Reciprocating air compressors play a pivotal role in various industrial applications, demanding a steadfast focus on their dependable operation. This research addresses the challenge of fault diagnosis in reciprocating air compressors by leveraging the potential of ensemble voting classifiers. Vibration signals were captured under five compressor conditions, covering both normal operation and multiple fault scenarios, such as inlet valve fluttering, outlet valve fluttering, check valve faults, and combined inlet and outlet valve issues. A range of features, including statistical and autoregressive moving average features were extracted from the acquired vibration signals to effectively differentiate among these conditions. Feature selection was executed using the J48 algorithm to identify the most pertinent indicators. During the classification phase, an exhaustive evaluation of individual classifiers, such as K-Nearest Neighbors, J48, support vector machine, multilayer perceptron (MLP), Naïve Bayes, logistic regression, random forest (RF), and logistic model tree, was conducted. The top-performing classifiers were subsequently selected for incorporation into ensemble voting systems. Within the ensemble voting framework, various configurations were explored, ranging from two to five classifiers, employing multiple voting strategies to optimize fault diagnosis performance. The findings of this study showcase that the combination of MLP and RF classifiers for statistical features produced the maximum classification accuracy of 98.67%.