Artificial intelligence and natural language processing capabilities applied to system engineering

What it can be the role for Artificial Intelligence (AI) and Natural Language Processing (NLP)?

• Can AI help, support and even propose design, starting from textual requirements?
• Can AI techniques be used to process natural language and transform into textual requirements?

The above questions can be asked, and extended, at any level within requirements engineering environment, facing growing challenges as systems complexity grows.

If considered, for example, a car (aka system), it shows much more complexity compared to another engineered decades ago; the number of HW systems and SW within a car has grown exponentially.

We can start to capture some challenges (not exhaustive list) engineers at all levels of a system development, face when performing requirements engineering.

For example:

• In most cases, requirements and information that supports it, is distributed and not synchronized, quickly driving to inconsistencies.
• Traceability links between different levels of requirements and sources are not always clear the reason for them and hard to maintain.
• Steep learning curve to learn good practises, standards and tools that may support requirements engineering.

It is the goal for requirements engineers to analyse and capture perfect requirements. Considerations for perfect requirements can be grouped as follows:

• Correct (free of mistakes): standards and good practises capture rules and practises to capture requirements free of mistakes.
• Well structured: requirements patterns can be used to promoted consistency in a group of requirements and facilitating the verification of requirements correctness and completeness.
• Textual requirements should be consistent with glossaries and any existing models.

INCOSE Guide for Writing Good Requirements (GfWR) captures a number of rules and good practises to capture correct requirements.

However, engineers need time to learn, get familiar with the guide and practise. Normally, this is associated to a "steep" learning curve.

Model-Based Systems Engineering (MBSE) can play a pivot role to help engineers in early phases of a system development to capture, analyse and simulate requirements in a model. MBSE promotes as values to reduce risk, tackles systems complexity and improves quality. More can be read at: What is Model-Based Systems Engineering

MBSE is Systems Engineering (SE), but supported by models. Models that can be captured using modelling languages, such as Arcadia, System Modelling Language (SysML), and others.

Arcadia implements an ontology. An ontology that captures concepts and definitions, and the relationship between concepts. Hence, it is inherent in a model, traceability between model elements.

Textual requirements can be created directly in a modelling tool (e.g., Capella) or imported into a model and created traceability links.

However, a model can be considered a great value, single traceability links between requirements and model elements it can also be considered as a shortfall, as it does not map individual requirements slot to model elements.

Any effort for creating traces between all requirements pattern slots and model elements should be handled by a tool to ease and facilitate the task.

The above challenges and considerations can be extended to what AI can do to support requirements engineering.

Some AI capabilities for model-based requirements engineering could be:

• Reduce the steep learning curve for requirements engineering and model-based.
• Ability to learn new rules, new patterns, verify consistency, correctness, and completeness of perfect requirements.
• Support engineers in decisions during the whole design process.
• Continuous monitoring and reporting of updates and changes.

AI can play a major role, in a near future, by trying to address the above and more challenges faced by “traditional” requirements engineering, but also model-based system engineering to accelerate and support engineering activities.

#ai #nlp #arcadia #capella #incose #systemsengineering #sysml #requirements #mbse #ontology