The digitization transformation of traditional machinery and advances in artificial intelligence have led to the development of data-driven machinery fault diagnosis methods. However, limited by the number of machinery equipment, it is challenging for small or medium sized manufacturing enterprises (SMEs) to collect sufficient data to support the effective execution of these methods. In addition, due to the potential conflicts of interest and risks of privacy leakage, direct sharing of raw data between enterprises is often impractical. To this end, a blockchain-empowered secure federated domain generalization (FDG) framework is proposed in this paper, aiming to achieve distributed collaborative machinery fault diagnosis. In this framework, blockchain technology is first employed to replace the central server in federated learning (FL) system, effectively mitigating the single-point-of-failure issue of the FL system. Second, a committee-based consensus mechanism is designed to verify the correctness of the global model. To achieve domain generalization (DG) in federated setting, two regularizers are incorporated into the proposed framework, which restrict the information contained in representation and perform implicit distribution alignment. Experimental studies on two datasets demonstrate that the proposed method outperforms state-of-the-art FDG methods in terms of diagnosis accuracy. The high reliability of the proposed framework makes it more suitable for practical industrial scenarios.