Eddy Currents & Excess Losses in CRGO Steel

Electromagnetic Thermodynamics and Microstructure of CRGO Silicon Steel

The magnetic core of a power transformer serves as the fundamental flux-coupling circuit between windings, operating under cyclic non-linear electromagnetic excitation regimes. The overall efficiency and power density of the transformer are decisively conditioned by the microstructural and crystallographic behavior of the ferromagnetic material utilized. The predominant industry standard for medium, high, and extra-high voltage (EHV/UHV) transformers is cold-rolled grain-oriented silicon steel (CRGO), optimized via thermomechanical processing to align its easy-magnetization crystallographic axes along the rolling direction.

CRGO steel exhibits a body-centered cubic (BCC, $\alpha$-Fe lattice) crystal structure, where the $\langle 100 \rangle$ crystallographic axis represents the direction of minimum magnetocrystalline anisotropy energy, whereas the $\langle 110 \rangle$ and $\langle 111 \rangle$ directions represent intermediate and hard anisotropy axes, respectively. Through a rigorous sequence of cold rolling and secondary recrystallization annealing (inhibited by manganese sulfide, $MnS$, or aluminum nitride, $AlN$, precipitates), the so-called Goss texture is developed, characterized by the $\{110\}\langle 001 \rangle$ crystallographic orientation. In this configuration, the $\{110\}$ planes lie parallel to the sheet surface and the $\langle 001 \rangle$ directions align parallel to the longitudinal rolling direction (RD).

Where $E_a$ is the magnetocrystalline anisotropy energy density ($J/m^3$), $K_1 \approx 4.8 \times 10^4 \ J/m^3$ and $K_2 \approx 1.5 \times 10^4 \ J/m^3$ are the anisotropy constants of iron at room temperature ($298 \ K$), and $\alpha_1, \alpha_2, \alpha_3$ denote the direction cosines of the magnetization vector $\mathbf{M}$ relative to the unit cell axes. The mean angular misalignment of the Goss texture relative to the rolling direction defines the quality grade of the steel: in conventional CRGO steels, the angular dispersion ranges between $5^\circ$ and $7^\circ$, whereas in high-permeability (Hi-B) steels, it is reduced to ranges between $2^\circ$ and $3^\circ$.

The addition of silicon ($Si$) at nominal concentrations of $3.0\%$ to $3.4\%$ by weight increases the electrical volume resistivity ($\rho$) from approximately $10 \times 10^{-8} \ \Omega\cdot m$ (for pure iron) to $(45 - 50) \times 10^{-8} \ \Omega\cdot m$. This resistivity increase drastically mitigates the free transport of eddy currents. However, silicon concentrations exceeding $3.5\%$ by weight induce severe mechanical brittleness due to the formation of ordered intermetallic phases $Fe_3Si$ and $FeSi$ ($B2$ and $D0_3$), rendering industrial-scale cold-forming processes non-viable. Additionally, silicon lowers the longitudinal saturation magnetostriction constant ($\lambda_{100}$), minimizing magnetic field-induced elastic deformations and the resulting acoustic noise emissions.

The magnetic domain structure (Weiss domains) in CRGO consists of main $180^\circ$ domains separated by Bloch walls, and $90^\circ$ closure domains in the vicinity of surface discontinuities and inclusions. The characteristic domain width ($2d_0$) depends on the balance between surface magnetostatic energy and domain wall elastic energy. The application of surface-tension dielectric coatings (such as phosphate and silicate coatings, commercially designated as Carlite) imposes permanent biaxial tension on the sheet ($\sigma_t \approx 4 - 10 \ MPa$), reducing the mean width of $180^\circ$ domains and suppressing secondary transverse closure domains, thereby radically lowering anomalous eddy current losses.

