Let's use the same discrete-event model of the wall clock factory and set up a "Score & Rank" optimization routine. For your convenience, the manufacturing process is summarized again in the table below. The information has been stored in a list of dictionaries named processes, with one dictionary per process. The keys of this dictionary correspond to the table column headers.

Table with process names and their duration statistics, namely mean and standard deviation.

The plot_results() method used to generate the plots in this exercise has been pre-loaded and is shown below.

def plot_results(objective_func):
    # Score
    fig, axes = plt.subplots(1, len(processes), sharey=True, figsize=(10, 8))
    for p in range(len(processes)):
        sns.scatterplot(ax=axes[p], x=process_duration_all[:, p], y=objective_func, c=objective_func, cmap="turbo_r")
        axes[p].set_title(processes[p]["Name"], rotation = 20, horizontalalignment='left')
        axes[p].set_xlabel("Duration [min]", rotation = -10)
        axes[p].grid()
    axes[0].set_ylabel("Objective function score")
    plt.show()

    # Rank
    index_sort = np.argsort(objective_func)
    fig, axes = plt.subplots(1, len(processes), sharey=True, figsize=(10, 8))
    for p in range(len(processes)):
        sns.lineplot(ax=axes[p], x=np.linspace(0, NUM_SIMULATIONS, NUM_SIMULATIONS),
                     y=process_duration_all[index_sort, p], color="orchid")
        axes[p].set_title(processes[p]["Name"], rotation = 20, horizontalalignment='left')
        axes[p].set_xlabel("Score-ranked scenarios", rotation = -10)
        axes[p].grid()
    axes[0].set_ylabel("Duration [min]")
    plt.show()

The Monte-Carlo sampling loop will produce a series of possible process trajectories and score them, as shown in the figure. Monte Carlo trajectories for different process scenario.

Model modularity to optimize continuous development

Create Your Free Account