Comprehensive Review of Dielectric, Impedance, and Soft Computing Techniques for Lubricant Condition Monitoring and Predictive Maintenance in Diesel Engines

Abstract

Lubricant condition analysis is a valuable diagnostic tool for assessing engine performance and ensuring the reliable operation of diesel engines. While traditional diagnostic techniques—such as Fourier transform infrared spectroscopy (FTIR)—are constrained by slow response times, high costs, and the need for specialized personnel. In contrast, dielectric spectroscopy, impedance analysis, and soft computing offer real-time, non-destructive, and cost-effective alternatives. This review examines recent advances in integrating these techniques to predict lubricant properties, evaluate wear conditions, and optimize maintenance scheduling. In particular, dielectric and impedance spectroscopies offer insights into electrical properties linked to oil degradation, such as changes in viscosity and the presence of wear particles. When combined with soft computing algorithms, these methods enhance data analysis, reduce reliance on expert interpretation, and improve predictive accuracy. The review also addresses challenges—including complex data interpretation, limited sample sizes, and the necessity for robust models to manage variability in real-world operations. Future research directions emphasize miniaturization, expanding the range of detectable contaminants, and incorporating multi-modal artificial intelligence to further bolster system robustness. Collectively, these innovations signal a shift from reactive to predictive maintenance strategies, with the potential to reduce costs, minimize downtime, and enhance overall engine reliability. This comprehensive review provides valuable insights for researchers, engineers, and maintenance professionals dedicated to advancing diesel engine lubricant monitoring.

1. Introduction

In diesel engines, monitoring oil condition is essential for ensuring the longevity, efficiency, and reliability of internal combustion systems. Lubricants play a critical role in reducing friction, managing heat, and preventing wear [1,2,3]. However, the degradation of oil—driven by factors such as fuel type, operating conditions, and additive interactions—can lead to reduced fuel efficiency, increased emissions, and accelerated wear [4,5,6,7,8]. Traditional analysis methods, including Fourier transform infrared spectroscopy (FTIR), are widely used to evaluate oil condition by detecting contaminants and measuring key properties such as viscosity and total base number (TBN), a measure of oil’s ability to neutralize acidic byproducts [9,10]. Despite their effectiveness, these techniques require specialized expertise, considerable time, and significant financial resources, limiting their practicality in many applications [11,12,13].

Advanced techniques such as impedance spectroscopy, dielectric analysis, and soft computing have emerged as promising solutions to the limitations of traditional oil analysis methods. Dielectric spectroscopy measures a lubricant’s electrical permittivity (ability to store energy in an electric field) and dielectric loss (energy dissipation), which correlate with degradation. Impedance analysis evaluates resistance to alternating current, reflecting conductivity changes due to contaminants. Soft computing (e.g., neural networks) applies AI to interpret complex data patterns, enabling predictive maintenance. Specifically, impedance [14,15,16,17,18,19,20,21,22,23] and dielectric [24,25,26] spectroscopy offer non-destructive, cost-effective, and highly precise means to assess the electrical properties of lubricants, which closely reflect their physical and chemical states. Soft computing approaches—including genetic algorithms, artificial neural networks (ANNs), and support vector machines (SVMs) [27,28,29,30,31,32,33,34]—leverage artificial intelligence and machine learning to interpret complex datasets, thereby reducing reliance on expert input and enhancing the accuracy of oil property predictions. Moreover, the integration of soft computing with impedance and dielectric techniques has demonstrated strong correlations between electrical characteristics and oil contaminants [13,35,36,37,38,39,40]. In this review, we critically examine these studies [13,14,15,16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40].

Transitioning from reactive to proactive, condition-based maintenance has underscored the critical importance of real-time, continuous monitoring of lubricant condition [41]. In situ measurements, which facilitate early detection of potential issues, combined with electrical techniques such as impedance and dielectric spectroscopy, can significantly reduce downtime and costly repairs [42,43]. Moreover, advances in multi-sensor data fusion and smart sensor systems have enabled real-time assessment of lubricant quality, estimation of remaining oil life, and a reduction in environmental impact [44,45,46]. Given that machine breakdowns can lead to serious safety hazards and substantial financial losses [47,48,49,50], these innovations are particularly valuable in sectors such as energy production, automotive, and aerospace. Despite these advancements, challenges remain in fully harnessing the potential of impedance, dielectric, and soft computing approaches for engine oil condition monitoring. Further research is warranted to address issues such as complex data interpretation, the development of robust models for limited datasets, and the integration of these techniques into existing maintenance platforms [51]. Moreover, since the relationships between specific lubricant components and their electrical properties may not exhibit straightforward linear behavior, more sophisticated models will be required to accurately predict oil condition [52].

This review examines the combined application of impedance spectroscopy, dielectric analysis, and soft computing in engine oil condition monitoring. We assess the practical utility of these techniques in forecasting lubricant properties, detecting wear conditions, and optimizing maintenance schedules by critically evaluating recent research and advancements. The article also addresses the challenges and future directions in this field, underscoring the potential of these methods to transform oil monitoring and contribute to the development of more reliable and efficient mechanical systems. Through this investigation, we aim to provide researchers, engineers, and maintenance professionals with valuable insights that pave the way for further advancements in lubricant condition monitoring.

2. Review Process

A systematic research strategy was implemented to ensure a thorough examination of the most relevant literature (see Figure 1). The methodology comprised the following steps:

A.

Keyword Search and Initial Screening

• A literature search was performed using the following keywords:

  -

  (“Condition monitoring” OR “Status monitoring”); “Lubricant”; “Diesel engine”

  -

  (“Dielectric” OR “Impedance”)

  -

  (“Soft computing” OR “Machine learning” OR “neural network” OR “Machine vision”)

• To capture current developments, the search was limited to publications from 2014 to 2024.
• Restricting the results to Q1, Q2, and Q3 journals yielded an initial pool of 413 articles, from which 174 were retained.

B.

Exclusion Criteria

• Review articles and non-English publications were excluded to focus exclusively on original research studies.
• Articles not directly addressing the application of dielectric, impedance, or soft computing techniques in lubricant condition monitoring were also removed.
• After applying these criteria, approximately 100 articles remained for further analysis.

C.

