorPvFraction_ < relaxedMaxPvFraction_)
+ return (this->error_ < relaxedTolerance_ && errorSum_ < sumTolerance_) ;
+ else if (this->numIterations() > numStrictIterations_)
+ return (this->error_ < relaxedTolerance_ && errorSum_ < sumTolerance_) ;
+
+ return this->error_ <= this->tolerance() && errorSum_ <= sumTolerance_;
+ }
+
+ void preSolve_(const SolutionVector& currentSolution OPM_UNUSED,
+ const GlobalEqVector& currentResidual)
+ {
+ const auto& constraintsMap = this->model().linearizer().constraintsMap();
+ this->lastError_ = this->error_;
+ Scalar newtonMaxError = EWOMS_GET_PARAM(TypeTag, Scalar, NewtonMaxError);
+
+ // calculate the error as the maximum weighted tolerance of
+ // the solution's residual
+ this->error_ = 0.0;
+ Dune::FieldVector componentSumError;
+ std::fill(componentSumError.begin(), componentSumError.end(), 0.0);
+ Scalar sumPv = 0.0;
+ errorPvFraction_ = 0.0;
+ const Scalar dt = this->simulator_.timeStepSize();
+ for (unsigned dofIdx = 0; dofIdx < currentResidual.size(); ++dofIdx) {
+ // do not consider auxiliary DOFs for the error
+ if (dofIdx >= this->model().numGridDof()
+ || this->model().dofTotalVolume(dofIdx) <= 0.0)
+ continue;
+
+ if (!this->model().isLocalDof(dofIdx))
+ continue;
+
+ // also do not consider DOFs which are constraint
+ if (this->enableConstraints_()) {
+ if (constraintsMap.count(dofIdx) > 0)
+ continue;
+ }
+
+ const auto& r = currentResidual[dofIdx];
+ const double pvValue =
+ this->simulator_.problem().porosity(dofIdx)
+ * this->model().dofTotalVolume(dofIdx);
+ sumPv += pvValue;
+ bool cnvViolated = false;
+
+ for (unsigned eqIdx = 0; eqIdx < r.size(); ++eqIdx) {
+ Scalar tmpError = r[eqIdx] * dt * this->model().eqWeight(dofIdx, eqIdx) / pvValue;
+ Scalar tmpError2 = r[eqIdx] * this->model().eqWeight(dofIdx, eqIdx);
+
+ this->error_ = Opm::max(std::abs(tmpError), this->error_);
+
+ if (std::abs(tmpError) > this->tolerance_)
+ cnvViolated = true;
+
+ componentSumError[eqIdx] += std::abs(tmpError2);
+ }
+ if (cnvViolated)
+ errorPvFraction_ += pvValue;
+ }
+
+ // take the other processes into account
+ this->error_ = this->comm_.max(this->error_);
+ componentSumError = this->comm_.sum(componentSumError);
+ sumPv = this->comm_.sum(sumPv);
+ errorPvFraction_ = this->comm_.sum(errorPvFraction_);
+
+ componentSumError /= sumPv;
+ componentSumError *= dt;
+
+ errorPvFraction_ /= sumPv;
+
+ errorSum_ = 0;
+ for (unsigned eqIdx = 0; eqIdx < numEq; ++eqIdx)
+ errorSum_ = std::max(std::abs(componentSumError[eqIdx]), errorSum_);
+
+ // scale the tolerance for the total error with the pore volume. by default, the
+ // exponent is 1/3, i.e., cubic root.
