Routing via ML: Glass Bottle Sorting Example

To integrate machine learning into the simulation, R studio was used to train the ML algorithm and write the routing formula for Simul8 to read. The data was imported into R and the model was trained using an ensemble machine learning method called random forest. We used the randomForest package in R. The model was then saved in R as shown in the script below.

#Load in the data set and packages

library(dplyr)

library(randomForest)

#Change this directory to one where you have saved GlassData.csv

Data = read.csv('C:\\Users\\YourName \\Downloads
GlassData.csv')

#Train the linear regression algorithm

DefectRF ← randomForest(formula = Defect ~ ., data = Data)

# Save the model for the function needed in Simul8 use the code below

saveRDS(DefectRF,“GetDefectRF.rds”)

For Simul8 to apply the ML algorithm for routing, the below script is needed.

# Add routing function to read though Simul8

# These need to be the same exact names as in the data set used to train the Machine Learning model

Routing = function(df){

Weight = (df[1,2])

Cracks = (df[2,2])

Bubbles = (df[3,2])

algorithm = readRDS(C:\\Users\\YourName\\Downloads\\GetDefectRF.rds')

data = data.frame(Weight, Cracks, Bubbles)

return(predict(algorithm,data))

}

#The file you will use for Simul8

saveRDS(Routing,“ DefectRFRouting.rds”)

The names from the dataset are used when training the algorithm and correspond to the Labels of Work Items in the simulation when these enter the Sorting Station Activity. The Activity then calls the algorithm which will feed back a numerical result to determine the route the Work Item follows.

To connect the scripts with the Sorting Station Activity in Simul8, click on the Sorting Station Activity and open the Routing Out dialog in the Properties tab. Choose the By ML option.

In the Algorithm file location box select Browse and locate the routing function DefectRFRouting.rds. Then add the parameters used to train the algorithm and connect them with the values in the simulation such as Label or global variables. Click OK.

The data was imported into Python and the model was trained using an ensemble machine learning method called random forest. This is loaded from the sklearn library in Python. In this example we will use Jupyter® Notebook. Copy and paste the script below into your notebook:

#First, we need to load in the packages needed for random forest regression

import pandas as pd

from sklearn.ensemble import RandomForestClassifier

from sklearn.datasets import make_classification

import pickle

#Then we lock in the file path, when saving this as a py file by using the download section remove the “” as this is only needed in Jupyter Notebook

filepath = os.path.dirname(os.path.abspath(“file”))

# Read in the data and save as a name, in this case df for dataframe was used

df = pd.read_csv(filepath+r'\GlassData.csv')

features = [' Weight', 'Cracks', 'Bubbles']

X = df[features].values

y = df['Defect']

# Then make the model based on the x and y data arrays in this case a Random forest

clf = RandomForestClassifier(max_depth=2, random_state=0)

GlassRF = clf.fit(X, y)

# Save the algorithm as a sav file which we will use in the prediction function

filename = filepath+r'\ GlassRF.sav'

pickle.dump(GlassRF,open(filename,'wb'))

Now open a new Script, copy and paste the below script into the console, and run the script. Remember to change (file) to ("file") if running in a Jupyter Notebook. For Simul8 to be able to run it, save the scripts as a prediction.py file.

import pickle

import os

def prediction(list1,list2):

filepath = os.path.dirname(os.path.abspath(file))

filename = filepath+r'\ GlassRF.sav'

loaded_model = pickle.load(open(filename,'rb'))

result = loaded_model.predict([list2])

return result[0]

Open the Glass Bottle example simulation. Click onto the Sorting Station Activity and go to the Routing Out section in the Properties tab. This will open a new dialog. In this dialog go to the Machine Learning option. Then click the Advanced Settings tab and select Python, go back to Setup. Click on Browse and find the prediction.py script file you saved in step 1.