Classical Loss Tripartition and Spectrum: The Bertotti Model

Under cyclic magnetodynamic excitation, the volumetric energy dissipation per unit time within the core is quantified by the closed integral over the dynamic hysteresis loop cycle. The general formulation of specific core power losses ($P_t$, in $W/kg$) is governed by the statistical theory of electromagnetic losses formulated by Giorgio Bertotti, which overcomes the limitations of classic empirical Steinmetz formulations by decoupling the phenomenon into three physically distinct dissipative mechanisms:

Where $f$ denotes the fundamental excitation frequency ($Hz$), $B_p$ is the peak magnetic flux density ($T$), $\sigma$ is the electrical conductivity of the material ($S/m$), $d$ is the laminated sheet thickness ($m$), $\rho_m$ is the volumetric mass density of the steel ($kg/m^3$, typically $7650 \ kg/m^3$), $k_h$ is the quasi-static hysteresis coefficient, $\alpha$ is the Steinmetz exponent (frequently $1.6 \le \alpha \le 2.0$), and $kexc$ is the microscopic excess loss parameter.

Quasi-Static Hysteresis Losses

Static hysteresis losses ($P_h$) originate from microscopic irreversible thermodynamic dissipation processes as magnetic domain walls move through a ferromagnetic medium containing crystal lattice defects, such as dislocations, vacancies, grain boundaries, residual stresses, and non-metallic inclusions (e.g., $SiO_2$ or $Fe_3C$ precipitates). During the motion of a $180^\circ$ Bloch wall, it becomes mechanically pinned at these pinning sites.

To unpin the domain wall from the potential well, an increase in the external magnetic field is required until the local critical field is reached. Once this energy barrier is overcome, the wall advances discontinuously at high speed to the next potential well via a non-linear microscopic jump, a phenomenon known as the Barkhausen effect. The energy dissipated in this jump is irreversibly converted into phonons (heat) within the crystal lattice:

The area of the quasi-static hysteresis loop ($w_h$) depends exclusively on the metallurgical structure of the material and is independent of the time rate of change of the flux for regimes where the domain relaxation time is infinitely smaller than the excitation period. Mathematical modeling of $P_h$ under complex waveforms or non-linear saturation requires the continuous Preisach operator formulation:

Where $\hat{\gamma}_{\alpha\beta}$ is a bistable elementary hysteretic operator with switching thresholds $\alpha$ (upswitching) and $\beta$ (downswitching), and $\mu(\alpha, \beta)$ is the Preisach density distribution function specific to the tested CRGO material, experimentally identifiable through families of First-Order Reversal Curves (FORC).

Classical Eddy Current Losses (Foucault)

Classical eddy current losses ($Pcl$) derive rigorously from the macroscopic application of Maxwell's equations to a continuous, homogeneous, isotropic conductor of finite planar geometry. Consider a steel lamination of thickness $d$ along the $z$-axis (from $z = -d/2$ to $z = +d/2$), infinite width along the $y$-axis, and length along the $x$-axis. Assuming an effective magnetic permeability $\mu$, volumetric conductivity $\sigma = 1/\rho$, and a magnetic flux density vector parallel to the surface: $\mathbf{B}(t) = B_p \sin(\omega t) \hat{\mathbf{x}}$.

Assuming that the thickness $d$ is substantially smaller than the electromagnetic penetration depth (skin depth) $\delta = \sqrt{2/(\omega \sigma \mu)}$, the magnetic flux density is spatially uniform across the lamination cross-section: $B_x(z,t) \approx B_p \sin(\omega t)$. Integrating with respect to $z$ from the central neutral plane ($z = 0$, where $E_y = 0$ by symmetry):

The induced current density $\mathbf{J}(z,t) = \sigma \mathbf{E}(z,t)$ dissipates instantaneous power via Joule heating, with a local volumetric density $p(z,t) = J_y^2(z,t) / \sigma = \sigma E_y^2(z,t)$. Integrating across the total sheet thickness and time-averaging over a period $T = 2\pi/\omega$:

Dividing by the volumetric mass density $\rho_m$ yields the normalized specific loss formulation:

This derivation reveals the direct quadratic dependence on lamination thickness $d$ and fundamental frequency $f$, justifying the imperative to reduce commercial gauges from $0.35 \ mm$ (M4) down to $0.23 \ mm$ or $0.18 \ mm$ (laser-treated ultra-low loss grades).

