Track papers across labs, topics, and trends.
Search papers, labs, and topics across Lattice.
This paper introduces PL-KKT-hPINN, a novel framework that enforces nonlinear equality constraints in physics-informed neural networks (PINNs) using piecewise-linear projection, addressing the limitations of soft constraint enforcement during training. The method extends the existing KKT-hPINN framework, which only handles linear constraints, by ensuring strict adherence to nonlinear constraints while maintaining predictive accuracy. Experimental results on a continuous stirred-tank reactor (CSTR) case study reveal that PL-KKT-hPINN not only achieves significantly lower constraint violations but also demonstrates improved performance in low-data scenarios compared to traditional unconstrained neural networks.
Nonlinear constraints can now be strictly enforced in neural networks without sacrificing predictive accuracy, revolutionizing how we model complex physical systems.
While physics-informed neural networks (PINNs) have shown strong potential for process modeling, physical equations are only enforced as soft constraints during training, and thus, they do not guarantee constraint satisfaction at inference. We propose a framework, called piecewise-linear Karush--Kuhn--Tucker hard-constrained PINNs (PL-KKT-hPINNs), that strictly enforces nonlinear equality constraints through piecewise-linear projection. This extends the KKT-hPINN framewor, which exactly enforces linear equalities through the Karush--Kuhn--Tucker (KKT) conditions associated with orthogonally projecting neural network outputs onto the constraint feasible region. The method is demonstrated on a continuous stirred-tank reactor (CSTR) case study for both one and two inputs. Results show that PL-KKT-hPINN preserves predictive accuracy comparable to that of a standard neural network while achieving substantially lower constraint violations. In addition, the proposed model shows improved robustness in low-data regimes, yielding lower RMSE than the unconstrained neural network for limited training sample sizes. These results demonstrate that PL-KKT-hPINN provides a computationally efficient and physically consistent framework for surrogate modeling of nonlinear chemical engineering systems.