Stamped sheet metal parts have spring back, fenders don’t line up with hoods & doors, aircraft panels have gaps with adjacent panels, molded parts suffer shrinkage, assembled parts have misalignment, temperature variations cause warpage, and machined parts have built in stresses. Errors accumulate. Manufactured components, sub-systems, and systems deviate from design intent.
To the extent that the variability is stochastic, statistical quality control has been very successful. But modern manufacturing operations behave as complex adaptive systems – where human agents compound statistical randomness with non-linearity, self-organization, adaptation, and emergence. The solution must therefore be founded on Model-Based Cognition – an AI Architecture designed to enable plant floor operations to adapt and evolve with changing conditions.
Stamped sheet metal parts have spring back, fenders don’t line up with hoods & doors, aircraft panels have gaps with adjacent panels, molded parts suffer shrinkage, assembled parts have misalignment, temperature variations cause warpage, and machined parts have built in stresses. Errors accumulate. Manufactured components, sub-systems, and systems deviate from design intent.
To the extent that the variability is stochastic, statistical quality control has been very successful. But modern manufacturing operations behave as complex adaptive systems – where human agents compound statistical randomness with non-linearity, self-organization, adaptation, and emergence. The solution must therefore be founded on Model-Based Cognition – an AI Architecture designed to enable plant floor operations to adapt and evolve with changing conditions.
AurosIQ customers have keyed in on three well-defined targets for implementing Model-Based Quality by identifying areas that will benefit the most from an adaptive quality process:
1. Smart Work Instructions – Equally prevalent among automotive and aerospace customers in engineering, and manufacturing.
2. Model-Based Product Quality Planning – Predominantly automotive customers. Certain important advances have been adopted by aerospace customers, especially in production part approval.
3. Model-Based First Article Inspection – Predominantly aerospace customers. Several key ideas have been adopted into automotive quality processes, especially for compliance with environmental and safety regulations.
Customers picked the three targets based on some practical considerations.
The ensuing sections discuss each of the customer targets for Model-Based Quality with detailed examples. A substantive description of enabling technologies built into AurosIQ, that make Model-Based Quality possible, follows at the end.
Smart Work Instructions (SWIs)
AurosIQ encodes requirements, standards, and expert-defined processes/procedures as structured cognition models. These are then dynamically executed on the shop floor in real-time. The models validate quality as work progresses and generate intelligent work instructions – that adapt instantly to live conditions like operator inputs, equipment status, or part variations. The result: precision is built-in, not achieved through exhaustive inspection and rework.
Previously, plant floor mechanics received standard work instructions through Dassault DELMIA. They would then have to reference static manuals, maintained at multiple viewing stations across the plant floor, to look up the correct value of torque to be used during assembly. As production scaled, the supplier was faced with 18 complex specifications and 73 interrelated tables from 3 OEMs, requiring 1000s of torque lookups per day. Error rates scaled non-linearly. Further the procedure could not adapt to evolving conditions.
To solve this, the supplier has built a custom workflow integrating AurosIQ, DELMIA, and a proprietary Torque Calculator. This workflow spans engineering, planning, and shop floor operations. AurosIQ dynamically provisions current build conditions to drive over 10,000 torque instantiations executed daily. Consequently, static manuals have been replaced with real-time, smart instructions – engineers and mechanics now receive precise, contextual guidance as they work, closing quality gaps and surfacing long-buried process issues.
At a leading aerospace company, subject matter experts author cognition models that encode detailed specifications of what quality metrics to measure (position errors, gaps in assembly etc.) And where – including, in the form of large data tables – such measurements should be performed. In addition, a second set of cognition models that instruct a technician how to perform and record the measurements are also codified. As work progresses, actual measurements from the plant floor are fed into AurosIQ. The next set of work instructions reflect new insights from reported conditions on the plant floor. This leads to SWIs becoming increasingly smarter with experience.
Model-Based Product Quality Planning
In manufacturing and quality operations, AurosIQ finds wide use in planning, process validation, and production part approval – for supplier parts as well as internally manufactured parts. In addition, all AurosIQ customers who are focused on advanced product quality (APQP) have implemented lessons learned. These are used to drive updates to internal standards as well as to create and implement best practices.
The same QPPA also drives the modeling and execution of separate supplier sub-assessments (SSAs) created for each supplier. Through the Partner Portal, suppliers have access to their individual SSAs, which they use to upload documents and status reports for all gate deliverables including preventive action, containment, corrective action, and validation.