Abstract Review and Final Selection

• The abstracts of the remaining 100 articles were carefully reviewed to assess their relevance to the study’s objectives.
• Articles that primarily focused on sensor design or did not address the targeted techniques in lubricant condition monitoring were excluded.
• This process resulted in the final selection of 28 articles that fit precisely within the study framework and offered valuable insights into the application of these technologies in diesel engine lubricant monitoring.

D.

Data Extraction and Analysis

• The 28 selected articles (see Table 1) were categorized based on their focus on dielectric, impedance, or soft computing.
• Key findings, methodologies, and innovations from each article were extracted and summarized in comparative tables.
• A further analysis was conducted to identify emerging trends, challenges, and future directions in the field.

E.

Synthesis and Review

• A comprehensive synthesis of the extracted data has yielded an extensive overview of recent advancements in dielectric, impedance, and soft computing techniques for lubricant condition monitoring.
• The analysis highlights the potential of these methods to facilitate real-time, predictive maintenance, while also examining current limitations and opportunities for future research.

The systematic review of 28 studies (see Table 1) provides a foundation for evaluating the efficacy of impedance spectroscopy, dielectric analysis, and soft computing in lubricant monitoring, as detailed in the following sections. This review is intended to serve as a valuable resource for technicians, engineers, and maintenance professionals seeking to enhance engine reliability and performance through advanced monitoring technologies.

Table 1. Classification of articles studied in this research.

No.	Main Idea			Journal	Quartiles	Publish Year	Ref.
	Dielectric	Impedance	Soft Computing
1	*		*	Sensors and Actuators B: Chemical	Q1	2014	[36]
2		*		International Journal of Electrochemical Science	Q3	2015	[15]
3			*	Tribology International	Q1	2017	[27]
4	*			Tribology international	Q1	2018	[24]
5	*		*	Measurement	Q1	2019	[13]
6		*		IET Science, Measurement & Technology	Q2	2019	[14]
7	*			Environmental Science and Pollution Research	Q1	2020	[25]
8		*		Ocean Engineering	Q1	2020	[16]
9		*		IEEE Transactions on Industrial Electronics	Q1	2020	[17]
10		*		IEEE Sensors Journal	Q1	2021	[18]
11		*		Micromachines	Q2	2021	[19]
12			*	Tribology International	Q1	2021	[28]
13			*	Mechanics & Industry	Q3	2021	[29]
14	*			Materials	Q2	2022	[26]
15	*		*	Sensors and Actuators A: Physical	Q1	2022	[37]
16		*		Energies	Q1	2022	[20]
17		*		IEEE Transactions on Instrumentation and Measurement	Q1	2022	[21]
18			*	Neural Computing and Applications	Q1	2022	[30]
19			*	Expert Systems with Applications	Q1	2022	[31]
20	*	*		Sensors	Q1	2023	[35]
21	*		*	Lubricants	Q2	2023	[38]
22		*		IEEE Sensors Journal	Q1	2023	[22]
23			*	Wear	Q1	2023	[32]
24			*	ACS Omega	Q2	2023	[33]
25	*		*	Lubricants	Q2	2024	[39]
26		*		Sensors and Actuators A: Physical	Q1	2024	[23]
27			*	SAE International Journal of Fuels and Lubricants	Q3	2024	[34]
28		*	*	IEEE Sensors Journal	Q1	2024	[40]

*: Identifies the main idea of each article.

Table 1. Classification of articles studied in this research.

No.	Main Idea			Journal	Quartiles	Publish Year	Ref.
	Dielectric	Impedance	Soft Computing
1	*		*	Sensors and Actuators B: Chemical	Q1	2014	[36]
2		*		International Journal of Electrochemical Science	Q3	2015	[15]
3			*	Tribology International	Q1	2017	[27]
4	*			Tribology international	Q1	2018	[24]
5	*		*	Measurement	Q1	2019	[13]
6		*		IET Science, Measurement & Technology	Q2	2019	[14]
7	*			Environmental Science and Pollution Research	Q1	2020	[25]
8		*		Ocean Engineering	Q1	2020	[16]
9		*		IEEE Transactions on Industrial Electronics	Q1	2020	[17]
10		*		IEEE Sensors Journal	Q1	2021	[18]
11		*		Micromachines	Q2	2021	[19]
12			*	Tribology International	Q1	2021	[28]
13			*	Mechanics & Industry	Q3	2021	[29]
14	*			Materials	Q2	2022	[26]
15	*		*	Sensors and Actuators A: Physical	Q1	2022	[37]
16		*		Energies	Q1	2022	[20]
17		*		IEEE Transactions on Instrumentation and Measurement	Q1	2022	[21]
18			*	Neural Computing and Applications	Q1	2022	[30]
19			*	Expert Systems with Applications	Q1	2022	[31]
20	*	*		Sensors	Q1	2023	[35]
21	*		*	Lubricants	Q2	2023	[38]
22		*		IEEE Sensors Journal	Q1	2023	[22]
23			*	Wear	Q1	2023	[32]
24			*	ACS Omega	Q2	2023	[33]
25	*		*	Lubricants	Q2	2024	[39]
26		*		Sensors and Actuators A: Physical	Q1	2024	[23]
27			*	SAE International Journal of Fuels and Lubricants	Q3	2024	[34]
28		*	*	IEEE Sensors Journal	Q1	2024	[40]

*: Identifies the main idea of each article.

3. Diesel Engine Lubricant Condition Monitoring: Impedance Spectroscopy Applications

The reviewed studies (see

Table 2. Overview of impedance-based approaches for lubricant condition monitoring.