Anomalous or Excess Losses

Historically, the direct summation of quasi-static hysteresis losses and classical eddy current losses proved significantly lower than the total losses experimentally measured via Epstein frame wattmeter methods ($P_{ total, measured } > P_h + Pcl$). This energetic discrepancy was formally termed the loss anomaly.

The microscopic model by Pry and Bean (1958) demonstrated that magnetic flux does not vary uniformly throughout the lamination continuum, but rather concentrates exclusively within magnetic domains and transfers via the localized motion of $180^\circ$ Bloch walls. The velocity of domain wall displacement $v_w(t)$ under sinusoidal excitation induces a local electric field gradient infinitely steeper in the immediate vicinity of the wall than predicted by classical continuum theory:

Where $L$ denotes the inter-wall domain spacing and $B_s$ is the saturation flux density. The micro-eddy current confined within the micro-volume adjacent to the wall generates a local magnetic braking field. Bertotti generalized this physics by formulating the statistical theory of Magnetic Objects (MO), defined as coherent aggregates of domain walls interacting with local medium heterogeneities:

Where $G = 0.1356$ is a dimensionless constant associated with eddy current field dissipation in infinite media, $S$ is the lamination cross-sectional area perpendicular to the magnetic flux direction, and $V_0$ is an intrinsic material parameter with dimensions of magnetic field ($A/m$), which quantifies the stochastic interaction between magnetic objects and the microscopic internal pinning field distribution.

CRGO Grade (AISI / EN Classification)	Nominal Thickness $d$ (mm)	Peak Induction $B$ at $800 \ A/m$ (T)	Specific Losses $P_{1.7/50}$ (W/kg)	Specific Losses $P_{1.7/60}$ (W/kg)	Stacking Factor (%)
M4 (AISI 35G155 / EN 10107)	0.35	1.82 - 1.84	1.25 - 1.35	1.64 - 1.77	$\ge 96.0$
M3 (AISI 30G130 / EN 10107)	0.30	1.84 - 1.86	1.05 - 1.15	1.38 - 1.51	$\ge 95.5$
Conventional Hi-B (27Q110)	0.27	1.90 - 1.93	0.95 - 1.05	1.24 - 1.38	$\ge 95.0$
High-Permeability Hi-B (23Q090)	0.23	1.92 - 1.95	0.80 - 0.90	1.05 - 1.18	$\ge 94.5$
Laser-Scribed Hi-B (20Q080-L)	0.20	1.93 - 1.96	0.70 - 0.78	0.92 - 1.02	$\ge 94.0$
Amorphous Metal (Fe-Si-B Metglas 2605SA1)	0.025	1.56 - 1.58	0.18 - 0.25	0.24 - 0.33	$\ge 86.0$

Non-Linear Effects, Harmonic Distortion, and Direct Current Bias (DC Bias)

In contemporary power systems, the widespread proliferation of power electronics (large-scale solar PV inverters, HVDC links, FACTS devices, and variable frequency drives) introduces non-sinusoidal magnetization regimes and voltages with high total harmonic distortion ($THD_v$). The non-linear response of the dynamic $B(H)$ magnetization curve converts these excitations into severe anomalous power losses.

Harmonic Excitation and Minor Hysteresis Loops

When the applied voltage waveform contains higher-order harmonics ($h = 3, 5, 7, \dots$), the time derivative of magnetic flux density $\partial \mathbf{B}/\partial t$ exhibits multiple zero-crossings during a fundamental cycle. Physically, this forces transient reversal of domain wall movement prior to completing full macroscopic polarization, generating minor hysteresis loops nested within the main saturation loop.

Under pure sinusoidal voltage, $B_h \propto V_h / (h f_1)$. If the voltage contains harmonic distortion with components $V_h$, the induced harmonic field is approximated by $B_{p,h} \approx \frac{V_h}{h \, V_1} B_{p,1}$. Substituting into the classical eddy current formulation:

However, excess losses do not follow this linear decoupling. The interaction of multiple frequencies accelerates the local domain wall velocity according to a time integral of the flux derivative:

Direct Current Offset Magnetization (DC Bias)

The injection of direct current into transformer phases—caused by geomagnetically induced currents (GICs), direct-coupled inverter topologies, or asymmetrical monopolar faults in HVDC lines with ground return—shifts the dynamic hysteresis cycle along the magnetizing field axis ($Hdc = N Idc / lfe$).

