Key Takeaways

• Use simulation as a lab method where students predict, validate, and explain system behaviour, not as a plot generator.
• Select EMT or RMS simulation based on the question and time scale, then require students to state what that model detail cannot represent.
• Keep models physics-based and transparent, and grade validation checks plus reporting quality so results stay defensible and transferable.

Students learn faster when they must predict, test, and explain results, not just watch a lecture or copy a schematic. A large meta-analysis of 225 STEM studies found active learning raised exam scores by about 6% and cut failure rates by 55%. Simulation fits that pattern when you use it as a structured lab, with checks, limits, and clear reporting. Used as a black box, it does the opposite and trains students to trust plots they cannot defend.

The most effective simulation teaching uses disciplined, physics-based models plus validation habits that students repeat until they become automatic. You’re not trying to replace hardware labs or textbook math. You’re building the missing bridge between them, so learners can reason from assumptions to waveforms, and from waveforms back to engineering choices with confidence.

“Simulation models help students link equations to power system behaviour they can test safely.”

Define what simulation models teach in power system courses

Simulation models teach cause and effect across an electrical network, not just component equations in isolation. Students learn how voltage, current, and power move through a system after a change such as a fault, a switching event, or a control action. The lesson is always conditional on assumptions, so modelling becomes a way to think clearly about limits.

Start by naming the learning target in plain language, then map it to what students must observe. If the target is “fault current depends on network impedance,” the observation is a current waveform and an impedance path, not a completed diagram. If the target is “protection needs selectivity,” the observation is timing and coordination, not a pass or fail result. That framing keeps simulation from becoming a button-click exercise.

Simulation also teaches students what not to assume. Ideal sources, perfect measurements, and lossless components produce clean plots that look correct but teach the wrong instincts. Good course design forces students to track parameter choices, initial conditions, and solver settings, then explain how those choices shape behaviour. That habit pays off later when they face messy field data and conflicting requirements.

Choose EMT and RMS simulation based on learning goals

The main difference between EMT and RMS simulation is the time detail each one keeps, and that detail decides what you can teach. EMT resolves fast electromagnetic transients and switching effects, so it suits converters, harmonics, and protection waveforms. RMS smooths fast dynamics into phasors, so it suits load flow, voltage control, and stability studies across longer time windows.

Use RMS when the lesson is system-level relationships and you need fast runs for many cases, such as parameter sweeps or contingency studies. Use EMT when the lesson depends on waveform shape, switching instants, or control interactions that vanish in a phasor model. Power systems curricula now must treat power electronics as normal grid equipment, not a special topic, since wind and solar produced 13% of global electricity in 2023. That share shows up in control behaviour and fault response, which pushes many teaching labs toward EMT at least some of the time.

Match fidelity to the question you’re asking, then make that match visible to students. When learners can say “RMS hides switching ripple, so I should not interpret this as a harmonic result,” they’ve learned something that transfers. When they cannot, they will misread a plot with total confidence, which is the failure mode to design against.

What you want students to understand	Model detail that usually fits the task
How voltage setpoints and reactive power targets affect a feeder	RMS studies with steady-state or slow control dynamics keep runs fast
Why a converter trips during a disturbance despite “normal” power flow	EMT waveform detail captures current limits, control saturation, and switching effects
How protection coordination depends on timing and measurement filtering	EMT supports relay inputs and transient behaviour that phasors can hide
How operating points shift across many contingencies	RMS lets you run many cases and compare patterns without long runtimes
What modelling assumptions change the answer the most	Either approach works if students must justify assumptions and validate outputs

Plan simulation-based labs that build skills in stages

Simulation labs work best when each lab adds one new modelling skill while keeping the rest familiar. Students need repetition in setup, checking, and reporting, then a controlled increase in complexity. That pacing reduces copy-and-paste work and makes it clear what concept is being tested. The goal is steady competence, not a single impressive capstone run.

Structure each lab around the same workflow so students build habits, then swap the technical content. A simple template keeps attention on the engineering rather than on interface details. A staged plan also makes grading more consistent because artefacts look similar across groups. Use a single lab handout format that always asks for the same five deliverables.