No.	Overview of	Description	Ref.
1	Research Objectives	Measure small water content (<0.05% vol) in mineral oil using EIS and dielectric constant; develop an in situ, low-cost IS-based system.	[14]
	Key Findings	• Linear correlation between water content and impedance/dielectric constant. • Feasibility of an IS-based system without sample treatment.
	Applications and Innovations	In situ moisture monitoring in oils; equivalent circuit modeling for property analysis.
	Future Directions	Optimize electrode geometry for precise capacitance; address challenges in the impedance module.
2	Research Objectives	Evaluate impedance spectroscopy for assessing sesame and almond oils as lubricants in four-ball tribometer wear tests compared with mineral oils.	[15]
	Key Findings	• Impedance distinguishes between fresh and used oils. • Almond oil outperforms sesame oil. • Wear levels correlate with impedance differences. • Vegetable and mineral oils are differentiated via impedance spectra.
	Applications and Innovations	Non-destructive oil condition monitoring using impedance spectroscopy.
	Future Directions	Focus on tribological testing rather than diesel engine applications.
3	Research Objectives	Improve inductive sensor sensitivity for non-ferromagnetic debris detection using a double-wire solenoid coil and inductance-resistance detection.	[16]
	Key Findings	• Detects 20 μm iron and 60 μm copper particles. • Combines inductance and resistance for material/size differentiation.
	Applications and Innovations	Marine machinery health monitoring via debris detection.
	Future Directions	Enhance shielding and signal processing; improve sensitivity for smaller non-ferrous particles.
4	Research Objectives	Develop a high-sensitivity impedance debris sensor with a high-gradient magnetic field and a square channel for increased throughput.	[17]
	Key Findings	• Detects 25 μm iron and 85 μm copper particles. • Dual-parameter (inductance/resistance) detection.
	Applications and Innovations	Hydraulic equipment health monitoring; portable oil condition devices.
	Future Directions	Improve noise filtering and anti-interference circuits; address variability from irregular particle shapes.
5	Research Objectives	Measure non-ferrous particle sizes in oil using real and imaginary impedance components; establish an elliptical function relationship for material-independent sizing.	[18]
	Key Findings	• Elliptical function correlates impedance changes with particle size. • Validated with copper and aluminum particles.
	Applications and Innovations	Material-independent particle sizing in lubricants.
	Future Directions	-
6	Research Objectives	Design an impedance debris sensor using silicon steel strips and double rectangular channels for real-time oil monitoring.	[19]
	Key Findings	• Detects 46 μm iron and 110 μm copper particles. • Series-connected coils improve signal-to-noise ratio (SNR).
	Applications and Innovations	Real-time machinery health monitoring; portable applications.
	Future Directions	Improve SNR with filters and amplifiers.
7	Research Objectives	Optimize a parallel resonance circuit via capacitance adjustment for wear debris detection.	[20]
	Key Findings	• Optimal capacitance differs for ferrous and non-ferrous particles. • Enhances sensitivity for both particle types.
	Applications and Innovations	Engine health monitoring through oil analysis.
	Future Directions	Validate in actual oil samples; improve throughput.
8	Research Objectives	Develop a theoretical model for impedance increments (inductance/resistance) induced by debris; validate the model with experiments.	[21]
	Key Findings	• Model validated for iron and copper particles. • Sensitivity improved for small debris (50–100 μm).
	Applications and Innovations	Early fault warning in marine machinery.
	Future Directions	Incorporate irregular debris shapes; integrate analysis of debris and oil parameters.
9	Research Objectives	Differentiate mixed metal particles using an inductive–capacitive dual detection method.	[22]
	Key Findings	• Dual-mode detection is feasible. • Accuracy depends on standard data.
	Applications and Innovations	Accurate lubricant monitoring devices; marine equipment maintenance.
	Future Directions	Improve capacitive accuracy; simulate mixed signals; integrate image recognition.
10	Research Objectives	Design a Wheatstone bridge sensor with temperature compensation for ferromagnetic debris detection.	[23]
	Key Findings	• Detects 72 μm iron particles within 20–70 °C. • Reduces temperature drift.
	Applications and Innovations	High-temperature oil monitoring for ships and machinery.
	Future Directions	Improve detection of non-ferrous particles; address debris shape identification.

and

Table 3. Methodological overview of impedance-based techniques for lubricant condition monitoring.

No.	Details of	Description	Ref.
1	Impedance Parameter	Complex impedance, dielectric constant	[14]
	Measurement Technique	EIS (0.01–100 Hz) using gold electrodes; dielectric constant calculated from impedance.
	Key Technical Specifications	Electrodes with an area of 25 cm2, 0.8 mm gap and 200 mV RMS voltage.
	Contaminant Correlation	Linear correlation between water content (0.01–0.05% vol) and impedance/dielectric constant.
2	Impedance Parameter	Impedance magnitude	[15]
	Measurement Technique	Impedance spectroscopy (0.2–2000 kHz) performed on fresh and used oils.
	Key Technical Specifications	Frequency range of 0.2–2000 kHz; four-ball tribometer employed for wear testing.
	Contaminant Correlation	Metal particles increase conductivity, resulting in lower impedance in used oils.
3	Impedance Parameter	Inductance and resistance.	[16]
	Measurement Technique	Double-wire solenoid coil with 2 MHz excitation; LCR meter used to detect impedance changes.
	Key Technical Specifications	Microchannel with a 500 μm diameter; optimal frequency at 2 MHz.
	Contaminant Correlation	Differentiation between ferromagnetic (iron) and non-ferromagnetic (copper) particles via pulse direction and amplitude.
4	Impedance Parameter	Inductance and resistance.	[17]
	Measurement Technique	High-gradient magnetic field sensor with dual-parameter detection.
	Key Technical Specifications	Square channel design; excitation provided by silicon steel strips.
	Contaminant Correlation	Particle size and material are linked to the amplitude of the inductance/resistance pulses.
5	Impedance Parameter	Real and imaginary impedance components.	[18]
	Measurement Technique	Cylindrical coil integrated with an LCR meter and microfluidic chip for particle flow.
	Key Technical Specifications	Flow rate of 0.6 mL/min; elliptical function model established for size correlation.
	Contaminant Correlation	Particle size estimation independent of material (validated with copper and aluminum particles).
6	Impedance Parameter	Inductance and resistance.	[19]
	Measurement Technique	Sensor enhanced with silicon steel strips and double rectangular channels.
	Key Technical Specifications	Series-connected coils to optimize the signal-to-noise ratio (SNR).
	Contaminant Correlation	Differentiation between ferromagnetic (46 μm iron) and non-ferromagnetic (110 μm copper) particles.
7	Impedance Parameter	Inductance and resistance.	[20]
	Measurement Technique	Parallel resonance circuit with adjustable capacitance, monitored via LCR measurements.
	Key Technical Specifications	Optimal capacitance values: 1.45 nF for ferrous and 1.50 nF for non-ferrous particles.
	Contaminant Correlation	Capacitance optimization enhances sensitivity for both particle types.
8	Impedance Parameter	Increments in inductance/resistance.	[21]
	Measurement Technique	Inductive sensor with AC excitation supported by COMSOL modeling.
	Key Technical Specifications	Coil with a 3.5 mm inner diameter and 150 turns.
	Contaminant Correlation	Impedance increments correlate with debris size (30–180 μm) and material (iron/copper).
9	Impedance Parameter	Combined inductive and capacitive signals.	[22]
	Measurement Technique	Dual inductive–capacitive detection method with finite-element simulation.
	Key Technical Specifications	Differentiation of mixed particle signals achieved via dual detection modes.
	Contaminant Correlation	Enables material differentiation (ferrous versus non-ferrous) using combined responses.
10	Impedance Parameter	Inductance measured using a Wheatstone bridge.	[23]
	Measurement Technique	Two-wire helical coils with a conditioning circuit; analysis supported by COMSOL simulation.
	Key Technical Specifications	Excitation frequency of 10 kHz; operating temperature range from 20 to 70 °C.
	Contaminant Correlation	Capable of detecting ferromagnetic particles (72–400 μm iron) despite temperature variations.

