• Written by Azhar Aliyar, Director

A maintenance manager checks the work order system to see which assets are due for routine maintenance, but the AI system is not providing the expected answers, highlighting the need to understand what data is required for AI in maintenance. The manager is using AsscherAi, a platform that lets a large organisation query its own data in natural language and get answers in real time, built by Harns Technologies. To get the right answers, the AI system needs data from specific systems, including work orders, asset history, spare parts inventory, and condition data where it exists.

Required systems for AI in maintenance

The work order system is a critical source of data for AI in maintenance, as it provides information on the maintenance activities that have been performed on each asset. The asset history system is also essential, as it provides a record of all the maintenance activities that have been performed on each asset over time. The spare parts inventory system is necessary to ensure that the required parts are available when needed. Condition data, such as sensor readings and inspection results, is also vital where it exists, as it provides real-time information on the condition of each asset.

For example, a shift supervisor checks the maintenance log to see which assets are due for routine maintenance, and the AI system uses data from the work order system, asset history system, and spare parts inventory system to provide the answer. The AI system can also use condition data to identify potential issues before they become major problems, allowing for proactive maintenance.

Impact of missing systems on AI answers

If one of the required systems is missing, the AI system will not be able to provide accurate answers. For instance, if the spare parts inventory system is not available, the AI system will not be able to determine if the required parts are available for a maintenance activity. This can lead to delays and increased downtime, as maintenance activities may need to be rescheduled or postponed until the required parts are available.

Most people assume that the AI system can simply use the available data to provide an answer, but this is not the case. The AI system requires a complete set of data from all the required systems to provide accurate answers. If any of the systems are missing, the AI system will not be able to provide reliable answers, and this can have significant consequences for maintenance operations.

Agreed definitions for AI in maintenance

To ensure that the AI system provides accurate answers, it is essential that all stakeholders agree on the definitions of key terms, such as "maintenance activity" and "asset condition". These definitions must be clear and consistent across all systems and stakeholders to ensure that the AI system can accurately interpret the data. For example, the definition of "maintenance activity" may vary between different departments or systems, and this can lead to confusion and errors if not addressed.

A clear understanding of these definitions is critical to the success of the AI system, and it requires collaboration and agreement among all stakeholders. This can be a challenging task, especially in large organisations with multiple departments and systems, but it is essential to ensure that the AI system provides accurate and reliable answers.

Quality problems that surface later

Quality problems can surface later if the data used to train the AI system is not accurate or complete. For instance, if the data from the work order system is not up-to-date, the AI system may not be able to provide accurate answers about maintenance schedules. Similarly, if the data from the condition monitoring system is not reliable, the AI system may not be able to identify potential issues before they become major problems.

To avoid these quality problems, it is essential to ensure that the data used to train the AI system is accurate, complete, and reliable. This requires regular data quality checks and updates to ensure that the AI system is using the most accurate and up-to-date data. For more information on how to implement AI in production, visit our production page.

Limitations of AI in maintenance

AI in maintenance is not a replacement for human judgment and expertise. While the AI system can provide accurate answers and identify potential issues, it is not a substitute for the knowledge and experience of maintenance personnel. The AI system should be used to support and augment human decision-making, rather than replace it. If you have any questions about how to implement AI in maintenance, please contact us to discuss your specific needs.

Additionally, AI in maintenance may not be the right choice for all organisations or situations. For example, if an organisation has a simple maintenance schedule with few assets and minimal condition monitoring data, AI may not be necessary. In such cases, traditional maintenance scheduling methods may be sufficient, and the use of AI may not provide significant benefits.