Comparing Lithium Battery Fault Detection Across Models
FEB 12, 20269 MIN READ
Generate Your Research Report Instantly with AI Agent
PatSnap Eureka helps you evaluate technical feasibility & market potential.
Lithium Battery Fault Detection Background and Objectives
Lithium-ion batteries have become the dominant energy storage solution across diverse applications, from consumer electronics to electric vehicles and grid-scale energy systems. This widespread adoption stems from their superior energy density, extended cycle life, and declining manufacturing costs over the past two decades. However, the increasing deployment scale has simultaneously elevated safety concerns, as battery faults can lead to thermal runaway, fires, and catastrophic failures with significant economic and safety implications.
The complexity of lithium battery systems presents substantial challenges for fault detection. Battery degradation mechanisms are multifaceted, involving electrochemical reactions, mechanical stress, thermal effects, and electrical anomalies that interact in non-linear ways. Traditional detection methods often rely on single-parameter monitoring or rule-based thresholds, which prove insufficient for capturing the intricate fault signatures across different battery chemistries, operating conditions, and aging states.
Recent technological advances in data-driven modeling and machine learning have opened new pathways for battery fault detection. Various model architectures, ranging from classical statistical approaches to deep neural networks, demonstrate distinct capabilities in identifying fault patterns from battery operational data. Each modeling approach exhibits unique strengths in feature extraction, temporal dependency capture, and generalization across diverse fault scenarios. The comparative evaluation of these models remains critical for understanding their practical applicability and limitations.
The primary objective of this technical investigation is to systematically compare fault detection performance across different modeling frameworks when applied to lithium battery systems. This comparison aims to establish a comprehensive understanding of how various model architectures handle the complexity of battery fault signatures, their sensitivity to different fault types, and their robustness under varying operational conditions. The analysis seeks to identify optimal modeling strategies that balance detection accuracy, computational efficiency, and real-world deployability.
Furthermore, this research targets the development of actionable insights for selecting appropriate fault detection models based on specific application requirements, data availability, and computational constraints. By evaluating models across standardized metrics and diverse fault scenarios, the investigation aims to provide strategic guidance for enterprises developing battery management systems and safety monitoring solutions in an increasingly electrified industrial landscape.
The complexity of lithium battery systems presents substantial challenges for fault detection. Battery degradation mechanisms are multifaceted, involving electrochemical reactions, mechanical stress, thermal effects, and electrical anomalies that interact in non-linear ways. Traditional detection methods often rely on single-parameter monitoring or rule-based thresholds, which prove insufficient for capturing the intricate fault signatures across different battery chemistries, operating conditions, and aging states.
Recent technological advances in data-driven modeling and machine learning have opened new pathways for battery fault detection. Various model architectures, ranging from classical statistical approaches to deep neural networks, demonstrate distinct capabilities in identifying fault patterns from battery operational data. Each modeling approach exhibits unique strengths in feature extraction, temporal dependency capture, and generalization across diverse fault scenarios. The comparative evaluation of these models remains critical for understanding their practical applicability and limitations.
The primary objective of this technical investigation is to systematically compare fault detection performance across different modeling frameworks when applied to lithium battery systems. This comparison aims to establish a comprehensive understanding of how various model architectures handle the complexity of battery fault signatures, their sensitivity to different fault types, and their robustness under varying operational conditions. The analysis seeks to identify optimal modeling strategies that balance detection accuracy, computational efficiency, and real-world deployability.
Furthermore, this research targets the development of actionable insights for selecting appropriate fault detection models based on specific application requirements, data availability, and computational constraints. By evaluating models across standardized metrics and diverse fault scenarios, the investigation aims to provide strategic guidance for enterprises developing battery management systems and safety monitoring solutions in an increasingly electrified industrial landscape.
Market Demand for Battery Safety and Diagnostics
The global shift toward electrification in transportation, energy storage, and consumer electronics has intensified the focus on lithium battery safety and reliability. As battery systems become more complex and energy-dense, the risk of thermal runaway, capacity degradation, and catastrophic failure increases proportionally. This has created substantial market demand for advanced fault detection technologies capable of identifying anomalies across diverse battery chemistries, form factors, and operational environments.