This asymmetrical shift forces the core into the saturation knee during an entire half-cycle (half-cycle saturation). In the saturation region ($B > 1.9 \ T$ in CRGO), the differential permeability $\mu_d = dB/dH$ collapses drastically toward values approaching vacuum permeability ($\mu_0 = 4\pi \times 10^{-7} \ H/m$). This induces massive asymmetrical magnetizing current peaks ($i_{ mag }(t)$ with crest factors exceeding 10), transforming the magnetic field from a vector confined within the laminated core into a large-magnitude three-dimensional stray flux that links adjacent structural components (clamping plates, transformer tanks, tie-rods, and magnetic shields).

Reactive power consumption $Q$ surges by orders of magnitude, causing severe terminal voltage drops, low-order even and odd harmonics (notably $h = 2$ and $h = 3$), and catastrophic thermal heating from induced eddy currents in structural metallic parts.

Forensic Analysis of Core-Induced Thermal Failures and Dielectric Degradation

Electromagnetic pathologies and thermal degradation originating in the core represent a critical failure mode in large power transformers. Unlike mechanical winding failures induced by radial and axial short-circuit forces, core failures typically manifest as progressive electrochemical and thermal degradation processes over extended operating periods.

Interlaminar Insulation Degradation (Carlite-Type Coating)

The superficial interlaminar insulation of magnetic laminations (thickness $\sim 1.5 - 3.0 \ \mu m$, composed of a forsterite $Mg_2SiO_4$ base with an aluminum phosphate $AlPO_4$ and phosphoric acid topcoat) is designed to withstand interlaminar voltages of a few volts ($1.5 \ V$ to $5.0 \ V$ RMS per IEC 60404-6). However, under continuous local thermal stress ($T > 150^\circ C$), magnetostrictive vibrational fatigue ($100/120 \ Hz$ and acoustic harmonics), or excessive mechanical stress from over-clamping during yoke assembly, the dielectric coating micro-fractures and mechanically pulverizes.

The loss of galvanic isolation between adjacent laminations induces micro-arcing or direct galvanic contact welds. If two or more contact points occur within the same lamination stack, a high-cross-section short-circuited loop is established, linking the main core magnetic flux. The induced electromotive force within the closed loop is given by:

Since the ohmic loop resistance $Rloop$ is on the order of milliohmios ($m\Omega$), the fault current $Ifault$ reaches steady-state amplitudes of hundreds to thousands of amperes, dissipating localized Joule heat that creates hot-spots with temperatures rapidly exceeding $500^\circ C - 800^\circ C$.

Dielectric Oil Degradation and DGA Forensic Analysis

Extreme heat transfer from the core hot-spot to the surrounding dielectric fluid (naphthenic, paraffinic mineral oil, or synthetic esters) induces thermal cracking and catalytic cleavage of hydrocarbon molecular chains ($C_n H_{2n+2}$). Carbon-hydrogen covalent bonds ($C-H$, bond energy $\sim 413 \ kJ/mol$) and carbon-carbon bonds ($C-C$, $\sim 348 \ kJ/mol$; $C=C$, $\sim 614 \ kJ/mol$; $C \equiv C$, $\sim 839 \ kJ/mol$) break depending on the contact thermal gradient.

In accordance with international standards IEC 60599 and IEEE C57.104, the dissolved gas analysis (DGA) profile in oil allows precise thermodynamic identification of the core fault temperature range:

Fault Thermal Range	Dominant Combustible Gas	Diagnostic Ratios (IEC 60599 / Duval)	Core Physical Mechanism	Critical Action Required
Thermal Fault $T < 300^\circ C$ (T1)	Methane ($CH₄$) / Ethane ($C₂H₆$)	$\frac{C₂H₂}{C₂H₄} < 0.1$, $0.1 < \frac{CH₄}{H₂} < 1$, $\frac{C₂H₄}{C₂H₆} < 1$	Diffuse heating of yokes, initial degradation of interlaminar varnishes or locking paint.	Monthly periodic chromatographic monitoring; network harmonic verification.
Thermal Fault $300^\circ C < T < 700^\circ C$ (T2)	Ethylene ($C₂H₄$)	$\frac{C₂H₂}{C₂H₄} < 0.1$, $\frac{CH₄}{H₂} > 1$, $1 < \frac{C₂H₄}{C₂H₆} < 4$	Interlaminar shorted turns across multiple CRGO steel laminations. Severe inorganic coating degradation.	Load reduction to 70%; perform no-load loss test and core insulation resistance measurement.
Thermal Fault $T > 700^\circ C$ (T3)	Ethylene ($C₂H₄$) with traces of Acetylene ($C₂H₂$)	$\frac{C₂H₂}{C₂H₄} < 0.15$, $\frac{CH₄}{H₂} > 1$, $\frac{C₂H₄}{C₂H₆} > 4$	Localized fusion of the magnetic stack (core burning); short circuit between laminations and clamping frame.	Immediate forced outage. Internal endoscopic inspection and core rebuild.
Partial Discharges / Core Sparking	Dominant Hydrogen ($H₂$) with $CH₄$	$\frac{CH₄}{H₂} < 0.1$, $H₂$ exceeds baseline limits (> 100 ppm)	Galvanic floating of the core (loss of single physical ground), causing capacitive discharges to the tank.	Verification of the core and neutral external grounding circuit using a $2.5 \ kV$ megohmmeter.

Multiple Grounding Faults (Ground Loop Circulating Currents)

By normative design, a transformer core must be galvanically grounded at a single physical point via a copper strip connected to a dedicated bushing on the tank cover. This unified grounding prevents static potential elevation induced by capacitive coupling with high-voltage windings ($V_{ ind } = C_{w-c}/(C_{w-c} + C_{c-g}) \cdot V_{ winding }$).

If an inadvertent second ground connection forms at the opposite end of the core due to tie-rod insulation degradation, metallic sludge bridging at the bottom of the tank, or assembly errors, a closed physical loop of large area is established that links the main magnetic flux:

This voltage induces a continuous circulating current through the tank structure and core ground straps, typically ranging from $10 \ A$ to over $150 \ A$. This burns through tie-rod insulation, carbonizes adjacent oil, and generates alarming concentrations of ethylene ($C₂H₄$) and methane ($CH₄$) without triggering overcurrent or differential ($87T$) protective relays, which remain blind to internal core loop anomalies.

Advanced Mitigation Techniques in Design and Manufacturing

Optimizing the electromagnetic performance of magnetic cores demands a multidisciplinary framework combining crystallographic material modifications, domain micro-engineering via coherent radiation, and three-dimensional topological optimization of yoke joints.

$45^\circ$ Step-Lap Yoke Joints

The geometric assembly of limbs (legs) and yokes in stacked cores introduces discontinuities into the magnetic circuit. In traditional $90^\circ$ simple butt-lap joints, magnetic flux is forced to cross between laminations perpendicular to the rolling direction (along the hard crystallographic direction $\langle 110 \rangle$), causing severe flux line refraction, a drastic increase in interfacial magnetic reluctance, and localized saturation at cut tips.

Modern cutting and stacking technology employs $45^\circ$ mitered joints with progressive lamination staggering designated as Step-Lap (typically 5 to 7 steps per cycle). Staggering distributes the longitudinal cut gap along an axial gradient:

By staggering the joints by a spatial offset $\Delta x \approx 3.5 - 7.0 \ mm$ between adjacent laminations, magnetic flux is not forced to jump perpendicularly across a single high-reluctance gap into the adjacent lamination. Instead, it distributes progressively across a significantly larger effective geometric cross-section:

Implementing Step-Lap design reduces total joint losses in yokes by up to $25\% - 30\%$, mitigates magnetizing reactive power by $40\%$, and lowers transformer acoustic sound emissions by $3 to 6 \ dB(A)$ compared to conventional lap assemblies.

