Making Batch Predictions with Graphs¶

In [1]:

from ml4pd import components
from ml4pd.flowsheet import Flowsheet
from ml4pd.streams import MaterialStream
from ml4pd.aspen_units import Distillation
import numpy as np

from ml4pd import components from ml4pd.flowsheet import Flowsheet from ml4pd.streams import MaterialStream from ml4pd.aspen_units import Distillation import numpy as np

Specify Molecules in System¶

The participating molecules must first be specified.

In [2]:

components.set_components(['acetone', '2-butanone'])

components.set_components(['acetone', '2-butanone'])

Build Graph¶

For larger process models, the user might want to first build out the model before feeding in data. Flowsheet can support this to a certain extend (can only handle tree-like structures with one root node at the moment).

In [3]:

feed = MaterialStream()
dist_col = Distillation()

bott, dist = dist_col(feed)

flowsheet = Flowsheet(input_streams=[feed], output_streams=[bott, dist])
flowsheet.plot_model()

feed = MaterialStream() dist_col = Distillation() bott, dist = dist_col(feed) flowsheet = Flowsheet(input_streams=[feed], output_streams=[bott, dist]) flowsheet.plot_model()

Out[3]:

Streams and UnitOp objects automatically assign an id to every initiliazed instance. Here, D0 is the id of dist_col and feed0, bott, dist2 are the ids of feed, bott and dist respectively. User can also specify the id of instances upon initiliazaiton via object_id.

In [4]:

feed_stream = MaterialStream(object_id="feed_stream")
feed_stream

feed_stream = MaterialStream(object_id="feed_stream") feed_stream

Out[4]:

MaterialStream(object_id='feed_stream', before=None, after=None, check_data=True, verbose=False, stream_type='feed', pressure_units='atm', flowrates_units='kmol/hr', temperature_units='degC')

Feed Data to Graph¶

Data can be fed to individual node via a nested dictionary passed to Flowsheet.

In [5]:

feed_data = dict(

    molecules = {
        "name_A": ['acetone'] * 5,
        "name_B": ["2-butanone"] * 5
    },
    flowrates = {
        "flowrate_A": np.linspace(0.1, 1, 5).tolist(),
        "flowrate_B": np.linspace(0.1, 1, 5).tolist()
    },
    vapor_fraction = [0] * 5,
    pressure = np.linspace(0.1, 5, 5).tolist(),
)

col_data = dict(
    no_stages = np.arange(5, 20, 3).tolist(),
    pressure = np.linspace(0.1, 5, 5).tolist(),
    reflux_ratio = np.linspace(5, 15, 5).tolist(),
    boilup_ratio = np.linspace(10, 15, 5).tolist()
)

input_data = {
    "feed0": feed_data,
    'D0': col_data
}

flowsheet.run(inputs=input_data)

feed_data = dict( molecules = { "name_A": ['acetone'] * 5, "name_B": ["2-butanone"] * 5 }, flowrates = { "flowrate_A": np.linspace(0.1, 1, 5).tolist(), "flowrate_B": np.linspace(0.1, 1, 5).tolist() }, vapor_fraction = [0] * 5, pressure = np.linspace(0.1, 5, 5).tolist(), ) col_data = dict( no_stages = np.arange(5, 20, 3).tolist(), pressure = np.linspace(0.1, 5, 5).tolist(), reflux_ratio = np.linspace(5, 15, 5).tolist(), boilup_ratio = np.linspace(10, 15, 5).tolist() ) input_data = { "feed0": feed_data, 'D0': col_data } flowsheet.run(inputs=input_data)

c:\users\hvo\desktop\ml4pd\ml4pd\aspen_unit_ops\distillation.py:160: UserWarning: feed_stage not specified, setting it equal to half no_stages.
  self._check_feed_stage()

Inspect Results¶

In [6]:

bott.flow

bott.flow

Out[6]:

	flowrate_A	flowrate_B
0	0.011202	0.071821
1	0.025705	0.263926
2	0.049175	0.486160
3	0.078453	0.699710
4	0.097979	0.900540

In [7]:

dist_col.reboiler_duty

dist_col.reboiler_duty

Out[7]:

[25122.243568573776,
 104937.15072332504,
 202456.03258323885,
 309232.84900782304,
 454447.95065304945]