With AurosIQ the QLE drives gate deliverables, supplier quality, and production/assembly operations from a single source of modeled data. All relevant information is dynamically provisioned, in context, to the right processes and parties. Every change is automatically propagated by the QPPA to all dependent sub-assessments. Where requirements and standards are defined upstream, all downstream actions are automatically validated against authoritative specifications.
The company achieves this by first defining a set of Quality Critical Metrics (QCMs) in AurosIQ. Next, assessment controls are used to cascade QCM requirements consistently by program, across regions to the appropriate personnel. The validation checks generate quality related outputs: validation results, nonconformance management, rework/repair evidence, and corrective/preventive actions. These are then fed back to AurosIQ for issue resolution and continuous improvement. Benefits include better first-time quality and reduced warranty costs.
Model-Based First Article Inspection
In aerospace, FAI is a pre-production validation step to ensure that the first part produced using the intended tooling, materials, and processes meets all requirements, standards, and design specifications. Once the FAI is successfully completed and approved, full-scale production can commence. FAI is governed by SAE AS9102, which defines the structure and content of the First Article Inspection process and reporting. Key considerations for AurosIQ implementations include the following steps:
1. In current practice, engineering definition is expressed through text, tables, and drawings. Humans repeatedly interpret those documents to derive design requirements, manufacturing plans, inspection rules, and AS9102 reports. This introduces interpretation variability, manual translation effort, and compliance risk.
In AurosIQ, the traditional specification document is ingested and composed into cognition models. Engineering definition is encoded into unitized, executable logic governed by orchestration models. The model becomes the authoritative representation of engineering definition, and any document (including an AS9102-style report) is derived directly from that model.
2. Model-Based FAI ensures that the same parameter definitions govern all downstream contexts. Design intent is expressed in one orchestration. Inspection logic, including FAI requirements, is expressed in another orchestration that reuses the exact same parameter identities. Manufacturing planning is governed by yet another orchestration. FAI therefore does not require decoding a document. It becomes the execution of an inspection orchestration derived directly from the authoritative models.
3. AurosIQ also allows for an important distinction between requirement-level values and design-level targets. In many cases, nominal values specified in the engineering definition are more stringent than contractual requirements. Model-Based FAI explicitly represents these as separate but related constraints within the same parameter structure. FAI can then validate compliance against the appropriate constraint set without ambiguity.
4. When inspection is executed, measured values are captured directly against the governing model parameters. The result is a structured model data footprint of conformance. From that footprint, reporting artifacts required by the end customer, including AS9102 Forms 1, 2, and 3 or DoW-specific submission formats, can be generated automatically. Manual compilation of documented proof is eliminated and replaced by automated evidence gathering and reporting.
Adaptive quality improvement, smart work instruction, and delta FAI are enabled by the same model-driven process. Except, the focus shifts from the first unit to continuous inspection and continuous data collection across production units, or across units selected by an orchestration model. Specifications compliance becomes continuous.
Enabling Technologies
Model-Based Quality is powered by three key AurosIQ innovations.
For example, quality inspection procedures would be encoded in an orchestration model, with conditional steps based on evaluated parameters. When evaluated conditions call for an override of standard work instructions (quality escapements,) a flow control model gets triggered. Flow control models will then re-instantiate affected parameter values and route back to automatically archived prior save points (or last viable inspection.)
Model-Based Quality is the most cognitive of AurosIQ applications in the sense that it uses every single model type in the Model-Based Cognition arsenal.
Dynamic real-time provisioning delivers these models directly into execution as context demands it. When the relevant conditions are present, the governing logic inserts itself into the flow of work and defines, instructs, and evaluates decisions in real time. Because shared parameters connect models across domains, provisioning becomes coordinated rather than isolated, enabling logic to combine and recombine organically as complexity increases or context changes.
When complemented by top-down orchestration models, bottom-up sourcing and dynamic provisioning ensures adaptation with scalability.
Parameter threading is the mechanism by which intent, decision-making, and learning remain connected across complexity. Without it, logic fragments as scale increases. With it, intelligence operates coherently across tools, teams, and time. In addition, parameter threading is a critical element of automating the creation of audit trails, evidence of proof, and documentation. These are required for compliance with functional, safety, quality and regulatory requirements.
Insert text to encourage the exploration of other use cases using the use case wheel.
