Bottleneck detection in production is essential to improve the overall manufacturing capacity. The common bottleneck detection methods have drawbacks of either practicability or accuracy. There is a recent manual methodology called bottleneck walk which is simple, accurate and easy to implement. Though it is a reliable and practical manual method, other more automated non-manual methods for detecting bottlenecks should be considered in the future. In this paper, the simulation models were built in a discrete event simulation software named Plant simulation. The models are based on real production cases where the bottleneck walk methodology had been tested. In the first case, the particular result shows that machine 2 has been the bottleneck for a long period, which is what is found in the real production. In the second case, the bottlenecks are found and the potential improvements are made on it. Though two improvement plans are made, neither of them solves the bottleneck and im-proves the system output. The plan of reducing buffer size might probably have an improve-ment on financial aspect. And further improvements on all findings remain available on models built in this paper.