+ Scalar x = EWOMS_GET_PARAM(TypeTag, Scalar, EclNewtonSumTolerance);
+ Scalar y = EWOMS_GET_PARAM(TypeTag, Scalar, EclNewtonSumToleranceExponent);
+ sumTolerance_ = x*std::pow(sumPv, y);
+
+ // make sure that the error never grows beyond the maximum
+ // allowed one
+ if (this->error_ > newtonMaxError)
+ throw Opm::NumericalIssue("Newton: Error "+std::to_string(double(this->error_))
+ +" is larger than maximum allowed error of "
+ +std::to_string(double(newtonMaxError)));
+
+ // make sure that the error never grows beyond the maximum
+ // allowed one
+ if (errorSum_ > newtonMaxError)
+ throw Opm::NumericalIssue("Newton: Sum of the error "+std::to_string(double(errorSum_))
+ +" is larger than maximum allowed error of "
+ +std::to_string(double(newtonMaxError)));
+ }
+
+private:
+ Scalar errorPvFraction_;
+ Scalar errorSum_;
+
+ Scalar relaxedTolerance_;
+ Scalar relaxedMaxPvFraction_;
+
+ Scalar sumTolerance_;
+
+ int numStrictIterations_;
+};
+} // namespace Ewoms
+
+#endif
diff --git a/ebos/eclproblem.hh b/ebos/eclproblem.hh
index b86a8ec3d..faed62b7b 100644
--- a/ebos/eclproblem.hh
+++ b/ebos/eclproblem.hh
@@ -60,6 +60,7 @@
#include "ecldummygradientcalculator.hh"
#include "eclfluxmodule.hh"
#include "eclbaseaquifermodel.hh"
+#include "eclnewtonmethod.hh"
#include "ecltracermodel.hh"
#include "vtkecltracermodule.hh"
@@ -235,14 +236,35 @@ SET_SCALAR_PROP(EclBaseProblem, EndTime, 1e100);
// not millions of trillions of years, that is...)
SET_SCALAR_PROP(EclBaseProblem, InitialTimeStepSize, 1e100);
-// increase the default raw tolerance for the newton solver to 10^-4 because this is what
-// everone else seems to be doing...
-SET_SCALAR_PROP(EclBaseProblem, NewtonRawTolerance, 1e-4);
+// the default for the allowed volumetric error for oil per second
+SET_SCALAR_PROP(EclBaseProblem, NewtonRawTolerance, 1e-2);
-// reduce the maximum allowed Newton error to 0.1 kg/(m^3 s). The rationale is that if
-// the error is above that limit, the time step is unlikely to succeed anyway and we can
-// thus abort the futile attempt early.
-SET_SCALAR_PROP(EclBaseProblem, NewtonMaxError, 0.1);
+// the tolerated amount of "incorrect" amount of oil per time step for the complete
+// reservoir. this is scaled by the pore volume of the reservoir, i.e., larger reservoirs
+// will tolerate larger residuals.
+SET_SCALAR_PROP(EclBaseProblem, EclNewtonSumTolerance, 1e-4);
+
+// set the exponent for the volume scaling of the sum tolerance: larger reservoirs can
+// tolerate a higher amount of mass lost per time step than smaller ones! since this is
+// not linear, we use the cube root of the overall pore volume by default, i.e., the
+// value specified by the NewtonSumTolerance parameter is the "incorrect" mass per
+// timestep for an reservoir that exhibits 1 m^3 of pore volume. A reservoir with a total
+// pore volume of 10^3 m^3 will tolerate 10 times as much.
+SET_SCALAR_PROP(EclBaseProblem, EclNewtonSumToleranceExponent, 1.0/3.0);
+
+// set number of Newton iterations where the volumetric residual is considered for
+// convergence
+SET_INT_PROP(EclBaseProblem, EclNewtonStrictIterations, 8);
+
+// set fraction of the pore volume where the volumetric residual may be violated during
+// strict Newton iterations
+SET_SCALAR_PROP(EclBaseProblem, EclNewtonRelaxedVolumeFraction, 0.03);
+
+// the maximum volumetric error of a cell in the relaxed region
+SET_SCALAR_PROP(EclBaseProblem, EclNewtonRelaxedTolerance, 1e9);
+
+// Ignore the maximum error mass for early termination of the newton method.
+SET_SCALAR_PROP(EclBaseProblem, NewtonMaxError, 10e9);
// set the maximum number of Newton iterations to 14 because the likelyhood that a time
// step succeeds at more than 14 Newton iteration is rather small
@@ -281,6 +303,10 @@ SET_TYPE_PROP(EclBaseProblem, FluxModule, Ewoms::EclTransFluxModule);
// Use the dummy gradient calculator in order not to do unnecessary work.
SET_TYPE_PROP(EclBaseProblem, GradientCalculator, Ewoms::EclDummyGradientCalculator);
+// Use a custom Newton-Raphson method class for ebos in order to attain more
+// sophisticated update and error computation mechanisms
+SET_TYPE_PROP(EclBaseProblem, NewtonMethod, Ewoms::EclNewtonMethod);
+
// The frequency of writing restart (*.ers) files. This is the number of time steps
// between writing restart files
SET_INT_PROP(EclBaseProblem, RestartWritingInterval, 0xffffff); // disable