• Written by Muhammed Abdulla NC, CEO

A maintenance manager at a manufacturing plant must decide this week how to allocate their limited budget to minimise downtime and reduce costs, currently basing their decision on intuition and historical data. However, this approach often leads to inefficient allocation of resources. The manager needs to understand where the money is actually being spent and which cost reductions will have a lasting impact on production.

The difference between unit cost and total cost of an AI workflow

The cost of implementing an AI workflow for maintenance is often misunderstood, with many focusing on the unit cost of individual components rather than the total cost of the workflow. This narrow focus can lead to incorrect assumptions about the overall cost-effectiveness of the AI solution. In reality, the total cost of the workflow includes not only the unit cost of individual components but also the costs associated with data preparation, model training, and deployment.

A thorough understanding of the total cost is essential to making informed decisions about cost optimisation. By considering the entire workflow, maintenance managers can identify areas where costs can be reduced without compromising the effectiveness of the AI solution. This might involve optimising data processing, streamlining model training, or improving model deployment efficiency.

Caching, batching, and model routing as the three real levers

Caching, batching, and model routing are the three key levers that can be used to optimise the cost of an AI workflow for maintenance. Caching involves storing frequently accessed data in memory to reduce the need for repeated queries, while batching involves processing multiple requests together to reduce overhead. Model routing, on the other hand, involves directing incoming requests to the most suitable model based on factors such as data type and complexity.

By carefully tuning these levers, maintenance managers can significantly reduce the cost of their AI workflow without compromising its effectiveness. For example, caching can help reduce the load on the system by minimising the number of queries made to the database, while batching can help reduce overhead by processing multiple requests together.

The maintenance data this depends on

The effectiveness of an AI workflow for maintenance depends on the availability and quality of maintenance data, including work orders, asset history, spare parts inventory, and condition data where it exists. This data provides the foundation for the AI model to learn from and make predictions about future maintenance needs. Without accurate and comprehensive data, the AI model may not be able to identify areas where costs can be reduced or optimised.

A shift supervisor at the manufacturing plant, for instance, checks the maintenance log to identify patterns and trends in equipment failures, which helps inform decisions about resource allocation and maintenance scheduling. The category manager compares two suppliers of spare parts to determine which one offers the best value for money, taking into account factors such as price, quality, and delivery time.

What it changes about reactive work crowding out planned work

One of the key benefits of using an AI workflow for maintenance is that it can help reduce the incidence of reactive work crowding out planned work. Reactive work, such as emergency repairs, can be costly and disruptive, while planned work, such as routine maintenance, can help prevent equipment failures and reduce downtime. By analysing maintenance data and identifying patterns and trends, the AI model can help maintenance managers prioritise planned work and reduce the need for reactive work.

Additionally, the AI workflow can help identify instances where the same asset is recurring under different descriptions, which can help maintenance managers to identify and address underlying issues rather than just treating the symptoms. This can lead to significant cost savings and improved equipment reliability.

Measured with planned versus reactive ratio, mean time between failures, spares availability

The effectiveness of an AI workflow for maintenance can be measured using a range of key performance indicators (KPIs), including the planned versus reactive ratio, mean time between failures, and spares availability. These KPIs provide a clear picture of the impact of the AI workflow on maintenance operations and help maintenance managers to identify areas where further improvements can be made.

For example, a higher planned versus reactive ratio indicates that more maintenance work is being done proactively, rather than reactively, which can help reduce costs and improve equipment reliability. The mean time between failures provides a measure of equipment reliability, while spares availability helps to ensure that maintenance work can be completed quickly and efficiently. To learn more about how AsscherAi can help with production planning, please visit our website.

What this does not do

While an AI workflow for maintenance can be a powerful tool for cost optimisation, it is not a silver bullet. It does not replace the need for human judgment and expertise in maintenance decision-making. Most people get wrong the idea that AI can fully automate maintenance decision-making, which is wrong. AI should be seen as a tool to support and augment human decision-making, rather than replace it. If you have any questions about how AsscherAi can help with your maintenance operations, please contact us to learn more.