Trade Studies with Model Data

The Structured Comparison

Every engineering project faces choices: which material, which architecture, which vendor, which configuration. A trade study is the structured process of evaluating alternatives against weighted criteria to make a defensible, traceable decision.

In traditional practice, trade studies are PowerPoint exercises. Alternatives listed in columns, criteria in rows, scores assigned by committee consensus in a conference room. The data behind the scores — if it exists at all — is buried in separate analyses that nobody in the room has read. The result looks rigorous (it has a matrix!) but the scores are opinions dressed as numbers.

Digital engineering changes this by connecting the trade study to the system model, where alternatives are defined as configurations, criteria are computed from model data, and scores trace to evidence rather than consensus.

Anatomy of a Model-Based Trade Study

Step 1: Define Alternatives in the Model

Each design alternative is a variant configuration in the system model. Same system architecture, different parameter values or component selections. The model captures precisely what differs between alternatives — not as a description in a slide, but as typed properties that can be queried, compared, and computed against.

In an MBSE context, this means creating variant blocks or configuration profiles within the system model. Each alternative inherits the common architecture and overrides the parameters that distinguish it. The model enforces that alternatives are structurally comparable — they satisfy the same interfaces and occupy the same architectural role.

Step 2: Define Evaluation Criteria

Criteria are the dimensions along which alternatives will be compared. In a model-based trade study, criteria map to computable properties: total mass, power consumption, thermal margin, unit cost, reliability estimate, development schedule impact.

Each criterion needs three things:

• A calculation method — how is the value determined? (parametric equation, simulation, expert estimate, supplier data)
• A data source — where does the input data come from? (system model property, simulation result, test measurement, vendor datasheet)
• A weight — how much does this criterion matter relative to others?

The calculation method and data source are engineering decisions. The weight is a stakeholder decision — it reflects program priorities, not technical truth. Separating these explicitly is essential: data tells you how alternatives perform; weights tell you what performance matters.

Step 3: Evaluate Alternatives

For each alternative, compute or collect the criterion values. In a model-based approach:

• Mass, power, and other budgets are computed from the parametric model — change the component selection and the budget updates automatically.
• Performance metrics come from simulation runs configured against each alternative's parameters.
• Cost and schedule may come from external databases or supplier quotes, linked to the model through the digital thread.
• Reliability and risk may combine model data (component counts, interface complexity) with historical data from similar systems.

The key difference from traditional trade studies: the data is not hand-entered into a matrix. It flows from the model and its connected tools. When an alternative's parameters change, the evaluation updates.

Step 4: Score and Rank

With computed values for each criterion and alternative, scoring becomes a mechanical step. Normalize the raw values to a common scale (since mass in kilograms and cost in dollars are not directly comparable), apply the weights, and compute the weighted score for each alternative.

The ranking follows from the scores. But the ranking is not the decision — it is an input to the decision. The decision-maker still needs to consider:

• How sensitive is the ranking to the weights? If changing a weight by 10% flips the top-ranked alternative, the decision is fragile.
• Are there threshold criteria? Some criteria are not trade-worthy — if an alternative fails a regulatory requirement, no amount of performance advantage saves it.
• What are the risks? The highest-scoring alternative may also carry the highest uncertainty.

Data vs. Opinion: The Critical Distinction

The value of a model-based trade study is not the matrix format — you can create a matrix in PowerPoint. The value is the data source for each cell.

Opinion-based scoring: "I think Sensor A is more reliable than Sensor B based on my experience with a similar project five years ago." This may be correct, but it is untraceable, non-repeatable, and impossible to update when new information arrives.

Data-based scoring: "Sensor A has a mean-time-between-failure (MTBF) of 80,000 hours from the manufacturer's qualification test report, tracked in the system model as a component property with a source reference." This is traceable, repeatable, and updatable.

Not every criterion can be scored with hard data. Development risk, supply chain stability, and organizational familiarity often require expert assessment. The discipline is being explicit about which scores are data-derived and which are judgment-derived — and treating them differently when interpreting results. A trade study where seven of ten criteria are data-driven and three are expert judgment is far more defensible than one where all ten are committee consensus.

Sensitivity Analysis: What Actually Drives the Decision?

A completed trade study matrix shows which alternative scores highest. Sensitivity analysis shows whether that result is robust or fragile.

Weight sensitivity: Vary each criterion weight by a defined amount (for example, plus or minus 20%) and observe whether the top-ranked alternative changes. If it does, the decision hinges on a subjective input (the weight), and the team should invest more effort in validating that weight with stakeholders.

Data sensitivity: Vary uncertain input data within its confidence interval and observe the effect on scores. If the MTBF estimate for Sensor A has a 90% confidence interval of 60,000 to 100,000 hours, does the ranking change across that range?

Criteria sensitivity: Remove one criterion at a time and re-rank. If removing a single criterion flips the result, that criterion is the real decision driver — and it deserves extra scrutiny.

Sensitivity analysis turns a trade study from "here is the answer" into "here is the answer, and here is how confident we should be in it." That second part is where engineering judgment adds the most value.

MBSE Reinforcement: The Model Feeds the Trade Study

The connection between MBSE and trade studies is direct: the system model provides the structured data that makes trade studies quantitative rather than opinion-based.

• Parametric models in the system model define the mathematical relationships between component properties and system-level criteria. Change a component selection and the parametric model recomputes mass, power, cost, and other criteria automatically.
• Architecture variants in the model ensure that alternatives are structurally comparable — they satisfy the same interfaces and can be swapped without breaking the architecture.
• Traceability relationships connect each criterion to the requirements it addresses, making it clear why each criterion matters and what happens if an alternative scores poorly on it.
• Version control preserves the state of the trade study at the time the decision was made, so future engineers can understand not just what was decided but what data supported the decision.

This is where DGTLENG 104's emphasis on traceable, structured requirements pays off. Requirements captured in the model are not just text — they are the criteria sources for trade studies, connected through typed relationships to the model elements being compared.

Expand each alternative to see how criteria values trace to specific data sources. Notice that some sources are model-computed (parametric), some are simulation-derived (CFD), some are external (supplier quotes), and some are expert judgment (development risk). A model-based trade study makes these source distinctions explicit rather than hiding them behind a single score.

Assessment