Abstract: Fault diagnosis is a critical technology for ensuring production safety and maintaining product quality within the chemical industry. However, many contemporary fault diagnosis methods leveraging deep neural networks are typically trained using samples from specific operating conditions, often neglecting the domain generalization capabilities of these models. This oversight leads to a marked decline in diagnostic performance when these models encounter the diverse and complex conditions present in real industrial scenarios. This paper proposes a weighted-based domain-specific feature removal network (WBDSFRN) to address this challenge. WBDSFRN integrates a weighting-based domain-specific feature removal module, which effectively distinguishes between domain-invariant and domain-specific features during the training phase. In the testing phase, the model partially eliminates domain-specific features from the target domain, thereby mitigating domain shift and enhancing the model’s generalization capabilities. The proposed method’s effectiveness is rigorously evaluated through comparative studies and ablation experiments using the Tennessee Eastman process. The results demonstrate that the WBDSFRN significantly outperforms existing approaches, exhibiting robust diagnostic performance even under complex operating conditions.