Magnetic Domain Refinement via Laser Scribing

In high-permeability CRGO steels (Hi-B), the crystallographic grain size is large ($\sim 10 - 30 \ mm$), which promotes near-perfect Goss orientation ($\le 3^\circ$). However, it proves counterproductive for dynamic loss behavior because the spacing between $180^\circ$ Bloch walls ($2d_0$) expands proportionally with the square root of grain size:

A large domain width drastically increases the instantaneous wall velocity $v_w(t)$, causing anomalous eddy current losses ($Pexc$) to surge.

The laser domain refinement technique (Domain Refinement or Laser Scribing) projects a focused continuous laser beam (Nd:YAG or high-power fiber laser) transverse to the rolling direction at regular spatial intervals ($p \approx 4 - 8 \ mm$). The thermal energy of the laser pulse induces ultra-fast, localized heating without material ablation, creating a micro-field of permanent subsurface compressive elastic stresses due to the thermal cooling gradient:

These periodic internal stress bands act as artificial magnetoelastic energy barriers that subdivide wide $180^\circ$ magnetic domains, shrinking their mean spacing $2d_0$ to one-half or one-third of their original dimension. As a direct result:

It is imperative to distinguish between non-heat-resistant laser treatment (which loses its refinement effect after stress-relief annealing at $T > 500^\circ C$ required for wound cores) and heat-resistant methods like chemical etching or plasma-controlled plastic deformation, engineered to maintain micro-domain pinning through $800^\circ C$ thermal treatments.

Computational Modeling and Implementation with Vexten Suite

Rigorous quantification of core losses and their multiphysics coupling with the electrical grid is operationally resolved in Vexten Suite by integrating non-linear harmonic power flow modules (per IEEE 519 / IEEE C57.110), coupled dynamic thermal modeling (IEC 60076-7), and short-circuit / electromagnetic transient simulation (IEC 60909 / IEEE 141).

Harmonic Distortion Correction Algorithm and K-Factor

Under severe harmonic spectra originating from non-linear loads or renewable generators, the Vexten PowerFlow & Harmonics module executes dynamic loss partitioning by recalculating the core loss factor according to the normalized spectral coefficient tensor:

The software links this factor to transformer capacity derating by calculating the K-Factor and Eddy Current Harmonic Loss Factor ($FHL$), adjusting maximum allowable current to ensure winding and core hot-spot temperatures do not exceed thermal insulation limits:

Multiphysics Modeling Workflow in Vexten Suite

1. Definition of Magnetic Excitation Spectrum: In the power quality module, import field measurements or standard IEEE voltage harmonic spectra ($V_h$) alongside residual DC components ($Idc$) resulting from power flow coupled with renewable generation.
2. Assignment of Magnetic Material in Database: Select the specific CRGO steel grade (e.g., M3, 23Q090, or 20Q080-L with laser scribing) within machine properties. Vexten Suite automatically loads calibrated microscopic parameters: conductivity $\sigma$, hysteresis coefficient $k_h$, thickness $d$, density $\rho_m$, and Bertotti parameter $V_0$.
3. Joint Topology Configuration: Define yoke joint architecture within design settings: $90^\circ$ Butt-Lap, 5-step Step-Lap, or 7-step Step-Lap. The software automatically applies the corresponding interfacial reluctance dispersion factor ($\mathcal{K}_{joint}$).
4. Execution of Coupled Transient Thermal Analysis: The solver maps total volumetric electromagnetic losses $P_{ total }(x,y,z)$ as heat sources within the thermal network solver per IEC 60076-7, resolving 3D temperature profiles across upper yokes, lower yokes, and core limbs.
5. Validation against Limits and Preventive Forensic Diagnostics: The system benchmarks local temperature gradients and loss densities against IEEE C57.104 limits. If core temperatures exceed $130^\circ C$ continuous, automated power derating alerts are triggered, and estimated fault gas generation rates ($C₂H₄$, $CH₄$, $H₂$) are calculated to support high-precision Condition-Based Maintenance (CBM) planning.

Through this comprehensive computational architecture, design, commissioning, and reliability engineers can precisely predict thermal aging of the magnetic core stack, eliminate catastrophic failure risks from interlaminar circulating currents, and optimize the selection of advanced magnetic materials for ultra-high-efficiency power transformers.