Challenges and Limitations

The Honest Assessment

Digital threads and digital twins are powerful concepts, but they're not magic. This lesson covers the real challenges organizations face — and where the current state of the art falls short.

Data Quality

The Garbage-In Problem

A digital twin is only as good as its data. If sensor data is noisy, intermittent, or miscalibrated, the twin's predictions will be unreliable. If the system model has inconsistencies, the digital thread will propagate errors rather than catch them.

The Integration Tax

Connecting multiple data sources means dealing with different formats, different update frequencies, different quality levels, and different semantics. "Temperature" in one system might mean ambient air temperature; in another, it might mean component surface temperature. Resolving these semantic mismatches is tedious, unglamorous work — and it never ends.

Organizational Resistance

Culture Change Is Hard

Digital engineering requires engineers to work differently. Instead of owning a document, they contribute to a shared model. Instead of reviewing slides, they interrogate a model. These changes threaten existing workflows, expertise hierarchies, and comfort zones.

The "Not My Job" Problem

Who owns the digital thread? Who maintains the digital twin? In many organizations, these responsibilities fall between existing roles. Systems engineering may own the model, but manufacturing owns the as-built data, and operations owns the sensor data. Without clear ownership, the thread frays.

Training Gaps

MBSE tools have steep learning curves. SysML is not intuitive. Model repositories require new workflows. Organizations often underestimate the training investment — both in time and in ongoing support.

Technical Limitations

Model Fidelity vs. Cost

High-fidelity models are expensive to build, validate, and maintain. A detailed physics model of a jet engine takes years of development. The question is always: how much fidelity do you need? Too little and the model is useless; too much and the cost exceeds the value.

Tool Fragmentation

Despite standards like OSLC and FMI, tool interoperability remains painful. Vendors have limited incentive to make integration easy — it enables customers to switch tools. In practice, most digital threads involve custom integration code that requires ongoing maintenance.

Scalability

A digital twin of one engine is manageable. A fleet of 10,000 engines, each with its own twin running real-time inference, is a significant infrastructure challenge. Data volume, compute cost, and orchestration complexity grow faster than linearly.

Validation

How do you know the twin is right? Validating a digital twin against physical reality requires instrumented tests — which are the expensive thing the twin was supposed to reduce. There's a bootstrapping problem: you need data to build the twin, but you need the twin to reduce data collection costs.

Security and IP Concerns

Model Data Is Sensitive

A complete system model contains more intellectual property in one place than any single document. If the model is the authoritative source of truth, it's also the most valuable target for competitors or adversaries.

Supply Chain Exposure

Digital threads that extend to suppliers mean sharing model data across organizational boundaries. How much does the OEM share with Tier 1 suppliers? How do you maintain traceability across organizations with different tools and different security postures?

Where We Are Today

What Works Well

• MBSE for systems engineering in complex programs
• Digital twins for high-value physical assets with good sensor coverage
• Digital threads within single-tool ecosystems (one PLM vendor)

What's Still Maturing

• Cross-tool digital threads with full traceability
• City-scale and fleet-scale digital twins
• AI-augmented twins that learn and adapt autonomously
• Standardized data exchange that works without custom code

What's Still Research

• Fully autonomous digital engineering pipelines
• Self-validating digital twins
• Digital threads that span the full lifecycle from concept to disposal without manual intervention

The Pragmatist's Approach

Don't wait for the technology to mature fully. The organizations that will be best positioned are the ones building capability now — learning, failing, iterating — while the cost of experimentation is low. The goal isn't perfection; it's building the muscle for a model-centric future.

Assessment