• A one-sentence statement of the system question being tested
• A diagram showing what is modelled and what is omitted
• A short table of key parameters students are allowed to change
• Two validation checks tied to hand calculations or known limits
• A final explanation that connects waveforms to the original question

Staging also protects learning time. Early labs should run quickly and fail predictably when something is wrong, so students can debug with logic rather than guesswork. Later labs can add larger networks, more controls, and more edge cases once students can explain why the earlier models behaved the way they did.

“The most important judgement is simple: simulation is a teaching lab only when students can explain why the model behaves as it does, and when they can show basic evidence that it is not lying.”

Build physics-based component models students can inspect and change

Students learn modelling when they can see what a component assumes, and they can change parameters without breaking the system. Physics-based components, with transparent equations and clear parameter meaning, turn a simulation into a teachable object. The model becomes a set of claims that students can test, not a sealed artefact that produces plots.

Start with parameter sets that map directly to course concepts, such as R, L, C values, transformer percent impedance, or controller gains with units. Keep names consistent across labs, and require students to state where each value came from, even if it is provided. Ask learners to identify one parameter that affects magnitude, one that affects timing, and one that affects stability, then confirm each with a sensitivity run. That keeps attention on physical meaning instead of on interface clicks.

SPS SOFTWARE supports this style of teaching through open, editable component models and workflows that can align with MATLAB/Simulink model-based design. That matters most when you want students to inspect internals, change assumptions, and defend results line by line. Tool choice still matters less than transparency and discipline, so insist on models your students can read and reason about.

Teach power system behaviour using fault and switching studies

Fault and switching studies teach system behaviour because they expose network limits quickly and visibly. Students see how impedance paths set current, how voltage sags propagate, and how protection and controls interact. These studies also force attention to initial conditions and timing, which are the first places where modelling errors show up. Done well, they convert “rules of thumb” into observable cause and effect.

A concrete lab can use a simple medium-voltage feeder with a source, a transformer, a line, a load, and one breaker. Set an initial steady operating point, apply a single line-to-ground fault at the far end, then clear it with a breaker trip after a set delay. Students compare bus voltages, fault current peak, and energy in inductive elements before and after clearing, then repeat with a different fault resistance and a different trip delay. That single scenario teaches network impedance, protection timing, and transient recovery in one controlled setup.

Keep the teaching focus on interpretation, not on the drama of the waveform. Require students to identify which elements carried the fault current and which ones limited it, using the network diagram and parameter values. Require a short explanation of what would change if the network were weaker or if the load were more inductive, without adding new cases. That approach teaches reasoning, and it keeps the lab within a manageable scope.

Assess student learning with model validation and reporting rubrics

Assessment should reward correct reasoning and validation, not just a working simulation file. A strong rubric checks if students can confirm units, sanity-check magnitudes, and explain discrepancies between expected and simulated results. That pushes learners to treat simulation outputs as hypotheses that need testing. It also reduces grading noise, since you can score the logic even when minor setup differences exist.

Validation is easiest to teach as a small set of repeatable checks. Require one check before running dynamics, such as confirming power balance at the operating point or matching a hand-calculated short-circuit estimate within a defined tolerance. Require one check after the run, such as verifying that the breaker operation produces the expected current interruption pattern and that the model returns to a plausible steady state. Make students write each check as a statement they could apply again, not as a one-off calculation.

Reporting rubrics should also enforce traceability. Students should record solver settings, timestep choices, and key model assumptions in plain language. Marks should go to clear plots with labelled axes, a short explanation of why the plot answers the original system question, and a note about one limitation of the model. That combination builds engineers who can defend results under review, not students who can only reproduce a screenshot.

Avoid common mistakes that make simulation results misleading

Misleading simulation results usually come from hidden assumptions, weak validation, and overconfident interpretation. Students will trust a clean waveform even when the model is wrong, so teaching must put friction on that impulse. The fix is procedural: force explicit assumptions, demand basic checks, and grade explanations as hard as plots. Over time, that discipline becomes part of how students think.

Watch for a few predictable failure modes. Ideal sources and missing losses can produce unrealistically stiff behaviour, so require students to justify source impedance and load models. Poor initial conditions can fake a transient that looks like a fault response, so require an operating point check before any event. Solver settings can hide oscillations or create false ones, so require students to state timestep and tolerance choices and to rerun one case with tighter settings as a confidence check.

