One of the major drawbacks of combustion engines is their combustion noise. To mitigate this issue, exhaust systems including mufflers are commonly mounted on engines. As community noise and work environment regulations become increasingly more stringent, engine and muffler manufacturers must keep improving the acoustic performances of their products. While the main purpose of this system is to reduce the intensity of engine orders, the induced back pressure must be kept minimum to guarantee optimal engine operating conditions. Achieving such performances, however, implies the increasing complexity of muffler designs, often leading to the emergence of undesired noise produced by the flow circulating inside a muffler, or muffler self-noise. Addressing those issues early in the development process using an experimental process based on prototyping can be complex, time-consuming and expensive. Numerical approaches are an alternative. However, capturing non-linear dissipation mechanisms and thermal fluctuations of exhaust flows is challenging, while necessary to accurately predict flow noise. Transient and compressible Computational Fluid Dynamics and Computational AeroAcoustics (CFD/CAA) Lattice-Boltzmann based Methods (LBM) represent a valuable solution. Such methods have previously been successfully applied to quantify the flow noise generated by mufflers mounted on an experimental cold flow test bench. In this paper, the accuracy of the method for self-noise predictions of a muffler mounted on a single cylinder engine is proven comparing results with experimental noise measurements obtained in an anechoic room. The inlet boundary condition of the muffler is calculated using a 1-D system modeling tool, providing oscillating transient temperature and mass flow. Measured temperature on the surface of the muffler is prescribed in the simulation on the geometry walls. After validation of the results, the simulation data is further post-processed to identify flow-induced noise sources in the system and propose design changes to reduce self-noise. This study confirms that this numerical approach can be used in the production process to quantify and reduce flow noise in mufflers.