Predictive Maintenance with Machine Learning | Data Science & Engineering Recipes
Predictive Maintenance with Machine Learning | Data Science & Engineering Recipes
Github:
https://github.com/databowlr
https://github.com/databowlr/PdM/blob/main/Predictive_Maintenance_multilabel_classification.ipynb
https://github.com/databowlr/PdM/blob/main/Predictive_Maintenance_multiclass_classification.ipynb
This recipe is about applying supervised machine learning on a predictive maintenance problem. Tree based multi-class algorithms will be applied to select a cost sensitive solution. Cost calculations are synthetic and only for training purposes.
References:
Dataset:
https://www.kaggle.com/datasets/shivamb/machine-predictive-maintenance-classification
Dataset Description:
The dataset consists of 10 000 data points stored as rows with 14 features in columns
-UID: unique identifier ranging from 1 to 10000
-productID: consisting of a letter L, M, or H for low (50% of all products), medium (30%), and high (20%) as product quality variants and a variant-specific serial number
-air temperature [K]: generated using a random walk process later normalized to a standard deviation of 2 K around 300 K
-process temperature [K]: generated using a random walk process normalized to a standard deviation of 1 K, added to the air temperature plus 10 K.
-rotational speed [rpm]: calculated from powepower of 2860 W, overlaid with a normally distributed noise
-torque [Nm]: torque values are normally distributed around 40 Nm with an σ = 10 Nm and no negative values.
-tool wear [min]: The quality variants H/M/L add 5/3/2 minutes of tool wear to the used tool in the process. and a
'machine failure' label that indicates, whether the machine has failed in this particular data point for any of the following failure modes are true.
--Target : Failure or Not
--Failure Type : Type of Failure
CONTENT OF THIS VIDEO
00:00 Intro
00:42 Maintenance types
02:25 Examples of Predictive Maintenance
02:48 Regression and Classification usage for Predictive Maintenance
03:21 Advances of PdM in Manufacturing
04:26 Condition Monitoring; Inspection vs. Sensor based
06:19 Condition Monitoring in practice
08:37 Benefits of PdM
10:48 Cost sensitive Machine Learning
13:09 Costs due to part degradation & failure
14:58 Start of Code in Colab
20:25 Multilabel vs. Multiclass Classification
22:56 Random Forest, LGBM, XGBoost & Catboost for multi class classification
31:27 Random Forest, LGBM, XGBoost & Catboost for multi label classification
34:20 Multi-Class Confusion Matrix
36:00 Summary of Multi-Label & Multi-Class Classifiers
38:00 Catboost as example for cost sensitive learning
39:50 Multi-Class vs.Multi Label cost related False Positives and False Negatives, final Model selection
Typical failure types to be detected by predictive maintenance include: abrasive and corrosion wear, rubbing, flaws, and leak detection, among others. Mechanical ultrasound, vibration analysis, wear particle testing, and thermography are some of the most commonly used techniques.
In many predictive maintenance applications there is a mismatch between maximizing F1-score and minimizing maintenance cost. This is known as cost sensitivity of misclassification mistakes.
In this recipe, cost sensitive learning is applied with creating synthetic dataset with SMOTE due to the highly imbalanced dataset.
Multiclass classification can be categorized as a single-output learning model when the output class is represented by the integer encoding. It can also be extended to a multioutput learning scenario if each output class is represented by the one-hot vector. There are many different variants of multi class classification methods like one versus one or one versus rest.
Undetected failure can result in severe machine failure and will cause costly production downtime.Severity of production downtime due to non operational machines because of undetected failure types is multiple times higher and much more critical.
Highly imbalanced data with multiple failure types is very common for predictive maintenance related classification problems. We hope this recipe will give you a solid introduction for using machine learning models to solve predictive maintenance classification problems.
Don't forget to subscribe to the channel and hit the like button
Thanks for watching!
#supervisedmachinelearning #machinelearning #predictivemaintenance #multiclass #predictivemaintenance2022 #predictivemaintenancemachinelearning
Disclaimer: We do not accept any liability for any loss or damage which is incurred from you acting or not acting as a result of watching any of our publications. You acknowledge that you use the information we provide at your own risk. Do your own research.
Copyright Notice: This video and our YouTube channel contains dialog, music and images that are property of Data Bowl Recipes. You are authorized to share the video link and channel, embed this video in your website or others.