The most important judgement is simple: simulation is a teaching lab only when students can explain why the model behaves as it does, and when they can show basic evidence that it is not lying. SPS SOFTWARE fits that mindset when you use its transparent models to keep assumptions visible and debuggable, but the habit matters more than the platform. Keep simulation disciplined, and you’ll graduate engineers who trust results for the right reasons.

Key Takeaways

• Set scope and study intent first so model fidelity, solver choices, and outputs stay consistent with the questions you need answered.
• Use strict conventions for naming, units, signal flow, and subsystem ports so large power system models stay readable and reusable across teams and labs.
• Protect repeatability with shared libraries, small test harnesses, centralized scaling, and stored initialization and solver settings, then keep quality steady with a simple review checklist.

You can keep large electrical models clear, reusable, and testable with a few consistent structure rules.

“Good organization removes the hidden work that slows teams down, like hunting for parameters, guessing signal meaning, or fixing the same wiring mistake in five places.”

It also makes results easier to trust because assumptions stay visible instead of getting buried inside deep subsystems.

Model size is not the main problem; inconsistency is. A well-structured EMT or phasor model can grow for years without becoming fragile, as long as you treat model structure like an engineering interface and not just a drawing exercise.

Set scope and study intent for large power system models

The cleanest model organization starts with a strict scope statement that defines what questions the model must answer and what it will ignore. You should lock down study type, event set, accuracy needs, and the outputs you will use to judge success. That scope then sets the right level of switching detail, control bandwidth, and network size.

Write scope in terms of test cases and measurements, not in terms of blocks you plan to draw. Identify the boundary buses, the measurement points, and the disturbance types you will apply. Keep a short list of non-goals so you do not accidentally mix studies, such as protection timing validation and converter loss estimation, inside the same baseline model.

Standardize naming, units, and signal flow conventions early

Consistent naming and units turn a complex diagram into something you can scan and verify. Signal names should tell you what the value represents, its reference frame, and its units. Port direction should stay consistent across the whole model so you do not need to read every wire to understand causality.

Write these conventions down once and apply them to every new subsystem and library block. A small amount of up-front discipline prevents confusion later when multiple people touch the same models across labs, projects, or course terms.

• Use one bus naming pattern across all voltage levels
• Add unit hints in signal names such as kV, A, pu
• Keep control signals flowing left to right across diagrams
• Reserve one colour scheme for measurement and logging paths
• Document reference directions for power, current, and torque

10 best practices for organizing electrical system models

These practices focus on readability first, then reuse and testability. Each one reduces a specific failure mode such as duplicated logic, hidden scaling, or solver changes that silently alter results. Apply them in order when you refactor an existing model, or as a checklist when you start a new one.

1. Split models by voltage level and functional purpose

Partition the model so each layer has one clear job, such as transmission, medium voltage feeders, or low voltage converter connection. Keep each partition small enough that you can validate it with focused tests. Tie partitions together through defined buses and interfaces, not ad hoc wiring. This keeps changes local when a study scope shifts.

2. Keep top diagrams shallow with clear left to right flow

Use the top level to show structure, not detail. A shallow diagram with a consistent left to right signal flow lets you understand the full system in minutes. Group blocks so the power path is obvious and the control path is separate. Push detail down into subsystems so the top does not become a wiring map.

3. Use subsystems to hide detail and expose key ports

Subsystem boundaries should match engineering boundaries, such as a converter, a feeder segment, or a protection relay function. Expose only the ports needed to connect and test that subsystem. Keep internal measurement, scaling, and filter details inside the subsystem so the interface stays stable. Treat subsystem ports like a contract you do not casually break.

4. Separate EMT switching detail from average value sections

Mixing switching models and average value models without clear boundaries makes results hard to interpret. Keep high-frequency switching detail in dedicated areas so time step and solver choices remain obvious. Place average value equivalents in separate subsystems with the same external ports where possible. This supports quick study swaps without rebuilding the diagram.

5. Put reusable components in a shared library structure

Reusable models belong in libraries, not copied across projects. Library blocks keep fixes and improvements consistent, and they reduce the risk of silent divergence between similar subsystems. Keep libraries organized by function, such as machines, converters, networks, and protection. Add short descriptions so new users choose the right block on the first try.

6. Centralize base values, per unit scaling, and unit checks

Scaling mistakes often look like control instability or network faults, so treat unit management as a first-class design task. Store base values and per-unit conversion in one place and reference them everywhere. Add simple unit checks on key signals so errors show up early. Keep conversions close to interfaces, not scattered across the diagram.

