Case Study: Application of Wireless Condition Based Monitoring by Applying Machine Learning Models

In oil and gas and petrochemical industries, high and medium critical rotating equipment e.g., Compressors, Pumps etc. are normally equipped with fixed machine vibration monitoring systems for the online monitoring and protection of the equipment. However, there are multiple low and medium critical and legacy equipment which are without fixed vibration monitoring system and where installing fixed vibration system is not financially practical. However, looking at the scale of a cost-efficient machine monitoring system and changing the maintenance approach from Preventive to Predictive can have significant financial as well as operational benefit. The intent behind performing this study was to assess the utilization of wireless vibration monitoring with following applications in One of Giant Oil and Gas Production Field (X Field) in ADNOC Onshore. Rotating equipment with failure history e.g., Instrument Air compressors, HVAC Compressors etc. Multiphase Pumps with no fixed vibration monitoring system. Centralized monitoring of remote equipment by deploying Private LPWA (Low Power Wide Area) Network over existing Telecom Backhaul consisting of Fiber Optics and WiMAX wireless networks. In the pursuit of some wireless vibration sensors with long range wireless coverage and on-premises monitoring and analytics application, ADNOC team identified a newly developed solution by renowned industrial instrumentation OEM. This system includes wireless LoRaWAN vibration sensors along with anomaly detection system based on the data collected by wireless sensors. To evaluate the system comparative effectiveness, a Proof of Concept was carried out in X Field by installing the solution at a pump already equipped with fixed vibration monitoring system at a remote facility 25 Km away from Central Plant. The data from these sensors was wirelessly transmitted to LoRaWAN gateway installed 200 meters away from the pump. From the gateway the data was routed to on-premises application server installed in a Central Facility utilizing existing Telecom Backhaul. A decision-based application was used for monitoring, trending, and automatic anomaly detection. At first stage the system was kept at machine learning phase to allow the system to learn the normal behavior of the Pump. Based on this learning data, an AI (Artificial Intelligence) based model was developed which self-assign a decision threshold for anomaly detection and alarming. This solution based on LPWAN (Lowe Power Wide Area Network) technology, LoRaWAN, can be utilized for Condition Based Monitoring, Trending and Anomaly detection of low and medium critical rotating equipment, where installation of fixed vibration monitoring system is not feasible. LoRaWAN sensors provide reliable wireless link up to 1 Km in congested plant installations with no requirement of Line of Sight. One LoRaWAN Gateway can support up to 1,000 Sensors. A cost comparison was also performed with traditional wired and this wireless solution and the later was found to be more cost effective with simplicity in deployment and no major footprints. Most of the available LPWA (LoRaWAN) solution are based on 3rd party connectivity e.g., GSM/Satellite and Cloud based Application Servers. Private LPWAN built on existing SCADA/Telecom infrastructure and on-premises application/network servers are best suited for the application where complete ownership of network/data is required.