Regulatory pressures are accelerating this demand. Governments worldwide are implementing stricter safety standards for electric vehicles and energy storage systems, mandating real-time monitoring and predictive diagnostics. The automotive sector, which accounts for a significant portion of lithium battery consumption, requires fault detection systems that can operate reliably across multiple vehicle models and battery architectures. Insurance companies and fleet operators are also driving demand for diagnostic solutions that reduce liability and extend asset lifespans.
The energy storage sector presents another critical demand driver. Grid-scale battery installations require continuous health monitoring to prevent service interruptions and ensure return on investment. Commercial and residential storage systems similarly need diagnostic capabilities that can detect early-stage faults before they escalate into safety incidents or performance degradation. The proliferation of second-life battery applications further amplifies the need for robust diagnostic tools capable of assessing batteries with varied usage histories.
Consumer electronics manufacturers face mounting pressure to prevent battery-related product recalls and safety incidents. High-profile failures have heightened public awareness and regulatory scrutiny, making advanced diagnostics a competitive differentiator. The market increasingly values solutions that can generalize across different battery models without requiring extensive recalibration, reducing time-to-market and development costs.
The convergence of these factors has created a substantial addressable market for battery diagnostic technologies. Industry stakeholders seek solutions that balance accuracy, computational efficiency, and adaptability across battery types. The ability to compare and validate fault detection performance across models has become essential for technology adoption and standardization efforts.
Regulatory pressures are accelerating this demand. Governments worldwide are implementing stricter safety standards for electric vehicles and energy storage systems, mandating real-time monitoring and predictive diagnostics. The automotive sector, which accounts for a significant portion of lithium battery consumption, requires fault detection systems that can operate reliably across multiple vehicle models and battery architectures. Insurance companies and fleet operators are also driving demand for diagnostic solutions that reduce liability and extend asset lifespans.
The energy storage sector presents another critical demand driver. Grid-scale battery installations require continuous health monitoring to prevent service interruptions and ensure return on investment. Commercial and residential storage systems similarly need diagnostic capabilities that can detect early-stage faults before they escalate into safety incidents or performance degradation. The proliferation of second-life battery applications further amplifies the need for robust diagnostic tools capable of assessing batteries with varied usage histories.
Consumer electronics manufacturers face mounting pressure to prevent battery-related product recalls and safety incidents. High-profile failures have heightened public awareness and regulatory scrutiny, making advanced diagnostics a competitive differentiator. The market increasingly values solutions that can generalize across different battery models without requiring extensive recalibration, reducing time-to-market and development costs.
The convergence of these factors has created a substantial addressable market for battery diagnostic technologies. Industry stakeholders seek solutions that balance accuracy, computational efficiency, and adaptability across battery types. The ability to compare and validate fault detection performance across models has become essential for technology adoption and standardization efforts.
Current Challenges in Cross-Model Fault Detection
Cross-model fault detection in lithium batteries faces significant technical barriers stemming from the inherent heterogeneity of battery systems. Different battery models exhibit distinct electrochemical characteristics, thermal behaviors, and degradation patterns due to variations in cell chemistry, manufacturing processes, and design specifications. This diversity creates substantial challenges when attempting to develop unified diagnostic frameworks that maintain accuracy across multiple battery types without extensive model-specific calibration.
The primary obstacle lies in feature representation inconsistency across different battery architectures. Voltage curves, impedance spectra, and thermal signatures that indicate specific fault conditions in one battery model may manifest differently or remain absent in another. Traditional machine learning approaches trained on single-model datasets demonstrate poor generalization capabilities when deployed on unfamiliar battery configurations, often resulting in elevated false positive rates or missed fault detections that compromise system reliability.
Data scarcity and imbalance present another critical constraint. Acquiring comprehensive fault datasets across multiple battery models requires extensive testing under various failure modes, which is both time-consuming and costly. Real-world fault occurrences are relatively rare events, leading to severely imbalanced datasets that hinder the development of robust cross-model detection algorithms. This limitation is particularly acute for emerging battery chemistries where historical operational data remains limited.
Domain adaptation complexity further compounds these challenges. While transfer learning techniques offer potential solutions, the significant domain shift between different battery models often exceeds the adaptation capacity of conventional methods. The nonlinear and time-varying nature of battery degradation mechanisms introduces additional complexity, as fault signatures evolve differently across models throughout their operational lifecycles. Establishing universal fault indicators that remain valid across diverse battery architectures while accounting for temporal dynamics requires sophisticated feature engineering and model architectures.
