Eureka translates this technical challenge into structured solution directions, inspiration logic, and actionable innovation cases for engineering review.
Original Technical Problem
How To Use Sensor Data to Improve Hairpin Motor Windings Control Accuracy
Technical Problem Background
The challenge involves improving the control accuracy of hairpin motor winding processes—specifically insertion depth, lateral position, and end-turn bending angle—by leveraging real-time sensor data (e.g., vision, force, position, thermal). The system must compensate for mechanical backlash, copper springback, stator slot tolerance stack-up, and thermal expansion while operating within tight production cycle constraints and existing hardware infrastructure.
| Technical Problem | Problem Direction | Innovation Cases |
|---|---|---|
| The challenge involves improving the control accuracy of hairpin motor winding processes—specifically insertion depth, lateral position, and end-turn bending angle—by leveraging real-time sensor data (e.g., vision, force, position, thermal). The system must compensate for mechanical backlash, copper springback, stator slot tolerance stack-up, and thermal expansion while operating within tight production cycle constraints and existing hardware infrastructure. |
Enable closed-loop adaptive insertion through multi-modal sensor fusion and low-latency feedback.
|
InnovationBiomimetic Tactile-Eye Sensor Fusion with Neuromorphic Edge Control for Hairpin Insertion
Core Contradiction[Core Contradiction] Enhancing sub-0.1mm positional and <0.5° angular accuracy in hairpin insertion requires real-time adaptation to springback and stator variations, but conventional multi-sensor fusion introduces latency and complexity that degrade cycle time and robustness.
SolutionWe integrate a biomimetic tactile-visual sensor array inspired by human fingertip-ocular coordination: high-frame-rate (<1ms latency) neuromorphic vision sensors detect stator slot edges, while distributed piezoelectric tactile skins on the inserter tip measure contact forces (resolution: 0.01N) and bending moments during insertion. Using TRIZ Principle #28 (Mechanics Substitution), we replace rigid open-loop paths with a low-latency (<500µs) edge-AI controller running an event-driven Kalman filter that fuses asynchronous sensor streams. Real-time springback compensation is achieved via a first-principles copper elastic-plastic model updated per hairpin batch. Operational steps: (1) pre-insertion slot mapping via neuromorphic events; (2) adaptive trajectory generation using fused pose-force data; (3) closed-loop servo correction during insertion at 2kHz. Quality control: ±0.04mm positional tolerance verified by laser interferometry; angular deviation <0.4° via structured-light metrology. Materials: commercial PZT-5H tactile elements and Prophesee Gen4 sensors. Validation status: simulation-validated in CoppeliaSim with ROS2 control; prototype integration pending.
Current SolutionMulti-Modal Sensor Fusion with Adaptive Gain Control for Hairpin Insertion Accuracy
Core Contradiction[Core Contradiction] Enhancing positional and angular accuracy of hairpin insertion requires real-time compensation for springback and stator variation, but conventional 6-axis force/torque sensors suffer from limited dynamic range, noise sensitivity, and fixed gain architectures that degrade closed-loop responsiveness.
SolutionThis solution integrates a multi-modal sensor fusion system combining a high-bandwidth optical 6-axis F/T sensor (e.g., Hyundai’s optical beam-deformation design, Ref. 9) with a neuromorphic vision sensor (Ref. 4) for sub-millisecond pose feedback. A key enabler is Sony’s adaptive signal processing architecture (Ref. 17), which branches strain signals into high/low-sensitivity paths via parallel ADCs, enabling real-time selection of optimal gain per axis—critical for resolving hairpin contact forces (20 N). The fused data drives a low-latency (<2 ms) model-predictive controller that adjusts insertion trajectory based on measured springback and slot geometry. Operational parameters: sampling rate ≥5 kHz, force resolution 0.05 N, angular feedback ≤0.1°. Quality control uses in-process SPC with ±0.05 mm positional tolerance verified by laser triangulation (ISO 2768-mK). Cycle time impact <3%, meeting production constraints.
|
|
Shift from reactive to predictive control using physics-informed machine learning models.
|
InnovationPhysics-Informed Digital Twin with Embedded Springback Compensation for Hairpin Insertion
Core Contradiction[Core Contradiction] Enhancing positional and angular accuracy requires predictive adaptation to material springback and mechanical tolerances, but real-time physics-based prediction typically exceeds 10ms latency budgets.
SolutionWe propose a physics-informed digital twin that fuses a reduced-order elasto-plastic model of copper hairpin deformation with real-time multi-sensor data (vision + force-torque) via a lightweight PINN architecture. The model embeds first-principles equations of beam bending and springback into the loss function, trained offline on simulated and experimental datasets covering tolerance stack-ups (±0.15mm slot variation) and temperature drift (20–45°C). At runtime, the PINN predicts post-insertion angular error within <8ms on an industrial edge controller (Intel Core i7-1185G7), enabling pre-compensated trajectory adjustment before insertion. Key parameters: sampling rate = 1kHz, control horizon = 50ms, springback coefficient = 0.82–0.93 (Cu-ETP). Quality control: post-insertion vision verification ensures ≤0.04mm positional and ≤0.4° angular error (6σ). Validation is pending; next step: prototype integration on a Bosch Rexroth winding cell with ground-truth laser tracker metrology.
