CONDITION MONITORING WITH PREDICTION BASED ON OIL ENGINES OF URBAN BUSES – A CASE STUDY

The paper presents a case study and a model for condition monitoring of Diesel engines’ oil of urban buses, through the accompaniment of the evolution of its degradation, with the objective to implement a predictive maintenance policy. Along time, because the usage, there is some decay in the lubricant properties. However, in normal functioning conditions, the lubricants properties, at the time the manufacturers recommend its changing, regardless of they are within the safety limits. Then, based on the accompaniment of the lubricants’ oil condition, the intervals of oil replacement can be enlarged what implies the availability increasing and the corresponding production increasing of the equipment. The model presented in this paper shows its potential to be spread to other types of equipment and organisations that want can implement similar maintenance policies, to achieve the best availability based on the real equipment health conditioning conditions.


INTRODUCTION
Public transport, in general, and the city bus passenger transport, in particular, represents an important alternative to the use of individual transport.For this reason, it is essential to focus on the quality of service provided by the public transport network in order to make them attractive for its users.
Under the current context, where the users of public transport are increasingly demanding about the quality of services, the maintenance stands out as a competitiveness key factor.This paper presents an approach to the analysis of lubricating oils for Diesel engines, validated on some models of urban buses, as is described along the paper.
The condition monitoring maintenance appeared in the 70-80's to designating a new approach to planned maintenance based on the knowledge of the state of equipment, using condition monitoring techniques, [1].
An integrating approach of the Life Cycle of a Physical asset can be seen in Farinha (2018), namely including management standards, like ISO 55000X interconnected with some econometric models to evaluate their Life Cycle Costs [8].
There are several approaches to forecast the evolution of oil degradation.Newell discusses an approach based on trend analysis to maximize oil change intervals.This author considers the following common oil analysis tests and procedures: viscosity; total acid number (TAN); total base number (TBN); water content; specific gravity; particle count (visual method); spectrometric analysis; ferrographic analysis, [9].Macián et al. present an analytical approach aiming at more accurate wear determination from engine oil samples.The authors ask: "What level of wear rate is normal or abnormal for the engine studied?"To answer the question, they propose a comparative parameter Z; it takes into account both the deviation from the engine reference wear rate and the reference wear rate.The parameter also compares the current situation to the behavior of a larger population, such as all engines of the same model.The authors demonstrate the influence of oil consumption and additions to the contaminant concentration [10].
Vališ, Zák, and Pokora concentrate on metal particles, such as iron (Fe) and lead (Pb), as potential failure indicators.They apply a linear regression model to determine a linear course of Fe and Pb particle generation.They assume a stochastic process with time dependence.The authors conclude by the importance of time series comparisons: Auto Regression Integrating Moving Average (ARIMA); Auto Regression Moving Average (ARMA) methods [11].Changsong et al. present a study based on 50 oil samples collected and analyzed in sequence covering 250 motor hours.The results show that maintenance intervals can be longer and, at the same time, the cost-effectiveness maintenance ratio can be increased [12].
Macián et al. say that Low Viscosity Engine oils (LVO) are considered key contributors for improving fuel economy in Internal Combustion Engines (ICE).Attending that the use of LVO could imply a variation in tribological states found in ICE; this work's aim was to test LVO in real fleet, with emphasis on engine wear and oil Key Performance Indicators (KPI).This test comprised 39 buses, two engine technologies and four different lubricants [13].
Macián shows the results of a comparative fleet test, which the main objective was to measure the influence of Low Viscosity Oils (LVO) over the fuel consumption and CO2 emissions of urban buses.To perform this test, 39 urban buses, classified into candidate and reference groups depending on the engine oil viscosity, covered a 60,000 km mileage corresponding to two rounds of standard Oil Drain Interval (ODI).In the same way, for 9 buses of the 39 buses, the effect of differential LVO over fuel consumption and their interaction with engine LVO was assessed during the second ODI.Test results confirm that the use of LVO could reduce fuel consumption, hence CO2 emissions.However, special attention should be taken prior to its implementation in a fleet, particularly if the vehicles powered by engines with high mechanical and thermal stresses during vehicle operation because this could lead to friction loss increase, loss of the potential fuel consumption reduction of LVO and, in the worst scenario, higher rates of engine wear [14].
Results have shown that LVO presented an excellent performance along the ODI, even improving some characteristics of the baseline oils with higher viscosity values.Those results have shown that oil degradation is more dependent on engine technology, but in any case presented a penalization in terms of ODI reduction, a key indicator for end-users related to maintenance costs.In the case of Compressed Natural Gas engines, higher oil degradation in terms of oil oxidation and nitration was observed [15].
The increasingly stringent emissions regulations and the climatic change as consequence of greenhouse gases emissions have moved the research interest towards the optimization of the internal combustion engine.Hence, the reduction of the energy losses of the engine sub-systems like friction and parasitic energy consumption are of especial interest.Tormos et al. [16] present a model to determine the friction losses and the engine accessories energy consumption is developed based on parameters obtained in standard test benches.A description of the models to estimate friction in the piston assembly, bearings and valve train, and energy consumption of the coolant, oil and fuel pump is provided.Finally, a brief application to demonstrate the model potential in diagnosis and predictive applications is discussed [16].
Although not the topic of this paper, multivariate statistics is also useful, because oil analysis involves several variables; an approach like this may help diagnosis of the health of diesel engines.We refer some additional references: [17], [18] and [19].