Computational constraints in embedded battery management systems impose practical limitations on algorithm complexity. Cross-model detection frameworks must balance diagnostic accuracy with real-time processing requirements and limited onboard computational resources, restricting the deployment of computationally intensive deep learning models that might otherwise improve generalization performance across battery types.
The primary obstacle lies in feature representation inconsistency across different battery architectures. Voltage curves, impedance spectra, and thermal signatures that indicate specific fault conditions in one battery model may manifest differently or remain absent in another. Traditional machine learning approaches trained on single-model datasets demonstrate poor generalization capabilities when deployed on unfamiliar battery configurations, often resulting in elevated false positive rates or missed fault detections that compromise system reliability.
Data scarcity and imbalance present another critical constraint. Acquiring comprehensive fault datasets across multiple battery models requires extensive testing under various failure modes, which is both time-consuming and costly. Real-world fault occurrences are relatively rare events, leading to severely imbalanced datasets that hinder the development of robust cross-model detection algorithms. This limitation is particularly acute for emerging battery chemistries where historical operational data remains limited.
Domain adaptation complexity further compounds these challenges. While transfer learning techniques offer potential solutions, the significant domain shift between different battery models often exceeds the adaptation capacity of conventional methods. The nonlinear and time-varying nature of battery degradation mechanisms introduces additional complexity, as fault signatures evolve differently across models throughout their operational lifecycles. Establishing universal fault indicators that remain valid across diverse battery architectures while accounting for temporal dynamics requires sophisticated feature engineering and model architectures.
Computational constraints in embedded battery management systems impose practical limitations on algorithm complexity. Cross-model detection frameworks must balance diagnostic accuracy with real-time processing requirements and limited onboard computational resources, restricting the deployment of computationally intensive deep learning models that might otherwise improve generalization performance across battery types.
Existing Fault Detection Solutions Across Battery Models
01 Battery management system with voltage and current monitoring
Lithium battery fault detection can be achieved through comprehensive battery management systems that continuously monitor voltage, current, and other electrical parameters. These systems employ sensors and data acquisition modules to collect real-time operational data from battery cells or packs. By analyzing voltage variations, current fluctuations, and charge-discharge characteristics, the system can identify abnormal conditions such as overcharging, over-discharging, or internal short circuits. Advanced algorithms process the collected data to detect deviations from normal operating ranges and trigger alerts or protective actions.- Voltage-based fault detection methods: Lithium battery fault detection can be achieved by monitoring voltage characteristics of battery cells or packs. This approach involves measuring voltage variations, voltage differences between cells, and voltage thresholds to identify abnormal conditions such as overcharging, over-discharging, or cell inconsistency. Voltage monitoring systems can detect faults in real-time by comparing measured values against predetermined safe operating ranges and triggering alerts when deviations occur.
- Temperature monitoring and thermal fault detection: Temperature-based detection methods focus on identifying thermal anomalies in lithium batteries that may indicate potential faults or safety risks. These systems monitor temperature distributions across battery modules, detect hotspots, and identify thermal runaway conditions. By analyzing temperature data patterns and rates of temperature change, these methods can predict and prevent catastrophic failures before they occur.
- Machine learning and AI-based diagnostic approaches: Advanced fault detection systems employ machine learning algorithms and artificial intelligence to analyze battery operational data and identify fault patterns. These methods process multiple parameters simultaneously, including current, voltage, temperature, and impedance data, to build predictive models that can detect early signs of degradation or malfunction. Neural networks and deep learning techniques enable more accurate fault classification and prognosis.
- Impedance and internal resistance analysis: Fault detection through impedance spectroscopy and internal resistance measurement provides insights into battery health and performance degradation. This technique involves applying AC signals at various frequencies to measure the battery's impedance characteristics, which change as the battery ages or develops faults. Internal resistance monitoring can identify issues such as electrode degradation, electrolyte decomposition, and connection problems.
- Multi-parameter integrated monitoring systems: Comprehensive fault detection systems integrate multiple sensing technologies and diagnostic parameters to provide holistic battery health assessment. These systems combine voltage, current, temperature, and state-of-charge monitoring with advanced algorithms to detect various fault types including short circuits, open circuits, capacity fade, and safety hazards. Integration of multiple detection methods improves accuracy and reduces false alarms.