7. Use consistent parameter sets with defaults and limits

Parameter sprawl makes models fragile because small edits change behaviour in unexpected ways. Group related parameters into structured sets and keep defaults close to typical studies. Add limits and sanity checks to catch impossible values before simulation starts. Maintain a clear separation between physical parameters and tuning parameters.

8. Separate power network, controls, protection, and measurements

Separate domains so you can review and test each one without distraction. Keep the power network focused on impedances, sources, and switching, while controls and protection stay in their own areas. Route measurements through a dedicated logging layer so instrumentation does not clutter functional logic. This structure also makes it easier to compare control versions against the same network baseline.

9. Add small test harness models for each major subsystem

A test harness gives you a fast way to validate a subsystem without loading the full system model. The harness should provide boundary conditions, reference inputs, and checks for expected outputs. A simple harness might feed a converter model with a DC source, a grid Thevenin equivalent, and a step in current reference while logging DC link ripple and line current distortion. Keep harnesses versioned beside the subsystem so updates stay linked.

10. Store solver settings, initialization, and annotations with models

Solver changes can shift results even when the diagram looks identical, so settings must be treated as part of the model. Keep initialization steps close to the subsystem they apply to, and write annotations that state assumptions and limitations. Use consistent initial conditions so test cases are repeatable. Capture any required configuration so someone else can run the model without guessing.

“Subsystem boundaries should match engineering boundaries, such as a converter, a feeder segment, or a protection relay function.”

Practice	Main takeaway
1. Split models by voltage level and functional purpose	Clear partitions keep changes local and verification focused.
2. Keep top diagrams shallow with clear left to right flow	Top levels should explain structure quickly, not show wiring detail.
3. Use subsystems to hide detail and expose key ports	Stable interfaces reduce rework when internals change.
4. Separate EMT switching detail from average value sections	Clear modelling boundaries prevent hidden solver and fidelity conflicts.
5. Put reusable components in a shared library structure	Libraries prevent copied blocks from silently diverging across projects.
6. Centralize base values, per unit scaling, and unit checks	Central scaling avoids unit errors that look like system instability.
7. Use consistent parameter sets with defaults and limits	Structured parameters keep behaviour predictable and reviews faster.
8. Separate power network, controls, protection, and measurements	Domain separation makes testing and troubleshooting more direct.
9. Add small test harness models for each major subsystem	Harnesses keep subsystem validation quick and repeatable.
10. Store solver settings, initialization, and annotations with models	Repeatable runs require solver and initialization to travel with the model.

Design subsystem interfaces for reusable simulation models and labs

Reusable simulation models depend on interface discipline more than clever internal implementation. Define what each subsystem accepts and produces, then keep that interface stable across versions. Use clear port names, documented signal units, and explicit reference directions so connections stay correct even when the model is reused in another system.

Interface discipline also supports teaching and team work because students and new engineers can connect blocks without guessing intent. SPS SOFTWARE users often get the best results when subsystems behave like well-defined components, with parameter sets that travel cleanly between lab exercises and research studies. Keep optional features behind parameters, not separate ad hoc copies of the same block.

Use review checklists and model metrics to guide refactors

Refactoring works best when you review structure the same way you review protection settings or control gains. Use a short checklist that flags duplicated logic, hidden scaling, inconsistent naming, and unclear subsystem boundaries. Track a few simple metrics, such as number of duplicate blocks removed, number of interface ports simplified, and count of unit conversions pushed to boundaries.

Good model organization is visible in daily work because debugging becomes faster and test cases become easier to repeat. SPS SOFTWARE fits well when you want transparent, physics-based modelling where the structure stays readable as complexity grows. Treat organization as part of engineering quality, and the model will stay useful long after the first study is finished.

Key Takeaways

• Start with a measurable study goal, then match model detail to the specific transient or duty you must verify.
• Use EMT only when waveform timing and switching physics will change the decision, and use RMS for broad screening and longer time windows.
• Protect accuracy first with disciplined event timing, fault impedance, and boundary equivalents, then improve speed through focused network reduction and time step control.

Accurate fault and switching models will give you transient results you can trust.