CONDITION MONITORING WITH PREDICTION THROUGH OIL ANALYSIS
The main physical feature of lubricating oil is viscosity and its variation with temperature, given by the viscosity index and the density.
In recent years, there has been a demand for high-performance engine lubricants, especially in the aerospace and automotive industries.This has led to the development of synthetic lubricants which can be maintained at high temperatures without decomposition and have a low risk of combustion.
The synthetic oils are produced using highly refined processes and sophisticated formulations.They are derived from synthetic compounds based on PAO (polyolefin, polyester, polyglycol), non-synthetic PAO, esters, alkylated naphthalene and alkylated benzene.The use of synthetic oils has become more important in areas where the use of mineral oils does not meet the required needs.
Lubricating oils can cause serious environmental problems if they are discharged indiscriminately, polluting rivers and groundwater.The improper burning of oil adds oxides and toxic gases to the atmosphere.Accordingly, the manufacturers of additives and lubricating oils have been developing products with a longer life, as this tends to reduce oil discharges along the equipment life cycle.
A key feature of lubricants is their behavior with increasing temperature.They are not used at room temperature; the temperature and pressure are often high.The oils undergo change when the temperature increases, and their degradation under operating conditions is a problem involving significant economic losses.To report certain special properties of the oil, or to improve the existing ones, especially when the lubricant is subjected to severe working conditions, chemicals are added (additives).The degradation of a lubricant is not an instantaneous process -the loss of its physicochemical properties and contamination are progressive over time and with the use of equipment along its lifetime.Lubricant degradation is affected by [20]: oxidation; viscosity variation; contamination; loss of additives (anticorrosion, anti-wear, dispersing agents, etc.).
Today's high-performance lubricants do more than simply reduce friction and wear: they control the formation of deposits, control airborne contaminants, protect against corrosion, have a cleaning function, and maintain the proper operating temperature.

Oil analysis
Under certain conditions, a lubricant can deteriorate and no longer fulfill its intended function.It is generally a function of the length of service, system temperature, environmental conditions or the stress that it is suffering and can often be traced to the presence of dirt or water, acidity, insufficient flow, or inappropriate levels of viscosity.Any of these can cause lubricated components to malfunction.Even when the lubrication system is well designed and maintained, breakdowns can occur in the component, resulting in the deterioration of the lubricant.The deterioration can be chemical or physical, generated internally by the lubricant or by external phenomena.Physical deterioration, often called contamination, materializes as foreign matter in the lubricant, such as water, foundry sand, weld slag particles, metal shavings, dust and abrasive wear particles.
Lubricant analysis is regularly performed in some industries [21].It involves four basic steps: 1) Obtaining a sample  Collection of a representative sample of a lubricant, observing certain precautions such as: using clean and dry containers; taking extreme care during collection to prevent external contamination; taking samples at operating temperatures.2) Identifying a relevant sample.

3) Performing physical-chemical analysis:
 Degree of physical and chemical deterioration, i.e., the degree of contamination and degradation, can be evaluated using a set of standard and specialized tests, such as: measuring certain properties and comparing these with a baseline value.

5) Validating diagnosis:
 Frequency with which each lubricant ought to be checked depends on various operational factors such as: importance of the equipment; total time of service; scale of production; security; time to failure after detection.
In this section the analysis of service lubricants is addressed, with the monitoring of the evolution of the degradation of the Diesel oils of one the bus fleet; this will have three welldefined phases: 1) In a first phase they will be selected the vehicles that will be the target of analysis and monitoring in the evolution of the degradation of the oils; this monitoring will be done through the periodic collection of samples of the oils of the selected vehicles and sending to a proper place for their analysis; 2) In a second phase, an in-depth study of the results obtained in the analyses, as well as of the prediction algorithms to be used will be carried out; 3) In the third and final phase, an analysis will be made of planned maintenance plans with a view to their improvement.This shall take into account the results obtained in monitoring the evolution of oil degradation; this phase will also serve to present proposals for the improvement of planned maintenance schedules and the reduction of costs that come from it.
This monitoring was done through periodic collection of oil samples from the various vehicles selected and, since there was a small number of samples collected during the period in which this monitoring was developed, it was felt the need to use data from older samples belonging to the same homogeneous group.
These samples were sent to a laboratory based in a European Country, with all the characteristics of the vehicle and the oil, such as: Subsequently, the reports of the results obtained from the various analyses carried out on the samples collected were received (Figure 1).The data collected was entered into an Excel spreadsheet, in order to create a database where they could be analysed more easily.
The Figure 2 shows an Excel database example, per vehicle.One the variables considered most important and to which the necessary study was given was the Iron content (ppm).This allowed us to draw several conclusions, which are described later, on the state of degradation of the oil and the equipment.

