<p>Conventional maximum likelihood estimation (MLE) methods, when utilizing data (including fault data and time-censored data) obtained from equipment experimental monitoring to estimate unknown distribution parameters and establish fault distribution models, often suffer from an imbalance in the utilization of these two data types, leading to reduced model credibility. To address this issue, this paper proposes an improved likelihood function method based on data weighting. The proposed method first employs the probability density function and the reliability function to process fault data and time-censored data, respectively. Subsequently, it recommends weighting each data point individually after assessment; however, it allows for unified weighting as a flexible alternative when assessment resources are limited. Finally, a weight optimization algorithm is developed, and the estimation results are evaluated based on the asymptotic normality principle of MLE and the Fisher information matrix. Using experimental monitoring data from a certain type of CNC machining center as a case study, the proposed method is compared with results from Bayesian modeling approaches used in other literature. The comparison verifies the feasibility and superiority of the proposed method.</p>