Total Productive Maintenance is a useful approach using which you can streamline your asset maintenance activities and enable the machine operators along with the maintenance teams to be equally responsible and carry out regular checks and precautions to avoid major breakdowns or downtime of the assets in the manufacturing plant.
The concepts of TPM and how it helps in asset management are explained in detail in the following article: Drive Total Productive Maintenance with Digitalization
Though TPM provides the approach of streamlining and continuous improvement of asset maintenance activities, you can actually make the processes more automated and intelligent infusing AI and ML technologies into it. Let’s take the example of the asset maintenance process, where three key personas are involved as below:
Asset Reliability Engineer: Needs to understand the performance of the asset and take proactive decisions to maintain it minimizing the downtime
Asset Maintenance Planner: Need to plan the maintenance activities and schedule with the relevant maintenance technicians
Asset Maintenance Technician: Performs the maintenance on the machine replacing spare parts or carrying out regular maintenance activities.
The Asset Reliability Engineer needs to monitor the asset and detect any issues proactively to avoid major breakdowns and ensure the reliability of the asset. To aid the asset reliability engineer, some of the key enablers can be continuous asset condition monitoring and predictive analytics for asset health and performance. To enable these, the asset performance data from its PLC and sensors need to be collected and analyzed after contextualizing it. Depending on the asset type it may be different parameters such as vibration, temperature, power consumption, etc. The approach to collect and contextualize the asset parametric data is explained in the following article along with how modern architectural concepts such as Unified Namespace can be leveraged for the same: Leveraging Unified Namespace for Asset Management.
After collection, storage, and contextualization, the feature set or the parameters applicable for specific analysis need to be determined from the data. Based on that, the condition rules and the predictive models need to be determined to be used to fit the data and provide the relevant predictions. It also involves training the model with historical data sets which need to be collected and stored for model training. This might be an iterative process to build or select the right model or algorithm.
The predictions can be on the anomaly of the data to determine any potential issue based on the operational parameters of the assets or remaining useful life prediction to determine when the asset breakdown may happen next.
Once the insights based on the conditions and prediction start coming in, the maintenance process to create the maintenance notification and alert the asset reliability engineer is triggered automatically. The Reliability Engineer can do the root-cause analysis based on the data that are already collected for the machine performance and initiate early and preventive maintenance to avoid larger breakdown issues. While doing the root-cause analysis he can get the information related to the earlier failure and maintenance of that specific machine in a concise format as insight and recommendations. If a maintenance need is determined by the root-cause analysis the maintenance work order is created to carry out the maintenance work.
Thus using the predictive and condition-based analysis of machine performance data, the asset reliability engineer’s activities are automated and made more intelligent to take proactive action even before a breakdown issue happens to a machine.
The Maintenance Planner gets all the different work orders, scheduled maintenance plans and maintenance requests for different machines for the work area or plant he is responsible for. Based on the different work orders and the maintenance technicians available at a specific shift or day, the maintenance planner has to effectively plan the maintenance schedules to optimize the work and schedule. The planning and scheduling is a tedious process that needs to consider different factors such as technicians’ skills, availability, spare parts, etc. Optimization algorithms can be used make to an optimized plan considering the available resources. Based on the input data and the trained machine learning algorithm, the Maintenance Planner can view the maintenance schedules as proposed by the optimization model and can make changes if needed. For each of the maintenance tasks, the maintenance technicians are proposed by the algorithm and the planner can either accept the proposal or change as needed. Additionally, the maintenance planner can also view the Failure Mode and Effect Analysis (FMEA) to understand the types of failure of the specific assets and their effects to determine its criticality which can be also a factor considered for the planning.
Once the maintenance process gets triggered based on the prediction of asset health, the work order is created and assigned to the maintenance technician. The maintenance technician on getting the task based on the work order goes to the machine to carry out the maintenance activity. However, the maintenance technician may need to refer to SOPs and manuals for the equipment and processes he/she may work on. It isn’t easy to check large documents during the actual maintenance work at the field or at the equipment location. To solve this problem, a GenAI-based intelligent chatbot can be used to consume different types of documents for the machines and processes as those are available and provide concise answers for any natural-language-based questions that the maintenance technician may ask. This is enabled by a RAG-based GenAI model which can consume unstructured data as documents and tokenize them to provide relevant and contextual information based on natural language queries. The documents fed to the GenAI model may have detailed descriptions of the equipment and process, but the answers provided based on specific questions can be very specific with steps to follow as a field technician. Thus the maintenance technician can quickly get the information needed during the work and need not have to search in large documents for relevant information.
Machine Learning models for Predictive Analytics, anomaly detection and planning optimization, as well as GenAI-based natural-language chatbot can transform the asset maintenance process infusing intelligence and automation and making the process faster and better. The below diagram shows the different use cases for Intelligent TPM.
If you want to know more about how you can utilize the intelligent and AI-based functionalities for your manufacturing and maintenance operations check out Incture Cherrywork Intelligent Maintenance & Operations (IMO) solution, which provides a platform for Total Productive Maintenance integrated with your SAP S/4HANA, enabling real-time asset monitoring, condition monitoring and predictive analytics, GenAI Chatbot and easy integration with business processes and systems.
Cherrywork Industry 4.0 suite of digital applications have digital twins that enable organizations to optimize asset operations by providing real-time insights, predictive capabilities, and simulation capabilities. They help improve asset performance, reduce maintenance costs, enhance safety, and support data-driven decision-making throughout the asset lifecycle
Would you like to do the same for your organization? If yes, then reach out to us at talk2us@cherrywork.com
© 2024 Incture Technologies Pvt. Ltd. All rights reserved.
Cherrywork is a registered trademark of Incture Technologies Private Limited.
