Asset management is critical for improving operational efficiency and safety in the process industry. Predictive maintenance is a powerful data-driven approach to increase the reliability and lifespan of assets. Despite its vast benefits, organisations must be cautious about implementing predictive maintenance programmes.
Maintenance is important for operational efficiency in the manufacturing industry. It ensures that there are no unexpected disruptions in the production schedule of the enterprise. Proper maintenance goes a long way in increasing the lifespan of plant assets and avoids unexpected capital expenditure to replace equipment or machinery. Mechanical integrity of plant assets is one of the 14 elements of Process Safety Management. Sudden and catastrophic failure of equipment and machinery can lead to disastrous consequences for personnel, property and the environment.
Reactive Maintenance
Traditionally, maintenance has been reactive. There are no scheduled inspections and equipment is allowed to “run to failure”. Reactive maintenance comes into play after the equipment breakdown and involves restoring it back to full functionality. Reactive maintenance pushes the asset to its limits. But it is clearly unsustainable in the long term.
Preventive Maintenance
Preventive maintenance is regularly and routinely carried out to reduce the probability of equipment failure. It ensures the efficient and safe operation of the plant and reduces the frequency of unplanned downtime. Preventive maintenance can be both time-based as well as condition-based. Time-based approach schedules maintenance of equipment at predetermined intervals, usually on the basis of recommendations from the manufacturer. Condition-based maintenance on the other hand is triggered by signs of underperformance and an impending threat of failure. For instance, a pump or compressor may be pulled out for maintenance when the vibration reaches a certain threshold level. Scheduling of preventive maintenance is based on the expertise accumulated from long hours of operating the asset, both within the premises and elsewhere. Preventive maintenance reduces unplanned downtime, which is estimated to cost the chemical industry $20 billion every year.
Predictive Maintenance
Predictive maintenance is a data-driven approach to determine when assets need maintenance so as to improve their reliability and lifespan. Predictive maintenance continuously monitors the fitness of various assets in a plant. It involves collecting a lot of data of different kinds and using them to build a dynamic model. The model is then constantly analysed using machine learning algorithms to identify probable causes of the failure of the asset and make timely, accurate and specific recommendations for maintenance. Even the smallest deviation in performance is analysed to identify their root cause. This improves the mean time between failures and the longevity of the asset. It also reduces downtime and costs incurred on unnecessary maintenance. The strength of AI-based predictive maintenance lies in processing voluminous amounts of data and providing insights in real-time. Artificial Intelligence can be very useful in implementing predictive maintenance systems in a manufacturing plant. It can trawl through vast amounts of complex data from equipment, machinery and other plant assets and provide insightful information to improve their operation, maintenance and service life.
Data
To be effective, predictive maintenance needs a vast amount of data. These include process parameters like pressure, temperature and composition and also machinery-related data like speed, vibration, acoustics and bearing temperatures. infrared thermography is a non-intrusive testing technology that is widely used in predictive maintenance. With infrared cameras, maintenance personnel can detect above-normal temperatures in equipment. Historical data on past breakdowns and maintenance records are also important inputs to build a predictive maintenance model. General external environmental data like temperature and humidity are also taken into account. Data is collected by sensors and IoT devices.
How does it work?
Predictive maintenance uses data analytics to detect anomalies in operation. Both real-time and historical data are used to anticipate problems before they happen. One of the key tools used in predictive maintenance is time series analysis. Data is collected over time to monitor the performance and health of equipment. Time series analysis involves techniques such as trend analysis and spectral analysis. Trend analysis is used to identify long-term patterns in the data, while spectral analysis is used to identify patterns in the data that repeat over time. The goal is to find patterns that can help predict and ultimately prevent failures. Machine learning algorithms can be trained using historical data to look for tell-tale signs of imminent failure and initiate remedial actions. Time series analysis can also be used to identify the root cause of a failure, which can help organisations to develop more effective maintenance strategies.
Timing preventive maintenance requires a judicious mix of intuitive understanding and experience, a skill set that is not easy to come by. Predictive maintenance avoids this uncertainty with accurate and machine-specific recommendations. The confidence built up by predictive maintenance can be used to optimise and rationalise the inventory of spare parts.
Advantages
The annual maintenance cost in the chemical industry is approximately 1.8% to 2% of the plant replacement value. In a poorly maintained plant, this could be as high as 5%. Further, it is estimated that one-third of the maintenance costs are unnecessary or improperly carried out. Preventive maintenance is believed to be undertaken unnecessarily 50% of the time in many manufacturing industries. According to the U.S. Department of Energy, predictive maintenance is highly cost-effective, saving roughly 8% to 12% over preventive maintenance, and up to 40% over reactive maintenance. Studies have shown that predictive maintenance results in a tenfold increase in ROI. Also, it results in a 25% to 30% reduction in maintenance costs, a 70% to 75% decrease in breakdowns, and a 35% to 45% reduction in downtime. The biggest savings from predictive maintenance come due to reduced downtime. Predictive maintenance allows the maintenance frequency to be as low as possible to prevent unplanned reactive maintenance without incurring costs associated with unnecessary preventive maintenance.
Challenges
Implementing predictive maintenance in an enterprise is fraught with many challenges. It requires Big Data in real-time. Installing the requisite number of sensors on equipment and machinery is expensive. Collecting and integrating large volumes of data from different sources and processing them requires a robust IT infrastructure. Data quality is essential as incorrect and incomplete data can lead to faulty prognosis. Establishing a predictive model is a costly and time-consuming affair. It requires expertise in data analytics and machine learning and there is a shortage of skilled data scientists who can develop and maintain predictive models.
Epilogue
Despite its huge benefits, predictive maintenance is not everyone’s cup of tea. Many companies may be better off by strengthening the condition-monitoring of their key assets instead of jumping onto the predictive maintenance bandwagon. A certain level of maturity in maintenance is a precondition for implementing predictive maintenance.
Readers’ responses may be sent to:
k.sahasranaman@gmail.com or
chemindigest@gmail.com
