The rolling bearing is a precision mechanical component that has several advantages, including minimal friction loss, high load-carrying capacity, low power consumption, and superior mechanical efficiency. Hence, novel methodologies for monitoring and diagnosing these rotating mechanisms are arising. Artificial intelligence is considered one of the most effective methods for analyzing collected data, such as vibration signals, to diagnose the functioning state of an asset. However, in most situations, a significant obstacle to the use of AI models in almost all industrial cases is model explainability, which is often referred to as “black boxes”. To address this problem, a new method called FaultDXAI (Fault Diagnosis using eXplainable AI) is developed. This method utilizes a multi-kernels-size CNN paired with an interpretable methodology to classify problems in rotating equipment. By utilizing gradient-weighted class activation mapping (denoted as Grad-CAM) in conjunction with multi-scale channels CNN (MSC-CNN), the application enables the retrospective analysis of outcomes. This approach on the CWRU-bearing dataset achieved both favorable diagnostic performance and the ability to acquire the features utilized by professionals to identify situations in one domain and apply them in another domain.