PREDICTIVE MAINTENANCE STRATEGIES IN PROCESS PLANT EQUIPMENT: A REVIEW OF TECHNIQUES AND APPLICABILITY TO PETROCHEMICAL OPERATIONS IN NIGERIA

Abstract

Predictive maintenance (PdM) has become an important approach for improving equipment reliability and reducing unplanned downtime in process industries. However, many PdM studies focus on highly instrumented industrial environments and give limited attention to the infrastructure, skills, logistics and organisational conditions required for successful adoption in developing economies. This study combines a structured review of published literature with a qualitative case study of a Polypropylene Production Unit (PPU) in Nigeria to examine the technical capabilities and practical requirements for PdM adoption. Five major PdM techniques were examined: vibration analysis, infrared thermography, oil and lubricant analysis, ultrasonic and acoustic-emission methods, and artificial intelligence/machine-learning-based predictive models. The literature was synthesised based on fault-detection capabilities, data and infrastructure requirements, workforce skills, cost considerations and implementation challenges. The case study shows that the PPU combines preventive and corrective maintenance with routine “Look, Listen and Feel” (LLF) inspections, process instrumentation, and targeted use of vibration analysis, thermography, oil analysis, ultrasonic inspection and borescope examination. However, these activities are mainly initiated by observed abnormalities, maintenance requirements or engineering recommendations rather than through an integrated predictive analytics framework. The findings indicate that effective PdM adoption in such environments is better approached as a progressive process than as an immediate transition to fully connected, AI-driven monitoring. The paper therefore proposes a phased approach to PdM adoption that begins with reliable condition data and targeted monitoring, followed by systematic data integration and trend analysis, before progressing to online monitoring and advanced predictive models where sufficient data and infrastructure exist.