The problem
Switched reluctance motors need no rare-earth magnets, are robust and tolerate high temperatures, which makes them interesting for low-cost or rare-earth-free EV traction. Their main drawbacks are high torque ripple and acoustic noise caused by the doubly salient structure and strongly non-linear magnetics. Reducing ripple without losing efficiency or speed range is an open research problem.
Objectives
- Build a non-linear SRM model from flux-linkage data and an asymmetric half-bridge converter.
- Implement conventional current-chopping and angle-position control as a baseline.
- Implement torque-sharing functions with current or flux tracking.
- Optimise turn-on and turn-off angles and the sharing function over the speed range.
- Quantify ripple, RMS current, efficiency and speed range against the baseline.
What you need
Tools and toolboxes
- Simulink — Controller, torque-sharing functions, logging and scripted experiments.
- Simscape Electrical — Switched reluctance machine block with tabulated magnetics, power switches and diodes for the converter.
- Optimization Toolbox — fmincon or surrogate-based optimisation of firing angles and sharing-function parameters.
- Parallel Computing Toolbox — Optional: parsim to run large parameter sweeps faster. Plain sim() in a loop also works.
Software
- MATLAB R2023b or later
- Simulink
- Simscape Electrical
- Optimization Toolbox
- Optional: free FEA tool such as FEMM for magnetics data
Extra parts and data
- Published or FEA-derived flux-linkage and torque data for an SRM geometry — SRM behaviour depends on non-linear flux linkage versus current and rotor angle; a credible study needs real geometry data, for example from a published machine or a free FEA tool such as FEMM.
Key concepts
- Reluctance torque
- Torque arises from the change of co-energy with rotor angle; ignoring saturation it is T = ½·i²·dL/dθ, so it is independent of current direction.
- Asymmetric half-bridge
- Each phase has two switches and two diodes, allowing positive, zero and negative voltage to be applied for fast magnetising and demagnetising.
- Torque-sharing function
- A function of rotor angle that splits the total torque demand between the outgoing and incoming phases during commutation so their sum stays constant.
- Firing angles
- Turn-on and turn-off angles relative to the aligned position; advancing them gives current time to build up and decay at high speed but changes efficiency and ripple.
Step-by-step guide
- 1
Define the machine and targets
Choose a published geometry, for example a 3-phase 12/8 or 4-phase 8/6 SRM of a few kilowatts, and record its flux-linkage and torque tables against current and angle. Set targets labelled as examples: a DC-link voltage, base speed, maximum speed, and a goal such as halving peak-to-peak torque ripple relative to the baseline at base speed without increasing RMS current by more than a set margin.
Checkpoint: A data file with flux-linkage and torque tables and a written target specification.
- 2
Build and verify the machine model
Configure the Simscape Electrical switched reluctance machine block with the tabulated magnetics, or build a phase model from 2-D lookup tables of current against flux linkage and angle plus torque against current and angle. Verify static torque curves by locking the rotor at a set of angles and applying constant currents, and compare with the source data. Plot torque against angle at several currents to show saturation.
Checkpoint: Simulated static torque curves match the source data within a few per cent.
- 3
Build the converter
Build an asymmetric half-bridge per phase from Simscape Electrical switching devices and diodes, with a DC Voltage Source and DC-link capacitor. Add current sensors per phase. Use a fixed-step local solver with a step small enough for the hysteresis or PWM frequency, for example 1 µs, and confirm results do not change when the step is halved. Also check that applying negative voltage removes the current in the time the flux-linkage table predicts.
Checkpoint: Applying a voltage pulse to one locked phase gives current rise consistent with the flux-linkage table.
- 4
Implement the baseline control
Implement current-chopping control with hysteresis at low speed and single-pulse angle-position control at high speed, with fixed firing angles from the literature for this machine. Add an outer speed loop producing the current reference. Measure torque ripple as (Tmax − Tmin)/Tavg, RMS phase current, efficiency and maximum torque at several speeds. Keep the speed loop and current limit fixed for all later comparisons, so only the phase-current shaping changes.
Checkpoint: A baseline table of ripple, RMS current and efficiency at low, base and high speed.
- 5
Implement torque-sharing functions
Implement linear, cubic and exponential torque-sharing functions that distribute the torque reference between outgoing and incoming phases over an overlap angle. Convert each phase torque reference to a current reference by inverting the torque table, and track it with hysteresis current control. Optionally track flux linkage instead, which reduces sensitivity to inductance non-linearity. Plot the phase torque references and the actual phase torques for one stroke to show where tracking is poor.
