& Artificial Intelligence
Quick Release are the global leaders in Product Data Management, and now we are turbocharging our domain expertise with powerful AI to help our clients build lasting capability, accelerate and de-risk complex programmes, and improve engineering efficiency. Our approach is entirely focused on the real-world impact.
BoM Validation
by AI
Accurate validation of the Bill of Material (BoM) is required by all manufacturers of complex engineered products.
Traditionally this is conducted either manually by engineers that understand the product (e.g. BoM Audits, Build Matrices, Commodity Quantity Checks etc.) or through simply building the product (either virtually or physically) and finding the errors that result.
Both methods are highly costly in time and material waste.
This project aims to find a third way, by using the recent advances in Foundation Models to create a smart and repeatable method of performing fast and accurate AI powered BoM Validation at scale.
Sign up to keep up!
A monthly roundup of our activities on this topic, in a short digest.
- The challenges and opportunities we've been working on
- The practical steps we've taken to get there and implementation techniques
- Access to the demonstrations of completed experimentation
Project Challenges
Validation Categories
Do we have all the parts needed, structured in the right way, configured for use on the right vehicles, populated with the right data, and in a way that the digital world is representative of the intended physical one?
Approaches
Logical rules to assess the BoM, and BoM to vehicle part mapping. Are attributes consistent with each other, clustered, and where are there outliers? Are relationships between attributes sensible, and variant configuration rules suitable for the commodities? How does the current BoM compare to historic BoMs?
Interfaces and Communication
What are the points of validation, the steps involved in each, and what is any current (and future) communication back to project teams? How do you explain the AI’s logic in a way a human engages in and can trust?
Target
Outcomes
Chatbot Interface
Harvest data, deliver results and provide a query interface for engineering teams.
Logical Validation
Complex validation for 150% BoMs to check for parts & attributes against a product specification and 100% BoMs against their build specification.
Modular Rulesets
Rulesets that can be integrated from client and project past data during onboarding to build upon a central trained core.
Explainable Results
Self-validation to avoid hallucinations and a RAG (Retrieval Augmented Generation) approach to enabling citation for results.
Secure & Compliant
LLM’s robustly deployed in the cloud with data security at the forefront.