Fault studies only pay off when the model matches the event you’re trying to understand, not just the one you can simulate quickly. Power interruptions are costly enough that avoidable modelling errors matter, with a Lawrence Berkeley National Laboratory study estimating about $44 billion per year in outage costs for U.S. electricity customers. That kind of impact is why disciplined fault and switching event modelling is worth the effort.

“The practical stance is simple: start with the study goal, pick the lightest model that can still answer it, and only then optimize speed.”

Breaker operations, fault impedance, and protection timing sit right on the line between “good enough” and “misleading.” Getting those details right will save you from confident looking plots that point to the wrong engineering action.

Start with the fault and switching study goals

Define the goal in terms of a measurable outcome and a pass fail check. You should know if you’re validating protection operation, checking equipment duty, or confirming ride through behaviour. Each goal implies a different time window, network detail, and output set. Clear goals stop you from overbuilding models that run slowly but answer nothing.

Lock down a minimum set of study inputs before you touch model detail. This keeps the team aligned on what must be accurate and what can be simplified. It also makes reruns and reviews much easier, since you can see what changed and why. These five items are usually enough to start well:

• Define the fault types and switching events you must represent
• Set the exact event times and required sequencing constraints
• Choose the outputs that decide pass fail for your study
• Confirm the source strength assumptions at the study boundary
• Agree on acceptable run time and acceptable error bands

Goal clarity also forces a useful question early: do you need waveform detail, or do you need system level trends. If your answer is “both,” split the work into phases, since one model rarely serves both needs well. That split is also where most simulation time savings come from, without cutting corners on the part that matters.

Choose EMT or RMS simulation based on transient detail

EMT simulation is the right choice when switching transients, harmonics, and fast control interactions matter. RMS simulation is the right choice when you mainly need phasor magnitude and angle behaviour over longer periods. The selection should follow the time scale of the phenomenon you’re studying. Picking EMT for every case will slow you down and still won’t fix poor event modelling.

EMT uses small time steps to resolve high frequency content, so it captures breaker prestrike, transformer inrush, and converter switching effects when model detail supports it. RMS assumes sinusoidal steady behaviour within each step, so it suits load flow, slower voltage recovery, and stability style studies. A common workflow uses EMT for the first tens or hundreds of milliseconds, then shifts to RMS once the fast energy exchange settles. That handoff only works if you define what “settled” means in your outputs.

Study need	EMT simulation tends to fit	RMS simulation tends to fit
Breaker or switch transient duty	Captures steep recovery voltage and current chopping effects	Misses high frequency detail that sets peak stress
Protection timing based on instantaneous quantities	Matches time domain pickup and filtering behaviour	Needs careful approximations for fast elements
Long duration voltage recovery and stability	Runs slowly and can hide trends in heavy detail	Runs fast and highlights system level trajectory
Converter and harmonic interactions	Represents switching ripple and control coupling if modelled	Often reduces converters to averaged behaviour
Study turnaround time for many contingencies	Becomes costly unless the network is reduced carefully	Supports broad screening with reasonable computation time

Tooling matters less than model transparency when you need to justify results. SPS SOFTWARE supports physics-based EMT and RMS modelling where you can inspect and edit component behaviour, which helps teams stay consistent across study types. That consistency is a practical advantage when results must survive review and reuse. It also helps you avoid hidden assumptions that only show up after you’ve spent hours on runs.

Model short circuit faults with location impedance and timing

Fault simulation in power systems starts with three choices that control most outcomes: fault type, fault impedance, and the exact time of inception and clearing. Location matters because network impedance changes with distance and topology. Timing matters because the voltage angle at inception sets the first peak. If those inputs are vague, the results will be vague too.

Most studies should prioritize single line to ground representation, since that fault class dominates many systems. Single line to ground faults are often cited as about 70% of power system faults in instructional protection material. That statistic is useful because it tells you where modelling effort will pay back first. It also supports using multiple impedance values, since “solid” and “resistive” ground faults stress different parts of the system.

Fault impedance should reflect the physical path, not just a convenient number. Arc resistance, tower footing, cable sheath return, and contact surface conditions all shift current magnitude and DC offset decay. Clearing time should be tied to the protection and breaker sequence you expect, including any intentional delay. If the study target is equipment duty, you also need to model how the network upstream is represented, since a weak Thevenin source can cut peaks sharply.

