generate_synthetic_df.py

from scipy.stats import skewnorm
import numpy as np

from pyspark.sql import SparkSession
spark = SparkSession.builder.getOrCreate()

def generate_cycle_randomly(unit_id, cycle, model_variant, label):
    temp = 50+skewnorm.rvs(-8, size=1).item() + np.random.normal(0, 5)
    pressure = np.random.uniform(900,1200) + np.random.normal(0, 50)
    return (unit_id, cycle, model_variant, round(temp, 2), round(pressure, 2), label)

def generate_cycles(unit_id, model_variant, init_rul):
    res = []
    rul = init_rul
    for cycle, vals in enumerate(range(rul)):
        res.append(generate_cycle_randomly(unit_id=unit_id, cycle=cycle, model_variant=model_variant, label=rul))
        rul -= 1
    return res

data = generate_cycles(unit_id=0, model_variant='A', init_rul=45) + generate_cycles(unit_id=1, model_variant='B', init_rul=40)
df = spark.createDataFrame(data, schema=["unit_id", "cycle", "model_variant", "temp", "pressure", "label"])