In modern manufacturing, the main area where small and medium-sized businesses have an advantage over large manufacturers is in their flexibility to fulfill non-standard and low batch orders. These types of orders add a layer of complexity that involves determining schedule timing, costs and resources needed for each order—factors that can require special tools, custom machines or appliances.

When non-standard orders arise, these smaller or medium-sized manufacturers have traditionally engaged in risk-based estimation to assess the variables of these orders for budget planning purposes. This entails evaluating an order for three main indicators: accuracy, time effectiveness and applicability.

One of the most common risk estimation methods that leverages these indicators is the Monte Carlo simulation method.

The Monte Carlo simulation method uses intricate algorithms to model the probability of different events occurring. The inherent problem of using this system to predict risks in non-standard orders, however, lies in the subjective nature of the risks of such orders.

That is, one individual in charge decides the probability of each risk—a decision that could vary greatly depending on how they evaluate the orders and what biases and preconceived notions they hold, among other things. An example of a Monte Carlo simulation for a manufacturing project, and the factors it analyzed, is included below.

Figure 1: Example of Monte Carlo Simulation and the risks around a non-standard order

The benefits of this are numerous; for small or medium-sized enterprises, an incorrect assessment of project risks might result in a missed deadline. Missed deadlines, in turn, could lead to contract fines, lost customers and damage to a company’s reputation.

To help eliminate risk, the new system employs two different AI strategies: artificial neural networks and fuzzy logic. Artificial neural networks are computational systems structured to mimic human neuron connections. Put in use, they enable the modified Monte Carlo system to take in data through artificial neurons and continually evaluate parameters or risks against a business’s historical data.

However, because not all businesses have access to ample historical data, the hybrid IT system also employs fuzzy logic, which evaluates heuristics supplied by the business. These heuristics are the risks around a project, along with rules around each risk.

Fuzzy logic attaches values between zero and one to each heuristic, assigning a level of uncertainty to each in order to measure their degrees of truth through comparative evaluations and non-linear functions. These values are then incorporated into the modified Monte Carlo system.

While not as accurate as an artificial neural network, fuzzy logic is useful in that it does not need historical data to evaluate the probability for risks.

Figure 2: Workflow diagram of the proposed model

While it is hard to quantify the impact of this hybrid IT model for businesses due to variance in orders, risks and manufacturers, the Monte Carlo system, when modified with AI, can still have a significant impact on the processing and planning of orders for smaller or medium-sized manufacturers.

Though this modified approach is still being tested and refined, businesses will be able to quickly adopt it upon completion, as it does not require much in terms of hardware. Additionally, by substituting cost for time, organizations can modify the model to run time-based estimations for projects.

“The basic purpose of the method is to quickly and efficiently estimate the additional project costs associated with the occurrence of the foreseeable situation during project implementation,” said Grzegorz Klosowski, researcher from the Lublin University of Technology. “For managers, it means better planning, less stress, fewer unforeseen problems and greater customer satisfaction.”

Monte Carlo systems are widely used, and one modified with AI can help improve any project that requires risk analysis. It is no secret that effective data analysis is critically important, and the researchers’ modified Monte Carlo system will soon be available to provide actionable information to reduce the risks and uncertainty associated with today’s manufacturing operations.

Learn more about the Monte Carlo system in IEEE Xplore.