) advance electrical sensing methodologies for lubricant and oil condition monitoring, each addressing distinct aspects through impedance, inductance, and capacitance techniques. For instance, one investigation [15] demonstrates that impedance spectroscopy is a robust tool for evaluating vegetable oils (e.g., sesame and almond) compared to mineral oils in tribological tests, revealing that almond oil exhibits superior lubricity and that impedance measurements can effectively differentiate oil types and degradation states. In a complementary study [14], the precision of impedance spectroscopy in detecting ultralow water content (less than 0.05% vol) in mineral oil is showcased by proposing an in situ system with linear calibration, albeit with limitations imposed by electrode geometry. Equivalent circuit modeling represents the oil’s electrical behavior using resistors (resistance to current), capacitors (energy storage), and inductors (magnetic field interactions). For instance, degraded oil with water contamination may show lower impedance, modeled as a reduced resistance in the circuit [14].

Wear debris are microscopic metal particles (e.g., iron, copper) shed from engine components. Their concentration and size in oil indicate mechanical wear severity. Regarding wear debris detection, several investigations [16,17,19,23] introduce innovative sensor designs. A double-wire solenoid coil sensor, as reported in one study [16], detects sub-100 μm particles via inductance-resistance signals; another study [17] employs a square-channel sensor to enhance throughput, while yet another investigation [19] utilizes silicon steel strips in a dual-channel configuration for real-time, multi-parameter detection. Signal-to-noise ratio (SNR) is a measure of signal strength relative to background noise. High SNR ensures reliable data, critical for detecting subtle changes in oil condition. Additionally, one research work [23] describes a Wheatstone bridge with temperature compensation, although its application is limited to ferrous particles. Theoretical contributions include an elliptical model for material-agnostic sizing of non-ferrous particles presented in one study [18] and an eddy current-based impedance model validated experimentally in another investigation [21], though the latter is challenged by the irregular shapes of debris. Eddy currents are circular electric currents induced in conductive materials (e.g., metal debris) by a changing magnetic field. These currents alter the sensor’s impedance, enabling detection and sizing of wear particles.

Moreover, one study [20] optimizes resonance circuits through capacitance tuning, achieving dual sensitivity for both ferrous and non-ferrous particles, albeit requiring separate systems. Another investigation [22] explores an inductive–capacitive dual detection method for mixed particles, but its performance is hindered by low capacitive accuracy. Collectively, these studies highlight trends toward multi-parameter sensing, enhanced environmental robustness (e.g., temperature compensation as observed in [23]), and sensor miniaturization, while also addressing challenges such as debris shape variability, standardization requirements, and interference mitigation.

This work highlights significant advancements in using electrical impedance and inductive sensing for oil condition monitoring, particularly for detecting wear debris and assessing oil degradation. However, several gaps and challenges remain. A primary challenge across studies is the detection of very small wear particles, especially non-ferromagnetic ones. While some research demonstrates the detection of iron particles as small as 20 μm and copper particles as small as 60 μm [16], and even 25 μm iron and 85 μm copper [17], further improvements in sensitivity are needed, particularly for non-ferrous particles. The irregular shape of actual wear debris also poses a challenge, as current models often assume spherical particles, which can affect the accuracy of size and material estimation [18,21]. Signal noise interference is another significant hurdle, as it can mask the detection of small debris, necessitating the development of robust filtering and signal processing techniques [17]. Furthermore, temperature variations significantly impact sensor performance [21], and while some studies propose temperature compensation methods [23], ensuring reliable detection across a wide range of operating temperatures remains crucial. The complexity of fabricating highly sensitive sensors, often requiring external circuits or magnetic materials, presents another practical challenge [16]. For impedance spectroscopy, a limitation is that a single spectrum can correspond to multiple electrical circuits, although this is less of an issue when testing the same substance with a single variable [14]. For multi-parameter detection methods, the need for sufficient standard data for comparison is a challenge for accurate differentiation [22]. Developing in situ, cost-effective, and fast impedance-based systems for specific applications like diesel engine oil monitoring is a future direction [14]. Future directions emphasize the development of hybrid algorithms (e.g., incorporating image recognition as suggested in [22]), improved shielding strategies [16], and broader applicability in harsh environments (e.g., marine systems as noted in [21]). Ultimately, the goal is to transition these promising laboratory techniques to practical, real-world applications for machinery health monitoring and hydraulic equipment, requiring further validation in actual oil samples [20]. Together, these advancements pave the way for real-time, predictive maintenance in industrial and automotive applications. While impedance spectroscopy appeared in detecting contaminants like water and wear debris, dielectric spectroscopy offers distinct advantages in tracking chemical degradation, as explored in the following (Section 4).

4. Diesel Engine Lubricant Condition Monitoring: Dielectric Spectroscopy Applications

Table 4. Overview of dielectric-based approaches for oil condition monitoring.