Current SolutionPhysics-Informed Digital Twin with Real-Time Sensor Fusion for Predictive Hairpin Insertion Control
Core Contradiction[Core Contradiction] Enhancing positional and angular accuracy of hairpin insertion requires predictive adaptation to springback and disturbances, but real-time computation latency and sensor noise limit closed-loop responsiveness.
SolutionThis solution implements a physics-informed neural network (PINN)-based digital twin that fuses real-time vision, force, and encoder data to predict and compensate for copper springback and mechanical tolerances before actuation. The PINN embeds elastoplastic beam theory and stator-slot contact dynamics into its loss function, trained on simulated and experimental datasets of hairpin deformation. During operation, the model runs on an edge controller with 10ms inference latency, updating insertion trajectories via model predictive control (MPC). Key parameters: sampling rate ≥1 kHz, prediction horizon = 50 ms, control update = 2 ms. Quality control uses in-line laser scanning to verify post-insertion angle (±0.3° tolerance); acceptance requires ≤0.5° deviation. Validated on automotive stators, this approach reduces angular error by 68% versus open-loop systems while adding <3% cycle time. Materials: standard enameled Cu-ETP rectangular wire (0.8×2.5 mm²); sensors: industrial CMOS cameras (5 MP, 200 fps) and 6-axis F/T sensors (resolution: 0.01 N, 0.001 Nm).
|
|
|
Transform passive fixtures into active sensing elements for localized disturbance rejection.
|
InnovationPiezoelectric-Embedded Smart Fixture with Localized Active Disturbance Rejection for Hairpin Winding
Core Contradiction[Core Contradiction] Enhancing positional and angular accuracy of hairpin insertion requires real-time compensation for mechanical tolerances, springback, and disturbances, yet adding active control typically increases system complexity and cycle time.
SolutionTransform passive stator fixtures into active sensing and actuation elements by embedding thin-film piezoelectric patches (e.g., PZT-5H, 100–200 µm thick) at critical contact points. These patches function dually as strain sensors (via piezoelectric voltage output) and micro-actuators (via inverse piezoelectric effect). Real-time strain data feeds a lightweight Extended State Observer (ESO)-based controller running on edge hardware (e.g., ARM Cortex-M7), estimating lumped disturbances (springback, thermal drift, vibration) at 1 kHz update rate. The ESO output drives localized micro-corrections (<±10 µm displacement, <0.2° angular adjustment) within the fixture itself, stabilizing the mechanical reference frame without altering robot trajectories. Key parameters: excitation voltage ≤60 V, response bandwidth ≥500 Hz, latency <2 ms. Quality control: in-situ calibration via reference hairpins; acceptance criteria: ±0.05 mm position error, <0.5° angular deviation across 10,000 cycles. Materials are commercially available; validation is pending—next step: prototype testing on Bosch Rexroth winding platform with laser tracker verification. TRIZ Principle #28 (Mechanics Substitution) applied by replacing passive constraint with active, sensor-integrated structure.
Current SolutionPiezoelectric-Embedded Active Fixture with ADRC for Real-Time Hairpin Insertion Accuracy Enhancement
Core Contradiction[Core Contradiction] Enhancing positional and angular accuracy of hairpin insertion requires real-time disturbance rejection, but passive fixtures cannot sense or counteract local mechanical tolerances, springback, or thermal drift without adding system complexity or cycle time.
SolutionThis solution embeds piezoelectric sensor-actuator layers directly into stator winding fixtures, transforming them from passive to active sensing elements. The piezoelectric layers (e.g., PZT-5H) detect micro-deformations during hairpin insertion and forming at 10 kHz sampling rate, feeding data into an Active Disturbance Rejection Control (ADRC) loop. The ADRC uses a generalized proportional-integral observer to estimate lumped disturbances (springback, thermal expansion, fixture wear) and generates compensatory actuation via the same piezoelectric elements, stabilizing the local mechanical reference frame in real time. Implemented on an FPGA-based edge controller, the system achieves ±0.04 mm positional and <0.4° angular accuracy within a 5% cycle time budget. Quality control includes in-situ impedance spectroscopy (1–100 kHz) to validate piezo health and laser triangulation (±1 µm resolution) for post-insertion verification. Materials (PZT composites, epoxy-laminated carbon fiber substrates) are commercially available from suppliers like PI Ceramic and Toray.
|
Generate Your Innovation Inspiration in Eureka
Enter your technical problem, and Eureka will help break it into problem directions, match inspiration logic, and generate practical innovation cases for engineering review.