Nonlocal models allow for autonomous initiation and propagation of cracks and damage, but they generally are significantly more expensive to compute compared with local counterparts. We introduce a fast convolution-based method for a correspondence model of peridynamics (PD) in ductile failure. The meshfree discretization of PD models, generally used for problems involving evolution of fracture/damage, uses simple numerical quadrature to approximate the value of the integral operator at every point in the PD domain, at every time step. This quadrature method, however, is expensive, especially in 3D since every node has hundreds/thousands of members in its family, leading to sums with hundreds/thousands of terms having to be computed at every node and every time-step. In order to speed-up such computations, we have introduced Fast Convolution-Based Methods (FCBMs) that exploit the convolutional structure of PD integral operators (see, e.g., [1]). Significant (orders of magnitude!) speedups and reduced memory requirements are observed compared with the regular meshfree solvers. The correspondence PD model was selected to simulate large plastic deformations, and the Johnson-Cook model is used for ductile failure. The domain and boundary conditions are implemented via masking functions. Examples of the method for brittle fracture and corrosion are available in the open-source code PeriFast [2]. Results for several examples of elasto-plastic failure also show that the peridynamic model is better able than classical models to capture ductile failure. REFERENCES [1] Jafarzadeh, S., Mousavi, F., Larios, A., Bobaru, F., 2022. A general and fast convolution-based method for peridynamics: Applications to elasticity and brittle fracture, Computer Methods in Applied Mechanics and Engineering, 392, 114666 (2022). [2] https://github.com/PeriFast/Code