Predictive maintenance (PdM) has been hailed as a game-changer for maintenance operations in industrial companies, promising to reduce unscheduled downtime, optimize the costs of maintenance schedules, and prevent costly breakdowns.
While PdM technology has seen greater adoption and ongoing improvements in recent years, in many cases it is still failing to live up to expectations. For example, a survey conducted in late 2022 by Plant Services magazine found that only 22.5% of respondents who have implemented PdM programmes, found them to be either “effective” or “very effective” while 51.3% found them to be either “needing some improvement” or “not effective”. In this article we explore some possible causes of ineffective PdM programmes and recommend some steps that industrial companies can take to fully realise the potential for true predictive maintenance.
PdM Program Survey Results: Performance Comparison, 2014-2022. Source: plantservices.com
PdM technologies leverage data from various sources such as sensors, equipment logs, historians and maintenance records to build machine learning models that are designed to predict when equipment is likely to fail. Their end goal is to move industrial companies from a reactive or periodic maintenance model to a dynamic and proactive one.
However, the sensor data that predictive maintenance models are based on, can be their ‘Achille’s heel’: If the sensor data that is being fed into PdM software does not consistently reflect the true physical state of the assets, predictions made will be unreliable, and maintenance schedules will be incorrect. This can result in unnecessary maintenance work or, worse, equipment failure that could have been prevented. There are several reasons why data quality issues can occur, including:
- Data collection: Incorrect data can occur due to data acquisition errors or sensor malfunctions.
- Data completeness: If data missing or incomplete, it can skew PdM results. Missing data can occur due to data being siloed, equipment not being instrumented, a lack of maintenance records, or equipment downtime.
- Data inconsistency: If there are too many variations in the data’s format this can lead to misinterpretation of results.
- Poor data management practices: Without a clear data governance strategy, there are likely be significant flaws in the way the data is collected, stored, accessed, and managed.
To prevent PdM efforts from failing due to poor data quality, organizations can take the following steps:
- Assess Data quality: Organizations should conduct a data quality review to identify any issues that are harming PdM programmes and other digital initiatives. If this is something that you need help with, contact one of our team members who will be happy to advise you further.
- Implement data governance: Data governance and data ownership policies should be put in place to ensure data is consistently defined and collected.
- Invest in data quality management technology: APERIO is at the forefront of helping companies to improve all aspects of digital transformation by ensuring that their industrial data is accurate, complete, and consistent. APERIO’s flagship product DataWise™ performs this function by leveraging AI to detect data quality problems and anomalies at scale and in real time. It alerts operators to a wide range of data quality issues and then recommends remediation steps. This, in turn, greatly improves the effectiveness of PdM programmes, reduces false alarms, and minimizes the risk of unexpected equipment failures.
- Provide training: Employees should be trained on data management best practices to ensure they understand the impact of data quality and how to maintain it.
In conclusion, data quality is the foundation to any successful of PdM efforts. Organizations must first ensure that they have addressed the overall health of their industrial data or their PdM efforts will be doomed to fail. APERIO’s flagship product DataWise™ is designed to ensure the data health of industrial companies at scale and in real time. To learn more, schedule a DataWise™ demo or contact us.