The implementation of discrete-event simulation and demand forecasting using Temporal Fusion Transformers to validate spare parts inventory policy for petrochemicals industry

One of the important strategic factors petrochemicals plant maintenance is spare parts inventory policy, It significant for efficiency, reliability, and productivity industry. An unsuitable part policy will lead to a loss in long engineering machinery downtime due shortage parts. To implement which able fulfill future demand parts, calculation by various statistical theories working processes used custom policy. However, validate whether or not customized suitable, discrete-event simulation library SimPy mimic actual system. must be involved performance evaluation process The model consists many events depending on supply chain crucial event most complex event. This work applies forecasting technique using deep learning with prebuilt architecture called Temporal Fusion Transformers (TFT). averaged MAE point predictions from global 0.4874+/- 6.7744 validation dataset 0.6424+/-3.4963 test dataset. Our method predicts quantile forecast handle stochastic nature industry result outcome more accurate close an information analysis management team make decisions about before deploying it