Since the effusion of Industry 4.0 (I40) and smart manufacturing, predictive maintenance (PdM) has become critical to prevent severe system breakdowns and costly production downtime in various industries. Several state-of-the-art Artificial Intelligence (AI) approaches, such as machine learning models (ML), empower the PdM design concept to produce more accurate outcomes. In this study, we propose the development of a PdM platform that effectively detects future possible warnings and failures in the vibration speed sensor of a conveyor motor. We build the platform backend using recurrent neural networks (RNN) regression ML models to predict vibration speed values and their trends. We train our RNN models using univariate time-series historical vibration sensor data recorded over the years. The experimental results demonstrate the robustness of our platform built on RNN models compared to other traditional regression machine learning models such as Extreme Gradient Boosting (XGboost). We compare the effectiveness of the PdM platform with two RNN models: a basic RNN model and a long short-term memory (LSTM) model. Unlike other regression PdM systems focusing on predicting devices Remaining Useful Life (RUL), our PdM platform forecasts and generates new vibration sensor data based on the desired future time range. It also pinpoints set warning and failure values and indicates when these faults are more likely to occur. Factories can utilize this feature for analytics purposes.