Jul 15, 2022
No image
Completed
ML Solution for Predicting Degradation Level of Hydraulic System Components
$10,000+
4-6 months
United States
2-5
Service categories
Service Lines
Artificial Intelligence
Big Data
Cloud Consulting
IT Services
Domain focus
Manufacturing
Challenge
A large manufacturing company reached out to us to build a technology solution that could address their business challenge of repeated malfunctioning of the specific components of the hydraulic system. Specifically, the manufacturer dealt with periodic failures of the cooler, valve, pump, and hydraulic accumulator, which also brought extra maintenance costs and affected the operations of the entire facility.
A large manufacturing company reached out to us to build a technology solution that could address their business challenge of repeated malfunctioning of the specific components of the hydraulic system. Specifically, the manufacturer dealt with periodic failures of the cooler, valve, pump, and hydraulic accumulator, which also brought extra maintenance costs and affected the operations of the entire facility.
Solution
We composed the four ML models that allowed the customer to predict the degradation level of the key components with over 90% accuracy. We achieved this by processing the records from 17 IIoT sensors in the dataset and completing a data analysis to achieve better results based on the available data; preventing the problem of overfitting since we had a huge amount of features in the system, which could negatively impact the model’s ability to generalize. Also by building the four ML models using the XGBoost classifier, StratifiedKFold, and RandomForestClassifier algorithms and comparing their effectiveness to choose the best fit for the prediction models and by implementing the model performance monitoring system for identifying when its performance drops and requires retraining or tuning
We composed the four ML models that allowed the customer to predict the degradation level of the key components with over 90% accuracy. We achieved this by processing the records from 17 IIoT sensors in the dataset and completing a data analysis to achieve better results based on the available data; preventing the problem of overfitting since we had a huge amount of features in the system, which could negatively impact the model’s ability to generalize. Also by building the four ML models using the XGBoost classifier, StratifiedKFold, and RandomForestClassifier algorithms and comparing their effectiveness to choose the best fit for the prediction models and by implementing the model performance monitoring system for identifying when its performance drops and requires retraining or tuning
Results
By the end of this project, the customer was completely satisfied with the results. The ML models provided the means for predicting component failures and allowed for proper maintenance of the hydraulic system in advance and, hence, benefited the company with cost reduction, improved productivity, safety, and lower downtime. Our team continues to monitor the performance of models and is ready to retrain them on demand. Together with the management team, we also conducted staff training on how to use the ML solution properly in production and benefit from it the most.
By the end of this project, the customer was completely satisfied with the results. The ML models provided the means for predicting component failures and allowed for proper maintenance of the hydraulic system in advance and, hence, benefited the company with cost reduction, improved productivity, safety, and lower downtime. Our team continues to monitor the performance of models and is ready to retrain them on demand. Together with the management team, we also conducted staff training on how to use the ML solution properly in production and benefit from it the most.