No.	Overview of	Description	Ref.
1	Research Objectives	Detect gasoline contamination using THz dielectric properties.	[24]
	Key Findings	- Absorption coefficient increases with contamination (error: 0.21–1.2%). - Refractive index decreases with contamination. - 0.5 THz identified as the most effective frequency.
	Applications and Innovations	- Fixed THz setups for real-time monitoring. - Distinguishes contamination levels from 0% to 12%.
	Future Directions	Develop miniaturized THz systems for engine applications and enable multi-contaminant detection.
2	Research Objectives	Develop a recycled capacitive sensor for monitoring oil degradation.	[25]
	Key Findings	- Linear correlation between permittivity and mileage (R2 = 0.98). - Identified a critical permittivity threshold of 2.73.
	Applications and Innovations	- Sustainable aluminum heat sink sensor. - Correlation validated against FT-IR (ASTM E-2412).
	Future Directions	Integrate the sensor into engine crankcases.
3	Research Objectives	Validate dielectric monitoring in hydraulic systems.	[26]
	Key Findings	- Increase in dielectric constant correlates with varnish formation (MPC test). - Viscosity measurement error within 5%.
	Applications and Innovations	- Multi-parameter tuning fork sensor. - Capable of detecting sudden contamination events.
	Future Directions	-

and

Table 5. Methodology details: articles focus on dielectric.

No.	Details of	Description	Ref.
1	Dielectric Parameter	Refractive index (n), absorption coefficient (α)	[24]
	Measurement Technique	Terahertz time-domain spectroscopy (THz-TDS)
	Key Technical Specifications	- Frequency range: 0.5–2.5 THz - 1550 nm laser source - Transmission mode with cuvette-based sampling
	Contaminant Correlation	Absorption at 0.5 THz predicts gasoline contamination levels (RMSE = 1.2%)
2	Dielectric Parameter	Permittivity (ε)	[25]
	Measurement Technique	Capacitive sensing
	Key Technical Specifications	- Operating frequency: 1 MHz - Parallel plate design (2.5 mm gap) - Electrodes made from recycled aluminum
	Contaminant Correlation	Linear correlation with oxidation and sulfation (R2 = 0.98)
3	Dielectric Parameter	Dielectric constant	[26]
	Measurement Technique	Tuning fork sensor
	Key Technical Specifications	- Simultaneous viscosity and dielectric measurement - Field calibration over 12 months
	Contaminant Correlation	A 0.1-unit increase in ε corresponds to a 5% viscosity change

summarize studies that explore innovative sensor technologies for oil condition monitoring, each employing distinct methodologies and applications. One investigation [24] focuses on terahertz time-domain spectroscopy (THz-TDS) to detect gasoline contamination in engine oil. Terahertz time-domain spectroscopy (THz-TDS) uses terahertz radiation (0.1–10 THz) to probe molecular interactions in oil. This study revealed statistically significant trends (p < 0.0001) in both the refractive index and absorption coefficient, with absorption at 0.5 THz achieving high predictive accuracy (R² = 0.93–0.998). However, its applicability is limited to SAE 5W-20 oil and contamination levels of 12% or lower. To better understand the information in the tables, root mean square error (RMSE) is a statistical metric quantifying prediction accuracy. The proximity of this criterion to zero is more desirable. Another study [25] introduces a cost-effective capacitive sensor fabricated from recycled aluminum to assess engine oil degradation via permittivity measurements, which were validated against FT-IR spectroscopy. This sensor demonstrated a strong linear correlation (R² > 0.98) between permittivity and simulated mileage, establishing a critical threshold (ε = 2.73) for oil lifespan. In addition, a separate investigation [26] evaluated a tuning fork sensor for hydraulic oil in machinery. This sensor effectively tracked changes in the dielectric constant and viscosity over a 12-month period, accurately detecting contamination events and correlating well with laboratory analyses (error < 5%). Nonetheless, its exclusion of diesel engine oils limits its broader applicability.

A systematic review of these studies (Table 4 and Table 5) highlights the feasibility of dielectric and spectroscopic techniques for oil quality assessment, offering a balance between cost-effectiveness [25], measurement precision [24], and real-time applicability in field conditions [26]. Notable advancements include the integration of recycled materials for sustainable sensor development [25], enhanced discrimination of multiple contaminants through spectroscopic analysis [24], and multi-parameter sensing capabilities for comprehensive oil monitoring [26]. However, ref. [24] achieves high accuracy (R² = 0.93–0.998) in controlled lab conditions; its applicability is restricted to fresh oils and narrow contamination ranges (0–12%), neglecting real-world factors like oxidation and multi-contaminant interactions. Similarly, ref. [25] demonstrates cost-effective permittivity measurements but relies on simulated degradation, overlooking thermal oxidation effects and requiring field validation for engine integration. Study [26], though effective for hydraulic oils, lacks diesel-specific correlations (e.g., soot, acid formation) and standardized metrics (TBN, FT-IR). Future research should focus on sensor miniaturization, expanding contaminant detection capabilities, and bridging the gap between laboratory-grade precision and the robustness required for industrial applications. It is clear that lubricant monitoring (whether impedance-based, dielectric-based, or other methods) involves complex datasets that require advanced analytical tools. Some of these studies are reported in Section 5.

5. Condition Monitoring of Diesel Engine Lubricant: Soft Computing Applications

The systematic review of soft computing applications in diesel engine oil condition monitoring (see

Table 6. Key contributions of soft computing in engine health monitoring.

