Artificial Intelligence (AI) combines computer science and robust datasets to enable problem solving. The complex interdependencies of geographically diverse and dispersed supply chains can leverage AI to optimize the inflow of critical parts, components and systems to a factory.
The brevity of this article prohibits a deep look into the AI/supply chain relationship. Therefore, an example* will be presented to illustrate the power of AI and its potential for supply chain optimization: A single part, without substitutes, critical to the function of a guidance computer, will be used for a neural network node.
The critical-part example is a single nut machined from beryllium, with an ultra-fine thread (52 TPI) cut inside a 5” (127 mm) diameter.
Deeper Visibility
The normal vetting and monitoring of a critical-part supplier incorporates all the associated norms to ensure compliance with specifications. These include supplier surveys, manufacturing readiness assessments (MRAs), material certifications, first article inspection, supplier reputation and financial viability, and can incorporate analytics derived from data mining the machinery used to make the part. Data analytics provides immediate feedback for progress, performance and health prediction inside the supplier’s factory.
AI enables a machine to have deeper visibility than a human could conceptualize, beyond traditional supplier performance monitoring, metrics and measures. Upstream contributors to the supplier’s output have been invisible (or, at best, fuzzy) and are therefore beyond the control of the part buyer.
Upstream information is critical to a factory that operates with a just-in-time (JIT) inventory control method. For example, the machine house receives a beryllium billet from a material supplier, the material supplier receives the billet from a foundry and the foundry receives the raw material to form the billet from a mining operation. Any disruption along the supply trail impacts the performance and delivery of the buyer’s critical part.
The complexity and cost of looking back into the machine house material supply chain trail is prohibitively expensive and overwhelms the supply chain management of a complex product already burdened with data and information. And what would an end user do with the information when it was received? Understanding the complete supply chain trail and its complex interdependencies is necessary for successfully applying JIT inventory control and executing smart factory enablers.
AI can provide visibility, predictability and optimization remedies to sustain supply. The interdependencies along the complete supply chain trail leading to an end part are identified and used to build the AI model. Impact elements are assessed that would delay or stop the flow along the upstream supply chain trail. For example, extreme weather events that historically slowed or stopped beryllium mining operations are identified, quantified and used to train and refine the AI model. When the model is being used, AI incorporates upcoming weather events as an input into its model to predict if mining operations will be affected.
If a disruptive weather event is imminent, data is sent to a separate algorithm (AI or human) to compute a mitigation strategy and avoid delay. A human or automated system then executes the identified mitigation plan that sustains the raw material supply and notifies the downstream beryllium consumers of mitigation actions taken to sustain their operation. If there is no impact, the mine shutdown is invisible to the downstream users. To their eyes, the supply is laminar.
Another example of the power of AI and its potential for supply chain optimization is for labor union contracts. Here, AI analysis would determine the volatility and potential for labor disruptions to raw material delivery. In our case, the foundry might have an upcoming contract renewal, with a history of adversarial relations with its union. Predictors would analyze the potential and provide mitigation solutions, such as surging foundry production. Post-contract price increases attributable to increased labor costs could also be predicted.
The simple beryllium nut example’s neural network begins to grow as we identify more potential disrupters to the supply trail. Changing environmental regulations or social disruption factors also could be added.
A neural network is a method in AI that teaches computers to process data in a way that is inspired by the human brain. It is a type of machine learning process, using interconnected nodes or neurons in a layered structure that resembles the brain. When machine learning is added to an AI model, it creates an adaptive system that computers use to learn from their mistakes and improve continuously.
With proper data engineering of the historical and incoming data, neural networks can help computers make intelligent decisions with limited human assistance. This is because they can learn and model the relationships between input and output data that are nonlinear and complex.
Complexity
Without the computing, diagnostic and learning capability of AI, the task would be untenable to expand across a landscape with a complex product. The application of AI analysis and its domain size depends on the elements that impact a product’s manufacture. Once quantified, the model is considered built, and the machine learning component of AI is activated if desired.
Machine learning begins when neural networks grow and begin to interact through the connections of the innumerable supply chain nodes using synapse connectivity. The places where neurons connect and communicate with each other are called synapses. Each neuron in the supply neural network can have thousands of synaptic connections, and these connections can be with itself, neighboring neurons or neurons in other regions of the final assembly.
For example, the beryllium nut is connected to a threaded shaft, which in turn holds a bearing in place. Each of these components has its own supply impact complexity. As the neural network grows, the original programmed input assumptions begin to play out. AI begins to grow and assign greater impact associations between nodes through synaptic connections.
When synapses are working correctly, they enable neural nodes to communicate with each other and confirm supply chain “health.” Shortly after this period of synaptic growth and neural network maturity, AI starts to remove synapses that it no longer needs. Once AI forms a synapse, it can either be strengthened or weakened, depending on how often the synapse is used and what impact AI recognizes across the entire network as critical or benign. In addition, a data scientist monitors and tests the AI to ensure the AI is not becoming biased over time.
AI offers a huge leap to control the complex interdependencies for a supply chain to sustain the laminar flow of parts, components and systems into factories that produce highly complex products. Leveraging AI minimizes (or can eliminate) disruptions to an assembly floor through neural networks operating to learn and manage supply.
* The example is used solely to provide clarity. Any relationship with actual hardware is purely accidental.
