How to bring Artificial Intelligence to Plan Maintenance in Industrial Plant
- About Maintenance of Equipment in a Plant
Processes and systems in an industrial plant are highly complex and comprised of variety of equipment. Equipment is prone to breakdown with time and requires maintenance, creating disruption in the process. To avoid the traditional breakdown maintenance, concept of Preventive Maintenance was introduced.
Preventive maintenance (PM) is the regular and routine maintenance of equipment and assets to keep them running and prevent any costly unplanned downtime from unexpected equipment failure. In general practice, the PM scheduling is based on personal experience or OEM recommendation. However, with the advent of newer technology, currently we have availability of huge amount of data/information related to equipment like its past performance, failures and unplanned maintenance in plant ERP system and other data acquisition system like DCS & PLC. Using Data Analytics, this data can be segregated and constructively utilized to prepare an “Intelligent framework” , so that scheduling of maintenance can be done more effectively.
This article describes about a segmented based maintenance strategy, and sample data has been considered as an example case study to demonstrate the said concept.
In the described segmented based strategy first, different segments or clusters of equipment are created by applying Hierarchical Clustering Model on data which is a combination of Equipment past maintenance data created in ERP system , Equipment performance data logged in plant asset management system and most important domain specific attributes of equipment like criticality, service nature (Gas/Liquid/Powder/Solid), make (Reputed/Local), in-use period etc. and then accordingly adopt segment specific strategy for planning & optimizing the maintenance activities.
This clustering/segmentation of equipment will help us to strategically focus on the equipment based on past logged data & other domain specific attributes and plan frequency of maintenance accordingly for better channelization of resources in spare management , managing obsolescence , enhance overall productivity and reliability, resulting in optimization of operational and capital expenditure.
2. Concept of Clustering
It is basically a type of unsupervised machine learning method . An unsupervised learning method is a method in which we draw references from datasets consisting of input data without labelled responses. Generally, it is used as a process to find meaningful structure, explanatory underlying processes, generative features, and groupings inherent in a set of examples.
Clustering is the task of dividing the population or data points into several groups such that data points in the same groups are more similar to other data points in the same group and dissimilar to the data points in other groups. It is basically a collection of objects on the basis of similarity and dissimilarity between them. Concept is explained in below mentioned figure.
3. Need for applying Clustering Concept
Clustering concept helps to analyze large un-structured datasets by dividing the data into logical groupings. It helps you to glance through the data to pull out some patterns or structures before going deeper into analyzing the data for specific findings.
In similar fashion this concept can help to divide large number of equipment mainly installed in a large industrial plant based on past maintenance data & other attributes and then optimize the task of prioritizing the maintenance of equipment.
Applications of Clustering Model
Clustering has varied applications across industries and is an effective solution to a plethora of machine learning problems.
- Customer Segmentation where customers can be segmented using demographic or income or purchasing patterns
- Social Network Analysis to understand the dynamics of individuals and groups based on their interests and access to information available.
- Arranging Genomic Data into meaningful biological structures based on a common characteristic
- City Planning Clustering ensures a commercial zone is not within an industrial zone, or a residential area is not within an industrial zone. Hashtags on social media also use clustering techniques to classify all posts with the same hashtag under one stream
4. Hierarchical Clustering
Hierarchical clustering, as the name suggests is an algorithm that builds hierarchy of clusters. This algorithm starts with all the data points assigned to a cluster of their own. Then two nearest clusters are merged into the same cluster. In the end, this algorithm terminates when there is only a single cluster left.
The results of hierarchical clustering can be shown using dendrogram. The dendrogram can be interpreted as:
At the bottom, we start with 25 data points, each assigned to separate clusters. Two closest clusters are then merged till we have just one cluster at the top. The height in the dendrogram at which two clusters are merged represents the distance between two clusters in the data space.
Steps of strategy are reading data from ERP & other equipment’s maintenance system , applying clustering models to create segments of equipment’s based on selected attributes and then deploying model on API (Streamlit used in this case) for end use.
5. Brief Layout of discussed Maintenance Strategy:
6. Steps from Data fetching to final model deployment
Step1 : Importing required libraries
Step 2:Create title of app .Ask user to upload Maintenance Notification logged in ERP and list of equipment in a plant. Equipment uploaded file also contains “criticality attribute” .Criteria for criticality has been decided depending upon domain knowledge as described below.This can be tweaked as per user or process requirement.
Step 3: Calculating frequency of Maintenance notifications of Equipment
Step 4:Merging files frequency of Maintenance notifications of Equipment and list of equipment.
Step 5:Pre-processing of Data: Scaling of data
Step 6: Applying H-Based Model. Attributes selected are criticality & frequency of notification.
Step 7:Creating interactive sidebar to ask user to select number of clusters in which equipment need to be divided.
Step 8:Creating Clusters
Step 9:Assigning cluster labels to Equipment
Step 10:Plotting formed clusters
Step 11:Displaying summary of results
7. Snippets of Web Deployment of Model
8. Conclusion
Total 2417 no of equipment has been divided into 6 nos. of Clusters selected by user. Deployment of resources for maintenance activity, up-gradation of equipment, procurement of spares can be prioritized as per order assigned to these clusters.
While assigning ranking to the clusters, equal weightage have been given to both frequency of notification & criticality factor. However, this weightage can be tweaked as per user’s requirement while preparing the model or can be done afterwards manually.
Tangible Benefits:
1. Auto clustering of Equipment based on the criticality, Failure rate, service, age effect, and subsequent scheduling of equipment Preventive Maintenance will help in improving the equipment reliability and consequently the process down time will be reduced.
2. Aid in proper segregation of warranted and unwarranted PM will further help in better allocation and utilization of resources and ultimately the Optimization of Maintenance expenditure.
3. Auto categorization of Equipment will help ensuring the all-time availability of spares for process critical equipment and reduce the over inventory of non-critical spares.
Intangible benefits:
1. Improvement in process reliability will further enhance the trust of stakeholders at large.
2. The adoption of new technology and concepts of Industry 4.0 will help us align our work practices with future innovations and up-gradations.
3. Automation in existing system helps in effective decision making and minimizes the associated inherent errors.
4. Effective scheduling of PM reduces the Equipment downtime and odd hours maintenance, helping in managing better work-life balance.
Recommendations
Below described further enhancements can be carried out to reap maximum benefits out of the strategy :
1. Insights in the form of cluster of equipment created from the model can be further optimized by adding more attributes related to equipment like service nature of equipment (Whether Gas/Liquid/Solid/Multi-phase), make (whether reputed or local) , time period since in-use , type or technology of equipment ,location of installation etc. Further Natural Language Processing (NLP) can be used to extract relevant message stored in text description of maintenance notification to create further sub clusters within a cluster.
2. Developed Interactive web app or dashboard can be integrated with ERP & plant asset management system to automatically fetch maintenance logged data and diagnostics of equipment respectively. Interactive web app will enable user to visualize allocation of equipment to the clusters and also allow user to change number of clusters to be created depending upon its domain knowledge & organizational requirements.
3. Further for equipment allotted in segments, a segment specific ML classification model can be deployed to predict failure & estimate remaining useful life (RUL).
For any suggestions/feedback you can contact me via my LinkedIn profile
Loved the article? Become a Medium member to continue learning without limits. I’ll receive a small portion of your membership fee if you use the following link, with no extra cost to you.
https://medium.com/subscribe/@rohit.malhotra67