02 Machine learning and artificial intelligence based fault diagnosis
Advanced fault detection methods utilize machine learning algorithms and artificial intelligence techniques to identify battery anomalies. These approaches involve training models on historical battery data to recognize patterns associated with various fault conditions. Neural networks, support vector machines, and deep learning architectures can be employed to classify battery states and predict potential failures. The AI-based systems can detect subtle changes in battery behavior that may indicate early-stage faults, enabling predictive maintenance and preventing catastrophic failures.Expand Specific Solutions03 Thermal monitoring and temperature-based fault detection
Temperature monitoring is a critical aspect of lithium battery fault detection, as thermal anomalies often indicate internal faults or abuse conditions. Detection systems incorporate temperature sensors at multiple locations within the battery pack to monitor thermal distribution and identify hotspots. By analyzing temperature gradients, rate of temperature change, and absolute temperature values, the system can detect conditions such as thermal runaway, internal short circuits, or cooling system failures. Thermal imaging and infrared sensing technologies may also be employed for non-contact temperature measurement.Expand Specific Solutions04 Impedance spectroscopy and internal resistance analysis
Electrochemical impedance spectroscopy and internal resistance measurement provide valuable insights into battery health and fault conditions. These techniques involve applying AC signals at various frequencies or measuring DC resistance to characterize the battery's internal impedance. Changes in impedance patterns can indicate degradation mechanisms, electrolyte decomposition, electrode damage, or connection failures. By tracking impedance evolution over time and comparing with baseline values, the system can detect early-stage faults and estimate remaining useful life.Expand Specific Solutions05 State estimation and model-based fault detection
Model-based approaches utilize mathematical models of battery behavior to estimate internal states and detect faults through residual analysis. These methods employ equivalent circuit models, electrochemical models, or data-driven models to predict expected battery responses under various operating conditions. By comparing actual measurements with model predictions, the system can identify discrepancies that indicate fault conditions. State observers, Kalman filters, and particle filters are commonly used for state estimation. This approach enables detection of faults that may not be directly measurable, such as capacity fade, power fade, or internal structural damage.Expand Specific Solutions
Key Players in Battery Management Systems Industry
The lithium battery fault detection sector is experiencing rapid evolution as the industry transitions from early adoption to mainstream deployment, driven by the exponential growth of electric vehicles and energy storage systems. The global market is expanding significantly, with increasing demand for reliable battery diagnostics across automotive, industrial, and grid-scale applications. Technology maturity varies considerably among key players: established manufacturers like LG Energy Solution, DENSO Corp., and Robert Bosch GmbH leverage decades of automotive expertise to develop robust detection systems, while specialized firms such as ACCURE Battery Intelligence focus exclusively on AI-driven predictive analytics. Automotive leaders including AUDI AG and Volvo Personvagnar AB integrate fault detection into vehicle architectures, whereas battery producers like GS Yuasa International and Hefei Guoxuan High-Tech Power Energy emphasize cell-level monitoring. Research institutions including Shandong University, Beijing Institute of Technology, and University of Michigan contribute foundational algorithms, bridging academic innovation with industrial implementation as the technology matures toward standardized, cross-model diagnostic frameworks.
LG Energy Solution Ltd.
Technical Solution: LG Energy Solution has developed a comprehensive Battery Management System (BMS) incorporating multi-model fault detection approaches. Their system utilizes machine learning algorithms combined with electrochemical impedance spectroscopy (EIS) to detect internal short circuits, thermal runaway precursors, and cell degradation patterns[1][4]. The solution integrates real-time voltage monitoring, temperature sensing arrays, and current profiling across individual cells within battery packs. Their diagnostic framework employs both model-based methods using equivalent circuit models and data-driven approaches using neural networks to compare fault signatures. The system achieves early fault detection by analyzing deviation patterns between predicted and actual battery behavior, enabling predictive maintenance capabilities for electric vehicle and energy storage applications[7][9].
Strengths: Industry-leading scale with extensive real-world validation data from millions of deployed battery packs; integrated hardware-software solution. Weaknesses: Proprietary system with limited interoperability; high implementation costs for third-party integration.