Represent breaker and switch operations with realistic contact behaviour

Breaker modelling should match the stress you’re checking, not just the logic you’re implementing. An ideal switch that toggles open and closed at a time instant will often be fine for phasor studies. EMT fault analysis needs more care, since contact parting, arc extinction, and restrike can shape the first few milliseconds. Switching event modelling becomes misleading when the breaker is treated as perfectly clean.

Start with the simplest representation that still captures the key quantities. Controlled switching needs a model that respects current zero crossing, since mechanical opening time and pole scatter affect interruption. Transformer energization studies need prestrike behaviour to get inrush right, since the effective closing angle is rarely the commanded time. Capacitor bank switching can need preinsertion elements or damping if you’re evaluating transient overvoltage.

Contact behaviour also ties directly to how you align events in the simulation. A breaker command time is not the same as contact separation time, and a trip signal is not the same as current interruption. Model event delays explicitly, keep them consistent across phases, and document them as parameters. That habit makes sensitivity checks easier when someone questions why one run looks different from another.

Handle protection logic reclosing and transient fault clearing

Protection and reclosing logic must be represented as a sequence of measurements, decisions, and actuator delays, not just a single open command. Transient faults clear only if arc extinction and deionization are plausible within the dead time. If you skip these mechanics, you can accidentally “prove” a scheme works when it depends on timing that the field will never achieve. You’ll get the most value when protection and breaker models share the same timing assumptions.

Consider an overhead 25 kV feeder with a recloser protecting a lateral. A line to ground flashover occurs at 0.12 s with 15 ohms of fault resistance, the relay asserts a trip after 25 ms of filtering, and contacts part 35 ms later with a 400 ms dead time before reclosing. The simulated voltage recovery and the second close current will look completely different if the dead time is 200 ms, or if you assume instantaneous interruption at the trip time. That single timing chain often decides if the transient fault clears cleanly or becomes a sustained event.

Accurate relay behaviour does not require modelling every internal block, but it does require matching what the relay “sees.” Filtering, phasor estimation window length, and CT saturation can all shift operate time and element security. Align those assumptions with the study goal, then check sensitivity to the timing parameters you can’t control tightly. When results hinge on a few milliseconds, the right response is usually better modelling discipline, not more optimism.

Improve simulation speed while keeping switching transients accurate

Simulation speed improves most when you reduce unnecessary bandwidth and unnecessary network detail, while keeping the event physics intact. EMT runs slow mainly because of small time steps and large state counts. You can shorten runs by focusing high fidelity only around the faulted area and the switching devices that drive the transient.

“Speed work should never start until you know which waveforms must remain trustworthy.”

Network reduction is often the safest first move. Replace distant parts of the grid with Thevenin equivalents that match short circuit strength and X to R ratio over the frequency range you care about. Keep transformers, cables, and reactors that shape transient voltage and current near the switching point. Set a time window that ends once the quantity of interest settles, since modelling an extra second at EMT resolution can waste most of your runtime.

Time step selection deserves equal care. Too large a step will smooth peaks, distort interruption, and shift protection timing. Too small a step will bury you in computation with little gain. A good practice is to run one high-fidelity baseline case, then adjust reductions and step size until key peaks and timings stay within your acceptance bands.

Validate results and avoid common fault modelling mistakes

Validation means checking that the simulation behaves like a power system, not like a plot generator. You should verify that pre-fault load flow and voltages match expectations, and that fault current levels are consistent with short circuit calculations. Energy storage elements must show physically reasonable exchange, especially during switching. If those checks fail, speed and detail choices won’t rescue the study.

Common mistakes tend to cluster around timing and boundaries. Trip time is often confused with contact separation, and close time is often confused with effective electrical closing angle. Source equivalents get reused across cases even when topology changes, which quietly shifts fault level and DC offset. Fault impedance is set to zero for convenience, then the results are used to justify protection settings that will never see that condition.

Good fault simulation power systems work is mostly disciplined repetition, not heroic modelling. You’ll get better outcomes when every case has the same event definitions, parameter naming, and validation checks, since differences then become meaningful rather than accidental. SPS SOFTWARE fits well when you need transparent models that can be inspected and controlled, since trust builds from what you can explain, not what you can run. The strongest studies finish with a simple judgment: if the result cannot be defended from inputs to waveforms, it is not ready to guide an engineering choice.