No.	Overview of	Description	Ref.
1	Research Objectives	Optimization of preventive maintenance intervals using MLP and RBF neural networks.	[27]
	Key Findings	MLP and RBF networks were used to predict oil operating time. Key oil indicators (Na, Fe, Mo, and soot) were identified, and a soft failure threshold was observed at approximately 200 h of operation.
	Applications and Innovations	Enables predictive maintenance and cost savings through optimized oil change intervals.
	Future Directions	Not specified.
2	Research Objectives	Detection of acid, base, and water content in oil using GRNN.	[28]
	Key Findings	GRNN achieved errors of 6.9%, 4.2%, and 15.7% for acid, base, and water content predictions, respectively. The approach was effective even with small training datasets.
	Applications and Innovations	Enables real-time oil condition monitoring with simultaneous detection of multiple properties.
	Future Directions	Incorporation of temperature compensation; expansion of detectable properties (e.g., soot, sulfur).
3	Research Objectives	Development of an SVR model optimized with PSO for engine cylinder liner wear assessment.	[29]
	Key Findings	SVR model parameters were optimized using PSO. The model was validated against a BPNN using MSE and R2 metrics.
	Applications and Innovations	Enables modeling of engine wear under uncertain conditions.
	Future Directions	Not specified.
4	Research Objectives	Comparison of KNN and RBF-NN for wear detection.	[30]
	Key Findings	RBF-NN outperformed KNN, achieving 99.85% classification accuracy. Fe, Cr, Pb, and Cu were identified as key wear indicators.
	Applications and Innovations	Supports maintenance optimization and reduces testing costs by focusing on key indicators.
	Future Directions	Not specified.
5	Research Objectives	Assessment of the impact of metal contamination on engine health.	[31]
	Key Findings	RBF-NN achieved 99% classification accuracy. Fe, Cr, and Si were identified as critical contaminants.
	Applications and Innovations	Enables early fault detection and diagnosis of critical engine conditions.
	Future Directions	Application of these methods to other industrial components; comparison of different soft computing approaches.
6	Research Objectives	Development of an anomaly detection framework based on LSTM-SVDD for oil condition time-series data.	[32]
	Key Findings	The LSTM-SVDD model outperformed an LSTM-PCA-SPE approach. It achieved high prediction accuracy with a low RMSE.
	Applications and Innovations	Enables real-time oil condition monitoring and predictive maintenance.
	Future Directions	Optimization of the LSTM-SVDD model using heuristic algorithms.
7	Research Objectives	Comparative study of SVM, ANFIS, GPR, and other models for oil viscosity prediction.	[33]
	Key Findings	The RBF model exhibited the best performance (RMSE = 0.11; EF = 1.0) and demonstrated high generalizability across different training set sizes.
	Applications and Innovations	Eliminates the need for direct viscosity testing, enabling cost-effective condition monitoring.
	Future Directions	Expansion of datasets; inclusion of additional parameters (e.g., temperature).
8	Research Objectives	Comparative evaluation of ANFIS and MLP-NN performance.	[34]
	Key Findings	ANFIS achieved 94% classification accuracy, outperforming the MLP-NN model. Fe, Cr, Pb, and PQ were identified as key health indicators.
	Applications and Innovations	Reduces costs by eliminating redundant indicators.
	Future Directions	Integration of real-time data; development of user-friendly interfaces.

and

Table 7. Methodological overview of soft computing approaches for oil condition monitoring.

No.	Details of	Description	Ref.
1	Soft Computing Parameter	Particle concentrations (Fe, Na, Mo)	[27]
	Measurement Technique	MLP/RBF neural networks trained on over 350 oil samples.
	Key Technical Specifications	Ten-year dataset; AES/FTIR/LNF analysis.
	Contaminant Correlation	Correlates metal particles, soot, and additives with oil degradation.
2	Soft Computing Parameter	Capacitance (acid/base/water)	[28]
	Measurement Technique	GRNN with a capacitive sensor array featuring polyimide, PTFE, and Nafion coatings.
	Key Technical Specifications	Forty-eight samples; FFT noise filtering.
	Contaminant Correlation	Linear correlation between capacitance and acid/base/water content.
3	Soft Computing Parameter	Wear capacity	[29]
	Measurement Technique	SVR with PSO optimization combined with fuzzy membership functions.
	Key Technical Specifications	Gaussian RBF kernel; validated using MSE/R2 metrics.
	Contaminant Correlation	Links wear data to engine cylinder liner degradation.
4	Soft Computing Parameter	Metal concentrations (Fe, Cr, Pb)	[30]
	Measurement Technique	KNN and RBF-ANN classifiers applied to 681 oil samples.
	Key Technical Specifications	Seven key indicators; 40–80% training splits.
	Contaminant Correlation	Correlates Fe, Cr, and Cu with wear, viscosity, and overall contamination.
5	Soft Computing Parameter	Metal contaminants (Fe, Cr, Si)	[31]
	Measurement Technique	RBF-NN/SVM classifiers applied to 1948 oil samples.
	Key Technical Specifications	ASTM-standard measurements; sensitivity analysis.
	Contaminant Correlation	High impact of Fe, Cr, and Si on engine health classification.
6	Soft Computing Parameter	Time-series oil metrics (e.g., viscosity)	[32]
	Measurement Technique	LSTM for trend prediction combined with SVDD for anomaly detection.
	Key Technical Specifications	A total of 1440 × 9 daily data points; grid search optimization.
	Contaminant Correlation	Correlates particle counts, water content, and dielectric constant with anomalies.
7	Soft Computing Parameter	Viscosity prediction	[33]
	Measurement Technique	Comparison of an RBF model versus SVM, ANFIS, and GPR on 555 oil samples.
	Key Technical Specifications	Data rescaled to −1 to 1; 35-network RBF topology.
	Contaminant Correlation	Links metallic and nonmetallic elements (e.g., Fe, Al) to viscosity changes.
8	Soft Computing Parameter	Metal contaminants (Fe, Cr, PQ)	[34]
	Measurement Technique	MLP-NN (using 13 training algorithms) versus ANFIS (employing clustering methods).
	Key Technical Specifications	A total of 1948 samples; LSD method for feature reduction.
	Contaminant Correlation	Correlates Fe, Cr, and PQ with critical engine states.

) highlights the increasing adoption of advanced computational techniques for predictive maintenance and operational optimization. Studies such as [27,31,34] utilize artificial neural networks (MLP, RBF, LSTM) and hybrid models (e.g., LSTM-SVDD, ANFIS) to analyze oil degradation patterns based on datasets obtained from spectral analysis (AES, FTIR), sensor arrays, and real-time monitoring systems. These models effectively predict key indicators, including oil lifespan, wear particle concentrations (Fe, Cr, Na), and viscosity fluctuations, achieving classification accuracies of up to 99% in distinguishing engine health states (normal, caution, critical). Artificial neural networks (ANNs) are computational models inspired by biological neurons, trained to recognize patterns in data. Long short-term memory (LSTM) is a recurrent neural network (RNN) variant effective for analyzing time-series data (e.g., tracking oil degradation trends). Support vector machines (SVMs) are supervised learning models that classify data by finding optimal hyperplanes (e.g., distinguishing healthy vs. degraded oil states). Adaptive neuro-fuzzy inference system (ANFIS) is a hybrid model combining fuzzy logic (handling vague data, e.g., “high” vs. “low” contamination) and neural networks (learning complex patterns). ANFIS improves classification accuracy in noisy environments.

Key advancements in this domain include the identification of cost-efficient diagnostic markers, such as sodium and soot levels, and the integration of fuzzy logic for managing data uncertainties, as demonstrated in [29,32]. Advancements, including real-time capacitive sensors [28] and adaptive algorithms such as the RBF model [33], indicate promising developments in dynamic maintenance scheduling.