Robert Bosch GmbH
Technical Solution: Bosch has developed an advanced fault detection system that compares multiple diagnostic models including equivalent circuit models, electrochemical models, and AI-based anomaly detection algorithms[2][5]. Their approach uses a hybrid architecture that combines physics-based models for understanding degradation mechanisms with machine learning models trained on extensive field data from automotive applications. The system performs continuous impedance estimation, capacity fade tracking, and thermal anomaly detection. Bosch's solution features model ensemble techniques that cross-validate findings across different diagnostic approaches, reducing false positives while improving early fault detection accuracy. The platform supports both embedded implementation for vehicle BMS and cloud-based analytics for fleet-level monitoring[8][11].
Strengths: Robust validation through extensive automotive partnerships; excellent balance between computational efficiency and accuracy; strong integration with vehicle systems. Weaknesses: Primarily optimized for automotive applications; may require adaptation for stationary storage systems.
Core Algorithms for Model-Agnostic Fault Detection
A machine learning-based approach for sensor fault detection in battery management systems
PatentPendingIN202441010827A
Innovation
- A machine learning-based approach for sensor fault detection is developed, utilizing techniques such as the interclass correlation coefficient, precise equivalent circuit models, and machine learning parameter estimation, along with new sensors like pressure and acoustic sensors, to accurately isolate and diagnose faults, leveraging algorithms like ANN, RF, SPCA, SVM, and deep residual networks for robust and efficient fault detection.
Accurate fault detection/diagnosis of lithium-ion battery using machine learning
PatentPendingIN202241040783A
Innovation
- A machine learning-based diagnostic system that identifies early markers of potential short circuits and thermal runaway by analyzing first-order, second-order, and composite response parameters of the battery's time-varying electrical state, enabling real-time detection and prevention of hazardous conditions.
Safety Standards and Regulations for Battery Systems
The regulatory landscape for lithium battery fault detection systems is shaped by multiple international and regional safety standards that establish minimum requirements for battery management systems and diagnostic capabilities. IEC 62619 and IEC 62620 define safety requirements for secondary lithium cells and batteries used in industrial applications and electric vehicles, mandating specific fault detection mechanisms including overvoltage, undervoltage, overcurrent, and thermal monitoring. UL 2580 and UL 1973 provide comprehensive safety certification criteria for battery systems in electric vehicles and stationary energy storage applications, requiring validated fault detection algorithms capable of identifying cell-level anomalies before they escalate to safety hazards.
Regulatory frameworks increasingly emphasize the need for model-agnostic fault detection approaches that can adapt to different battery chemistries and configurations. The UN ECE R100 regulation for electric vehicle safety mandates that battery management systems must detect and respond to fault conditions within specified time thresholds, regardless of the underlying detection methodology employed. ISO 26262 functional safety standards require systematic validation of fault detection algorithms across various operating scenarios, establishing reliability metrics that detection models must achieve to ensure automotive-grade safety compliance.
Recent regulatory developments reflect growing concerns about thermal runaway prevention and early fault identification. The Chinese GB 38031 standard and European Union's proposed Battery Regulation both introduce stricter requirements for predictive fault detection capabilities, encouraging the adoption of advanced machine learning models that can identify degradation patterns before critical failures occur. These regulations specify minimum detection accuracy rates and maximum false positive thresholds, creating performance benchmarks that different detection models must satisfy.
Certification processes now require comprehensive documentation demonstrating how fault detection systems perform across diverse battery configurations and operational conditions. Testing protocols defined in SAE J2464 and ISO 12405 series standards mandate validation across temperature ranges, state-of-charge levels, and aging conditions, ensuring that detection models maintain effectiveness throughout battery lifecycle. Compliance verification increasingly involves third-party validation of detection algorithm performance, establishing independent benchmarks for comparing different modeling approaches in meeting safety objectives.
Regulatory frameworks increasingly emphasize the need for model-agnostic fault detection approaches that can adapt to different battery chemistries and configurations. The UN ECE R100 regulation for electric vehicle safety mandates that battery management systems must detect and respond to fault conditions within specified time thresholds, regardless of the underlying detection methodology employed. ISO 26262 functional safety standards require systematic validation of fault detection algorithms across various operating scenarios, establishing reliability metrics that detection models must achieve to ensure automotive-grade safety compliance.
