The research in the paper is about the hardware machine learning study of an intelligent neuro-fuzzy system (NFS). The NFS is embedded within a DSP-FPGA chip system. The well-known random optimization method is used as the learning algorithm for the NFS. It is applied to a laboratory-scale temperature control process to study the hardware learning ability. The experiment results show the intelligent hardware system is able to achieve the machine-learning task with good performance. This encourages us the future study of intelligent hardware systems.