AI Applications in AEC Industry

1import numpy as np
2import pandas as pd
3from sklearn.ensemble import RandomForestClassifier
4from sklearn.model_selection import train_test_split
5from sklearn.metrics import classification_report
6import ifcopenshell
7
8# Load IFC file
9def load_ifc_data(ifc_file_path):
10    print(f"Loading IFC file: {ifc_file_path}")
11    ifc_file = ifcopenshell.open(ifc_file_path)
12    return ifc_file
13
14# Extract features from BIM elements
15def extract_element_features(ifc_file):
16    print("Extracting features from BIM elements...")
17    
18    # Get all walls, slabs, columns, beams, windows, and doors
19    walls = ifc_file.by_type("IfcWall")
20    slabs = ifc_file.by_type("IfcSlab")
21    columns = ifc_file.by_type("IfcColumn")
22    beams = ifc_file.by_type("IfcBeam")
23    windows = ifc_file.by_type("IfcWindow")
24    doors = ifc_file.by_type("IfcDoor")
25    
26    # Combine all elements
27    all_elements = walls + slabs + columns + beams + windows + doors
28    
29    # Create feature list
30    features = []
31    labels = []
32    
33    for element in all_elements:
34        # Skip elements without properties
35        if not hasattr(element, "IsDefinedBy"):
36            continue
37            
38        # Get element properties
39        element_props = {}
40        element_props["GlobalId"] = element.GlobalId
41        element_props["Type"] = element.is_a()
42        
43        # Extract geometric properties
44        try:
45            # Get representation
46            if hasattr(element, "Representation"):
47                rep = element.Representation
48                if rep and rep.Representations:
49                    for r in rep.Representations:
50                        if r.RepresentationType == 'Brep':
51                            # Extract bounding box
52            
53                            # Get volume and surface area
54                            element_props["HasBrep"] = 1
55                        else:
56                            element_props["HasBrep"] = 0
57        except:
58            element_props["HasBrep"] = 0
59            
60        # Extract material properties if available
61        try:
62            if hasattr(element, "HasAssociations"):
63                for association in element.HasAssociations:
64                    if association.is_a('IfcRelAssociatesMaterial'):
65                        material = association.RelatingMaterial
66                        if material.is_a('IfcMaterialLayerSetUsage'):
67                            element_props["HasMaterial"] = 1
68                        else:
69                            element_props["HasMaterial"] = 0
70        except:
71            element_props["HasMaterial"] = 0
72            
73        # Add more features as needed
74        
75        # Add to feature list
76        features.append([element_props.get("HasBrep", 0), element_props.get("HasMaterial", 0)])
77        labels.append(element_props["Type"])
78    
79    return features, labels
80
81# Train a machine learning model to classify BIM elements
82def train_element_classifier(features, labels):
83    print("Training element classifier...")
84    
85    # Convert to numpy arrays
86    X = np.array(features)
87    y = np.array(labels)
88    
89    # Split data into training and testing sets
90    X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=42)
91    
92    # Train a random forest classifier
93    clf = RandomForestClassifier(n_estimators=100, random_state=42)
94    clf.fit(X_train, y_train)
95    
96    # Evaluate the classifier
97    y_pred = clf.predict(X_test)
98    print(classification_report(y_test, y_pred))
99    
100    return clf
101
102# Predict element types for new elements
103def predict_element_types(clf, features):
104    print("Predicting element types...")
105    
106    # Convert to numpy array
107    X = np.array(features)
108    
109    # Make predictions
110    predictions = clf.predict(X)
111    
112    return predictions
113
114# Main function
115def main():
116    # Load IFC file
117    ifc_file = load_ifc_data("path/to/building_model.ifc")
118    
119    # Extract features
120    features, labels = extract_element_features(ifc_file)
121    
122    # Train classifier
123    clf = train_element_classifier(features, labels)
124    
125    # Save the trained model
126    import joblib
127    joblib.dump(clf, "bim_element_classifier.joblib")
128    
129    print("Model saved as 'bim_element_classifier.joblib'")
130
131if __name__ == "__main__":
132    main()