Monitoring equipment conditions is of great value in manufacturing, which can not only reduce unplanned downtime by early detecting anomalies of equipment but also avoid unnecessary routine maintenance. With the coming era of Industry 4.0 (or industrial internet), more and more assets and machines in plants are equipped with various sensors and information systems, which brings an unprecedented opportunity to capture large-scale and fine-grained data for effective on-line equipment condition monitoring. However, due to the lack of systematic methods, analysts still find it challenging to carry out efficient analyses and extract valuable information from the mass volume of data collected, especially for process industry (e.g., a petrochemical plant) with complex manufacturing procedures. In this paper, we report the design and implementation of an interactive visual analytics system, which helps managers and operators at manufacturing sites leverage their domain knowledge and apply substantial human judgements to guide the automated analytical approaches, thus generating understandable and trustable results for real-world applications. Our system integrates advanced analytical algorithms (e.g., Gaussian mixture model with a Bayesian framework) and intuitive visualization designs to provide a comprehensive and adaptive semi-supervised solution to equipment condition monitoring. The example use cases based on a real-world manufacturing dataset and interviews with domain experts demonstrate the effectiveness of our system.