The integration of soft computing techniques, such as neural networks and machine learning algorithms, has shown significant potential in advancing diesel engine lubricant condition monitoring. However, several challenges must be addressed to fully leverage these technologies. Key gaps include data uncertainty due to variability in oil samples and limitations in laboratory instrumentation [27]. Sensor noise and the need for optimal sensor design remain critical challenges, as highlighted in studies utilizing capacitive sensor arrays [28]. Future research will focus on enhancing model resilience through multimodal artificial intelligence approaches, incorporating temperature compensation mechanisms, and expanding datasets to improve predictive accuracy. Collectively, these studies underscore the transformative role of soft computing in shifting maintenance strategies from reactive to predictive paradigms, thereby offering substantial cost reductions and improved reliability for industrial applications. The standalone strengths of soft computing, impedance, and dielectric techniques are further amplified when integrated, as demonstrated in hybrid approaches (Section 6).

6. Condition Monitoring of Diesel Engine Lubricant: Multiple Simultaneous Applications

In this section, dielectric/impedance sensors collect real-time electrical data (e.g., permittivity, conductivity). Machine learning models (e.g., ANNs) then process these data to predict oil lifespan, detect contaminants, or classify engine health. A recent study [35] investigated the potential of integrating dielectric and impedance-based parameters to assess the quality of fresh synthetic engine oils, focusing specifically on 5W30-grade oils from various brands. By analyzing electrical properties such as capacitance, impedance magnitude, and quality factor over a frequency range of 100 Hz to 1.2 MHz, the study identified capacitance as a robust, frequency-independent diagnostic marker—with a low variability (6% coefficient of variation)—making it suitable for broad-range applications. Additionally, impedance magnitude and quality factor emerged as critical indicators within specific frequency bands (4000 Hz–0.01 MHz and 2000 Hz–0.01 MHz, respectively), reflecting their sensitivity to changes in oil composition and additive interactions.

The dielectric behavior of the oils, measured under controlled conditions using precision LCR techniques, demonstrated superior selectivity compared to conventional TAN/TBN analyses, which are limited by their dependence on polar degradation byproducts. Moreover, statistical clustering and regression analyses further validated the discriminative power of these electrical parameters, underscoring their potential for real-time sensor development in automotive applications. Despite these promising results, the study also highlighted significant challenges, such as the need to validate findings in degraded oils and to establish explicit correlations between electrical signatures and chemical degradation markers. Future research should address these gaps to facilitate the translation of dielectric and impedance diagnostics into practical, algorithm-driven sensor technologies for dynamic engine environments [35].

Table 8. Overview of combined dielectric/impedance and soft computing approaches for oil condition monitoring.

No.	Overview of	Description	Ref.
1	Research Objectives	Develop thick-film potentiometric sensors for oil acidity and compare with AN and impedance spectroscopy.	[36]
	Key Findings	• Oil conductivity increases with oxidation. • Sensors detect acidity linearly (0–25 mg KOH/g). • Temperature affects sensor lifetime.
	Applications and Innovations	• Real-time oil acidity monitoring. • Use of impedance spectroscopy to track oxidation stages.
	Future Directions	• Optimize sensors for formulated oils. • Integrate with other monitoring techniques. • Conduct field trials in engines.
2	Research Objectives	Use dielectric properties to monitor oil contamination and develop a low-cost sensor via ANN.	[13]
	Key Findings	• Dielectric properties correlate with contaminants (e.g., Fe, Pb). • ANN achieves strong correlation at 7.40 GHz.
	Applications and Innovations	• Microwave-based real-time sensors. • ANN for contamination prediction.
	Future Directions	• Develop comprehensive multi-contaminant sensors. • Validate sensor performance in real engine conditions.
3	Research Objectives	Develop a capacitive sensor array integrated with ANN for multi-property lubricant monitoring.	[37]
	Key Findings	• Accurate measurement of water, diesel, base, and soot. • Temperature compensation achieved via thermocouple.
	Applications and Innovations	• Online diagnostics for diesel engines. • Enhanced accuracy through deep learning.
	Future Directions	-
4	Research Objectives	Evaluate soft computing models (RBF, ANFIS) to predict contaminants from dielectric properties.	[38]
	Key Findings	• RBF/ANFIS outperform MLP/SVM. • Higher frequencies (7.4 GHz) improve accuracy.
	Applications and Innovations	• Real-world engine oil analysis. • Comprehensive multi-element prediction.
	Future Directions	• Collaborate for diverse datasets. • Explore deep learning and optimization techniques.
5	Research Objectives	Predict pollutants (Fe, Cr, Pb, etc.) via dielectric properties using machine learning.	[39]
	Key Findings	• The RBF model achieves an RMSE of 0.01 and R2 > 0.99. • The model generalizes well across training ratios.
	Applications and Innovations	• Dielectric-based oil quality monitoring. • ML-enhanced predictive maintenance.
	Future Directions	• Validate with real-world data. • Incorporate fuzzy logic for enhanced robustness.
6	Research Objectives	Classify cross-contaminants in aviation oil using impedance measurements combined with ML.	[40]
	Key Findings	• NN classifier achieves 99.8% accuracy. • Impedance spectra outperform tan δ measurements.
	Applications and Innovations	• Real-time contamination detection. • Data augmentation for small datasets.
	Future Directions	• Refine transient regime analysis.

and

Table 9. Overview of combined dielectric/impedance and soft computing methodologies.