© Data Bowl Recipes
Predictive Maintenance with Machine Learning | Data Science & Engineering Recipes
Github:
https://github.com/databowlr
https://github.com/databowlr/PdM/blob/main/Predictive_Maintenance_multilabel_classification.ipynb
https://github.com/databowlr/PdM/blob/main/Predictive_Maintenance_multiclass_classification.ipynb
This recipe is about applying supervised machine learning on a predictive maintenance problem. Tree based multi-class algorithms will be applied to select a cost sensitive solution. Cost calculations are synthetic and only for training purposes.
References:
Dataset:
https://www.kaggle.com/datasets/shivamb/machine-predictive-maintenance-classification
Dataset Description:
The dataset consists of 10 000 data points stored as rows with 14 features in columns
-UID: unique identifier ranging from 1 to 10000
-productID: consisting of a letter L, M, or H for low (50% of all products), medium (30%), and high (20%) as product quality variants and a variant-specific serial number
-air temperature [K]: generated using a random walk process later normalized to a standard deviation of 2 K around 300 K
-process temperature [K]: generated using a random walk process normalized to a standard deviation of 1 K, added to the air temperature plus 10 K.
-rotational speed [rpm]: calculated from powepower of 2860 W, overlaid with a normally distributed noise
-torque [Nm]: torque values are normally distributed around 40 Nm with an σ = 10 Nm and no negative values.
-tool wear [min]: The quality variants H/M/L add 5/3/2 minutes of tool wear to the used tool in the process. and a
‘machine failure’ label that indicates, whether the machine has failed in this particular data point for any of the following failure modes are true.
–Target : Failure or Not
–Failure Type : Type of Failure
CONTENT OF THIS VIDEO
00:00 Intro
00:42 Maintenance types
02:25 Examples of Predictive Maintenance
02:48 Regression and Classification usage for Predictive Maintenance
03:21 Advances of PdM in Manufacturing
04:26 Condition Monitoring; Inspection vs. Sensor based
06:19 Condition Monitoring in practice
08:37 Benefits of PdM
10:48 Cost sensitive Machine Learning
13:09 Costs due to part degradation & failure
14:58 Start of Code in Colab
20:25 Multilabel vs. Multiclass Classification
22:56 Random Forest, LGBM, XGBoost & Catboost for multi class classification
31:27 Random Forest, LGBM, XGBoost & Catboost for multi label classification
34:20 Multi-Class Confusion Matrix
36:00 Summary of Multi-Label & Multi-Class Classifiers
38:00 Catboost as example for cost sensitive learning
39:50 Multi-Class vs.Multi Label cost related False Positives and False Negatives, final Model selection
Typical failure types to be detected by predictive maintenance include: abrasive and corrosion wear, rubbing, flaws, and leak detection, among others. Mechanical ultrasound, vibration analysis, wear particle testing, and thermography are some of the most commonly used techniques.
In many predictive maintenance applications there is a mismatch between maximizing F1-score and minimizing maintenance cost. This is known as cost sensitivity of misclassification mistakes.
In this recipe, cost sensitive learning is applied with creating synthetic dataset with SMOTE due to the highly imbalanced dataset.
Multiclass classification can be categorized as a single-output learning model when the output class is represented by the integer encoding. It can also be extended to a multioutput learning scenario if each output class is represented by the one-hot vector. There are many different variants of multi class classification methods like one versus one or one versus rest.
Undetected failure can result in severe machine failure and will cause costly production downtime.Severity of production downtime due to non operational machines because of undetected failure types is multiple times higher and much more critical.
Highly imbalanced data with multiple failure types is very common for predictive maintenance related classification problems. We hope this recipe will give you a solid introduction for using machine learning models to solve predictive maintenance classification problems.
Don’t forget to subscribe to the channel and hit the like button
Thanks for watching!
#supervisedmachinelearning #machinelearning #predictivemaintenance #multiclass #predictivemaintenance2022 #predictivemaintenancemachinelearning
Disclaimer: We do not accept any liability for any loss or damage which is incurred from you acting or not acting as a result of watching any of our publications. You acknowledge that you use the information we provide at your own risk. Do your own research.
Copyright Notice: This video and our YouTube channel contains dialog, music and images that are property of Data Bowl Recipes. You are authorized to share the video link and channel, embed this video in your website or others.
© Data Bowl Recipes
