Figure 1.
Industrial application of DTs across various lifecycle phases. DT, digital twin.
Figure 2.
Annual number of publications on DT technology from 2017 to 2025. DT, digital twin.
Figure 3.
Digital model, digital shadow DT integration levels. DT, digital twin.
Table 1.
Criteria used to classify digital model, digital shadow DT integration levels in electric drive applications.
| Criterion | Digital model | Digital shadow | DT |
|---|---|---|---|
| Physical-to-virtual data flow | No automatic physical-to-virtual flow | Yes | Yes |
| Online model update | No | Yes | Yes |
| Synchronisation assessment | No or offline only | Yes | Yes |
| Virtual-to-physical decision path | No | No | Yes |
| Typical ED example | Offline Simulink/FEM model | EKF-based monitoring or fault diagnosis | Synchronised model with supervisory action or maintenance decision |
Figure 4.
Hybrid optimisation and machine learning framework for fault diagnosis. DT, digital twin; FPA, flower pollination algorithm.
Figure 5.
Flowchart of the prognosis algorithm for RUL prediction RUL, remaining useful life.
Table 2.
Classification of DT studies in ED: Strengths and limitations.
| Group | References | Common strength | Common limitation |
|---|---|---|---|
| Online monitoring and digital shadow studies | Wang et al. (2019); Cherifi et al. (2022); Brandtstaedter et al. (2018); Ebadpour et al. (2023); Bouzid et al. (2020); Rjabtsikov et al. (2021) | Use online measurements, model updating, observers, reduced-order models, or physics-based simulation to monitor the physical system and improve consistency between measured and simulated behaviour. | Mainly support monitoring, state estimation, fault notification, or operator decision-making. Autonomous virtual-to-physical feedback is generally not demonstrated. |
| Fault diagnosis and classification studies | Adamou and Alaoui (2024); Zayed et al. (2023); Xia et al. (2021); Lopes et al. (2021) | Provide fault-diagnosis approaches using efficiency indicators, FEM-based data generation, hybrid physics-based/data-driven models, optimisation methods machine-learning classifiers. | Focused mainly on fault detection or classification. Corrective action, supervisory control, or closed-loop feedback to the physical drive is not reported. |
| Offline simulation and digital model studies | Gonzalez et al. (2020); Lopes et al. (2021); Bejaoui et al. (2021); Magadán et al. (2023) | Use high-fidelity models, FEM simulations, Simulink-generated datasets, or data-driven models to analyse system behaviour, generate fault data, or support prognostic modelling. | Mostly offline or open-loop approaches. Online synchronisation with the physical system and automatic model updating are limited or absent. |
| RUL prediction and prognostic studies | Sivalingam et al. (2018); Aivaliotis et al. (2019); Lei et al. (2016); Magadán et al. (2023); Venkatesan et al. (2019); Bejaoui et al. (2021) | Estimate degradation, health indicators, damage evolution, or RUL using physics-based, stochastic, ANN/fuzzy-logic, or deep-learning-based prognostic models. | Mainly focused on prediction and health assessment. The prognostic output is generally not connected to operational feedback, control action, or maintenance decision execution. |
| Framework and methodology-oriented studies | Sivalingam et al. (2018); Cherifi et al. (2022); Zayed et al. (2023) | Provide methodological elements for DT development, including DT frameworks, hierarchical modelling, hybrid simulation data-driven diagnostic pipelines. | Full DT implementation remains incomplete because online synchronisation, multiphysics integration, validation, or virtual-to-physical decision paths are not fully demonstrated. |
Figure 6.
Proposed DT framework for electric drive systems. DAQ, data acquisition; DT, digital twin; PLCs, programmable logic controllers.
Table 3.
Demonstrative elements of the IM DT case study.
| Framework element | Implementation in the use case |
|---|---|
| Physical system | Induction motor drive supplied by a Mitsubishi FR-D720S-014SC-EC inverter, with belt-pulley transmission, automotive alternator battery–rheostat adjustable load. |
| Measurement layer | Line voltages, line currents rotor speed were acquired using voltage dividers composed of 2.2 MΩ and 1 kΩ resistors, 0.1 Ω, 3 W, ±1% shunt resistors, TI AMC1300 isolation amplifiers an incremental encoder with a resolution of 1,000 pulses/rev. |
| DAQ and communication | The analogue voltage and current signals were sampled using an NI PCI-6251 DAQ card in combination with an NI BNC-2120 connector block, while the rotor speed was measured from the incremental encoder pulses through counter input ctr0, connected to PFI8 of the DAQ system. |
| Data processing | Time alignment, signal scaling despiking of PWM-induced spikes. |
| Offline commissioning | Electrical parameters estimated from DC, locked-rotor no-load tests; mechanical parameters estimated from coast-down tests. |
| Virtual model | Fifth-order induction motor model expressed in the stationary α -β reference frame |
| Online synchronisation | EKF-based correction of current, flux, speed load-torque states using measured signals. |
| Synchronisation assessment | Residual comparison between measured and estimated current and speed responses. |
| Application layer | Decision support for torque-boost selection based on flux magnitude, current magnitude, speed input power. |