Introduction
In Simio, complex decision logic can be simplified by using neural network regression to infer information when the relationship between inputs and outputs is complicated, such as estimating order lead times. Simulation runs can not only use neural network models for estimation and inference but also automatically generate synthetic training data to monitor their predictive performance and retrain them. By using neural networks to simplify decision logic within a simulation model, the focus shifts to modeling system components and their interactions. This makes the digital twin simulation model easier to build, understand, debug, and maintain.
Neural networks are a subset of machine learning algorithms, inspired by the human brain. A neural network model is comprised of node layers: an input layer, one or more hidden layers, and an output layer. A network with two or more hidden layers is referred to as a deep learning network. A feedforward neural network is the first and simplest form of a neural network, in which data moves in only one direction from the input layer to the output layer (without loopbacks). The network has parameters called weights and biases, which are established through a process called training. These weights and biases are then used to transform the inputs at each node into an output that is sent to the next layer of nodes.
A regression neural network predicts one or more numerical outputs given a set of numerical inputs. Although neural networks are widely known for modeling complex problems such as image recognition and generative AI applications like ChatGPT, they are also well-suited for regression applications, such as predicting a KPI in a system given the current system state. This makes them an ideal framework for embedding AI in digital twin simulation models.
Simio Digital Twins & Neural Networks
Simio is the first and only discrete-event-based digital twin simulation software to offer comprehensive, embedded AI features, fully supporting the creation and automatic training of regression neural networks within a model—without requiring Python or Java programming or integration with external third-party AI tools. Simio enables the definition of one or more neural network models, which can then be referenced for inference within a Simio model through a neural network element.
Please download the PDF below to continue reading…