Checkpoint: At low speed, ripple with a torque-sharing function is clearly lower than the baseline.
- 6
Find the speed limit of each method
Sweep speed and record where each torque-sharing function can no longer track the reference because the available voltage cannot change the current fast enough. Plot ripple against speed for every method. This identifies where angle optimisation must take over and is often the weak point of torque-sharing schemes. Record also the RMS current of each method at the same points, because a method that lowers ripple by drawing much more current is not necessarily better.
Checkpoint: Ripple-versus-speed curves for all methods, with the tracking limit of each marked.
- 7
Optimise firing angles and overlap
Define an objective combining ripple, RMS current and efficiency with explicit weights, and decision variables of turn-on angle, turn-off angle and overlap angle. Evaluate the objective by simulation and optimise at several speeds with fmincon or a grid-plus-surrogate approach. Store the optimal angles as lookup tables over speed and torque. Check that the optimiser is not finding solutions at the edge of the allowed angle ranges, which usually means the bounds are wrong.
Checkpoint: Optimised settings reduce ripple at every tested speed compared with fixed angles, with RMS current change recorded.
- 8
Validate over a drive profile
Run the optimised controller and the baseline over a speed–torque profile derived from a vehicle model on an urban cycle such as MIDC. Compare energy loss, mean ripple and peak currents. Check that the optimised tables generalise to operating points between those used for optimisation. Interpolate between optimised table points linearly and test a few points not used in the optimisation, so the generalisation claim is backed by data rather than assumed.
Checkpoint: Profile-level comparison of energy and ripple for baseline and optimised control.
- 9
Sensitivity and robustness
Perturb the magnetics data, for example scaling flux linkage by ±10 per cent, and introduce a current sensor offset. Re-evaluate ripple and efficiency without re-optimising. Report how much of the improvement survives, which tells the reader whether the method would hold up on a real machine. If the gain largely disappears, say so clearly; that is a useful result in its own right and tells future work where to focus effort.
Checkpoint: A robustness table showing ripple and efficiency under perturbed parameters.
- 10
Write up for publication
Document machine data, converter, baseline, torque-sharing functions, optimisation method, drive-profile results and robustness. Release models and data where licences allow. Discuss limitations such as no mechanical resonance model and no acoustic prediction, and suggest how radial-force modelling would extend the work. Make the contribution statement specific, for example the speed range over which ripple stays below a stated value, and compare it directly with the baseline and with published results for similar machines.
Checkpoint: A thesis chapter or paper draft with reproducible results and a clear statement of contribution.
Testing & validation
- Static torque curves match the source magnetics data.
- Baseline ripple, RMS current and efficiency measured over speed.
- Torque-sharing function tracking limits identified.
- Optimised angle tables compared with fixed angles.
- Drive-profile energy and ripple comparison and robustness study.
Expected results
A validated non-linear SRM drive model with baseline and torque-sharing controllers, optimised firing-angle tables, and a quantified trade-off between torque ripple, current and efficiency over a drive profile.
Research angle for your PhD
Research questions: how far can torque ripple be reduced across the full speed range without unacceptable increases in RMS current or loss, and which combination of torque-sharing function and angle optimisation is most robust to magnetics uncertainty? Metrics: ripple factor, RMS current, efficiency, profile energy and sensitivity. Baselines: fixed-angle current chopping and a published torque-sharing function. Data should come from a published machine or documented FEA so results can be checked. A method that extends low-ripple operation to higher speed, validated on a drive profile with sensitivity analysis and released models, is the kind of result suitable for a journal or major conference.
Take it further
- Add radial-force modelling to predict acoustic noise.
- Compare a model predictive torque controller with torque-sharing functions.
- Study a different pole combination with the same method.
- Add a DC-link capacitor ripple analysis for each control method.
Viva and review questions
Why is SRM torque ripple high?
Torque comes from the change of inductance with angle, which is strongly non-linear and concentrated near phase overlap, so as conduction moves between phases the summed torque dips unless currents are shaped carefully.
Why do torque-sharing functions fail at high speed?
The back-EMF rises and the time available per stroke falls, so the converter cannot change the phase current fast enough to follow the sharing profile.
Why use an asymmetric half-bridge?
It allows positive, zero and negative voltage on each phase independently, so current can be built and removed quickly, which the unipolar torque mechanism needs.
How do you make the optimisation result trustworthy?
By using real magnetics data, checking generalisation between optimisation points, and testing robustness to parameter perturbations.
Learn it at DIYguru
These DIYguru courses and programmes teach what this project needs.