No.	Details of	Description	Ref.
1	Dielectric/Impedance Parameter	Impedance spectroscopy for conductivity	[36]
	Soft Computing Details	Comparison with AN using the Kittiwake test and other metrics.
	Measurement Technique	Thick-film sensor fabrication; EIS using a Solartron analyzer.
	Key Technical Specifications	Measurements conducted at 50 °C and 80 °C.
	Contaminant Correlation	Increased conductivity correlates with oxidation stages; acidity correlates with oil condition.
2	Dielectric/Impedance Parameter	Dielectric constant (ε′) and loss factor (ε″)	[13]
	Soft Computing Details	ANN models the relationships between dielectric properties and contaminants.
	Measurement Technique	Dielectric probe kit combined with a vector network analyzer.
	Key Technical Specifications	Measurements at frequencies of 2.40 GHz, 5.80 GHz, etc.
	Contaminant Correlation	Established correlation between dielectric properties and contaminants such as wear metals.
3	Dielectric/Impedance Parameter	Dielectric properties (capacitance)	[37]
	Soft Computing Details	ANN tuned with stochastic global optimization (SGO) for accurate measurements.
	Measurement Technique	Interdigital capacitive sensor array paired with a thermocouple.
	Key Technical Specifications	Sixty-four samples; 10W-30 oil; temperature range: 90–110 °C.
	Contaminant Correlation	Measured dielectric variations correlate with contaminant concentrations.
4	Dielectric/Impedance Parameter	ε′, ε″, and tan δ at 2.4, 5.8, and 7.4 GHz	[38]
	Soft Computing Details	Multiple soft computing algorithms (RBF, ANFIS, SVM) developed and trained.
	Measurement Technique	Vector network analyzer for electrical property measurements at multiple frequencies.
	Key Technical Specifications	A total of 49 samples (33 existing + 16 new); measurements taken at 2.4, 5.8, and 7.4 GHz.
	Contaminant Correlation	Established relationships between electrical properties and various contaminants (e.g., Fe, Pb, Cu).
5	Dielectric/Impedance Parameter	ε′, ε″, and tan δ (2–8 GHz)	[39]
	Soft Computing Details	Evaluation of several soft computing models, particularly RBF, for pollutant prediction.
	Measurement Technique	Dielectric probe combined with a microwave analyzer.
	Key Technical Specifications	Seventy samples; Box–Behnken design; five measurements per sample.
	Contaminant Correlation	Correlation between dielectric properties and the presence of metals (e.g., Fe, Cr, Pb).
6	Dielectric/Impedance Parameter	Real and imaginary impedance, tan δ	[40]
	Soft Computing Details	Three classifiers (1-NN, SVM, EBT) trained for real-time classification.
	Measurement Technique	Microfabricated sensor coupled with impedance spectroscopy.
	Key Technical Specifications	Sixteen contamination classes; measurements at room temperature; aged oil.
	Contaminant Correlation	Cross-contamination levels establish correlations between impedance and contaminant levels.

illustrate the concurrent application of dielectric or impedance measurements integrated with soft computing tools. The reviewed articles emphasize a growing trend toward combining these electrical measurement techniques with advanced computational methods for real-time oil condition monitoring across diverse industrial applications. For example, studies [36,40] employ impedance spectroscopy to correlate oil conductivity with oxidation levels [36] or to classify cross-contaminants such as water and fuel in aviation oil [40], achieving high accuracy (up to 99.8%) with machine learning classifiers like 1-NN and SVM. In parallel, dielectric properties—including permittivity (ε′, ε″) and loss tangent (tan δ)—are extensively investigated in studies [13,37,38,39], where they are linked to contaminant concentrations (e.g., Fe, Pb, Cu) in lubricants. These investigations utilize artificial neural networks (ANNs), radial basis function (RBF) models, and support vector machines (SVMs) to decode the complex relationships between dielectric responses and oil degradation, with RBF models demonstrating superior performance (RMSE < 0.01, R² > 0.99). Innovations such as temperature-compensated interdigital capacitive sensor arrays [37] and thick-film potentiometric sensors [36] address challenges including thermal drift and cross-sensitivity, while soft computing techniques enable multi-property prediction (e.g., water, soot, base levels) and fault diagnosis.

Finally, based on the studies presented in Table 8 and Table 9, despite advancements, key limitations persist, including temperature sensitivity [36,37], limited generalizability across oil types [35], and small sample sizes [39]. Future research should prioritize field validation of sensors in degraded oils [35], integration of fuzzy logic to handle measurement uncertainties [39], and development of hybrid models (e.g., RNNs) for temporal analysis [40]. Disposable sensor technologies could enhance industrial adoption, while collaborative efforts to expand datasets [39] and optimize hyperparameters [38] will strengthen predictive accuracy. Collectively, these innovations underscore the transition toward AI-driven, multi-modal systems capable of real-time diagnostics, reducing downtime, and advancing predictive maintenance paradigms. These approaches offer scalable solutions for industries ranging from aviation to heavy machinery, although further field validation and sensor integration remain critical for broader industrial adoption.

7. Conclusions and Outlook

Recent advancements in impedance, dielectric, and soft computing techniques have significantly enhanced the capability to monitor lubricant conditions in diesel engines, enabling a shift from reactive to predictive maintenance strategies. Our comprehensive review of 28 studies highlights a range of innovative approaches—including thick-film potentiometric sensors, impedance spectroscopy, and machine learning models—that demonstrate strong correlations between oil properties, wear debris, and engine health indicators. These findings underscore the potential of data-driven sensor designs to detect oil degradation early, thereby enabling timely maintenance interventions.

The application of soft computing methods, such as artificial neural networks and support vector machines, has proven effective in accurately forecasting lubricant properties and classifying engine health states based on key markers. However, challenges remain, including the inherent complexity of oil chemistry, the need for further validation under diverse operating conditions, and the difficulty of standardizing sensors and data interpretation protocols for seamless integration into current maintenance practices. On this basis, future research should prioritize the following three key directions:

• Enhanced Sensor Design: Miniaturized, cost-effective sensors with improved shielding and multi-modal detection capabilities (e.g., inductive–capacitive hybrids) are needed to address sensitivity and environmental robustness. Integrating image recognition or advanced signal processing could mitigate challenges posed by irregular debris shapes and noise.
• AI-Driven Multi-Modal Systems: Combining impedance, dielectric, and spectroscopic data with adaptive AI models (e.g., fuzzy logic, recurrent neural networks) will improve predictive accuracy and handle measurement uncertainties. Collaborative efforts to expand datasets and optimize hyperparameters are essential for generalizable solutions.
• Field Validation and Standardization: Translating laboratory success to industrial applications requires rigorous validation in real-world environments, particularly for degraded oils and multi-contaminant scenarios. Establishing standardized metrics for electrical properties and degradation thresholds will bridge the gap between research and practical implementation.

These improvements are expected to reduce operational costs by extending oil change intervals and minimizing unscheduled downtime, while also promoting environmentally friendly practices through improved lubricant management. As these technologies continue to evolve, they will play a critical role in advancing predictive maintenance and enhancing overall engine performance in automotive, marine, and aerospace sectors. The transition to predictive maintenance paradigms promises not only economic benefits but also environmental sustainability through optimized lubricant management.