Recent regulatory developments reflect growing concerns about thermal runaway prevention and early fault identification. The Chinese GB 38031 standard and European Union's proposed Battery Regulation both introduce stricter requirements for predictive fault detection capabilities, encouraging the adoption of advanced machine learning models that can identify degradation patterns before critical failures occur. These regulations specify minimum detection accuracy rates and maximum false positive thresholds, creating performance benchmarks that different detection models must satisfy.
Certification processes now require comprehensive documentation demonstrating how fault detection systems perform across diverse battery configurations and operational conditions. Testing protocols defined in SAE J2464 and ISO 12405 series standards mandate validation across temperature ranges, state-of-charge levels, and aging conditions, ensuring that detection models maintain effectiveness throughout battery lifecycle. Compliance verification increasingly involves third-party validation of detection algorithm performance, establishing independent benchmarks for comparing different modeling approaches in meeting safety objectives.
Data Interoperability and Model Generalization Strategies
Data interoperability represents a fundamental prerequisite for enabling cross-model comparison in lithium battery fault detection systems. The heterogeneity of data sources, including variations in sensor configurations, sampling frequencies, battery chemistries, and operational environments, creates significant barriers to model generalization. Establishing standardized data formats and unified feature representations becomes essential for facilitating meaningful comparisons across different detection models. Industry initiatives toward common data schemas, such as battery management system communication protocols and standardized diagnostic parameters, provide foundational infrastructure for interoperability. However, practical implementation requires addressing semantic inconsistencies, unit conversions, and temporal alignment challenges inherent in multi-source battery datasets.
Model generalization strategies must account for domain shift phenomena that occur when detection algorithms trained on specific battery types or operating conditions are applied to different scenarios. Transfer learning techniques offer promising approaches by leveraging knowledge from source domains to enhance performance in target domains with limited labeled data. Domain adaptation methods, including feature alignment and adversarial training, help mitigate distribution discrepancies between training and deployment environments. Meta-learning frameworks enable models to rapidly adapt to new battery configurations through few-shot learning paradigms, reducing dependency on extensive retraining datasets.
Federated learning architectures present innovative solutions for collaborative model development while preserving data privacy across organizations. This approach allows multiple stakeholders to contribute to model improvement without centralizing sensitive operational data. Ensemble methods that combine predictions from models trained on diverse datasets can enhance robustness and generalization capability. Feature engineering strategies focusing on physics-informed representations rather than raw sensor data improve transferability across different battery systems by capturing fundamental degradation mechanisms.
Validation frameworks for assessing model generalization require carefully designed cross-validation protocols that simulate real-world deployment scenarios. Stratified sampling across battery types, usage patterns, and fault categories ensures comprehensive evaluation of model performance boundaries. Continuous monitoring and adaptive recalibration mechanisms maintain detection accuracy as battery populations age and operational conditions evolve, addressing the temporal dimension of generalization challenges in practical applications.
Model generalization strategies must account for domain shift phenomena that occur when detection algorithms trained on specific battery types or operating conditions are applied to different scenarios. Transfer learning techniques offer promising approaches by leveraging knowledge from source domains to enhance performance in target domains with limited labeled data. Domain adaptation methods, including feature alignment and adversarial training, help mitigate distribution discrepancies between training and deployment environments. Meta-learning frameworks enable models to rapidly adapt to new battery configurations through few-shot learning paradigms, reducing dependency on extensive retraining datasets.
Federated learning architectures present innovative solutions for collaborative model development while preserving data privacy across organizations. This approach allows multiple stakeholders to contribute to model improvement without centralizing sensitive operational data. Ensemble methods that combine predictions from models trained on diverse datasets can enhance robustness and generalization capability. Feature engineering strategies focusing on physics-informed representations rather than raw sensor data improve transferability across different battery systems by capturing fundamental degradation mechanisms.
Validation frameworks for assessing model generalization require carefully designed cross-validation protocols that simulate real-world deployment scenarios. Stratified sampling across battery types, usage patterns, and fault categories ensures comprehensive evaluation of model performance boundaries. Continuous monitoring and adaptive recalibration mechanisms maintain detection accuracy as battery populations age and operational conditions evolve, addressing the temporal dimension of generalization challenges in practical applications.
Unlock deeper insights with PatSnap Eureka Quick Research — get a full tech report to explore trends and direct your research. Try now!
Generate Your Research Report Instantly with AI Agent
Supercharge your innovation with PatSnap Eureka AI Agent Platform!