Oil analysis changings through prediction
In first step, it was applied the exponential smoothing method to the iron content (Fe) to determine the evolution of its degradation, as can be seen in Table 2 and Figure 3.The table and graph show a clear degradation in the iron content of the analyzed oils.Obviously, the prediction of the next values will involve increased degradation.When this variable has values like those shown in the table, the oil must be replaced, because the equipment is at a high risk level.
The main formula for exponential smoothing is given by:  where: St+1 Is the forecast for the next time; Is the real value recorded in the present time;

St
Is the forecasted value for the present time;  Is the smoothing parameter.In the second part of the development of algorithms for our model, we analyzed the data using the t-Student distribution, n≤30, ( Where, Is a fixed value used for comparison with the sample mean; Is the average sample; Corresponds to the critical T; Is the sample standard deviation; Is the sample size.
Where tα corresponds to the critical t of a tail, considering the desired confidence interval, and the degrees of freedom n-1.
Next, we followed up on the evolution of the degradation of this variable using the t-student distribution.The objective was to estimate the average value of iron (Fe).As Table 3 shows, the average Fe content was 51 (ppm).This is above the upper normal limits, suggesting a high level of degradation.Note that through additional methods, like the t-student test, it is possible to calculate other important data, such as the sample mean, sample standard deviation, and upper parameter limit for confidence intervals.Lastly, we used bilateral tests of hypotheses for the value of μ: H0: μ=µ0 H1: μ≠µ0 μ is considered a random variable whose distribution for small samples (n <30) is given by: In general, σ (standard deviation of the population) is unknown.The process is the following:  A one-tailed test uses one threshold value (associated with the chosen significance level) and rejects the hypothesis H0 -where T > T critical -when the value of the modulus calculated for the t statistic exceeds the critical value.
Finally, they were estimated the population average of the iron for the following significance levels: 0.001; 0.01; 0.05; 0.1; 0.2.As Table 4 shows, with a value of 80 (ppm) and a confidence interval of 99%, the hypothesis H0 is not rejected.But with a 90% confidence interval, H0 will be rejected because the value of t (1.59) is higher than the value of the confidence interval (1.41).Furthermore, with a confidence interval of 80% (0.90) and a sample average of 51.00, the average value is 37.59.

DISCUSSION
With this condition monitoring model several variables can be evaluated, that can help to understand the evolution of the degradation' state of the oils.The models exemplified here were applied in three ways: 1) Individually, to all vehicles (all parameters); 2) Homogeneous groups of different vehicles (all parameters); 3) To the group of vehicles that use biodiesel as fuel (all parameters).
The exponential smoothing was applied to the iron content (Fe) variable for a bus number XX3 to determine the evolution of its degradation.When this variable has high values, the equipment is at a high risk level, and the oil must be changed.The second model applied to monitor the degradation of the iron content was based on the t-student distribution: it estimates the average of iron content (Fe) -the average content is 99,80 (ppm).
It is also possible to calculate more information such as the sample mean, the sample standard deviation and the upper limit of the parameter to determine the confidence intervals.If the value of 150 (ppm) is found in the iron content variable, with a 99% confidence interval, the hypothesis H0 is not rejected.But, if the confidence level is 90%, H0 is rejected.The value of t (2,35) cannot be greater than the value of the confidence interval (1,53).
If the value of t is used from the t-student table with 80% confidence interval (0,90) and a sample mean of 99.80, a mean value for a population of 70,48 is obtained.
Because the oil itself and the enormous influence that it has in the Diesel engines condition, the accompaniment of its degradation permits to maximize the bus availability itself and the bus fleet in general.
The paper demonstrates that using condition monitoring maintenance, the intervals of the interventions can be increased and, consequently increased the bus fleet availability, reducing the maintenance costs.
Table 5 shows the data about the company under study, such as the number of buses that constitutes the fleet, their availability, the need of buses for production, the number of buses under maintenance and the number of buses that correspond to the reserve fleet, based on a systematic preventive maintenance policy.
Figure 4 (radar map) shows the Availability versus Production Requirement (buses necessary to carry out the careers) of the company during a year.

CONCLUSIONS
The implementation of a condition monitoring based maintenance policy may maximize the physical asset availability, reduce the costs of maintenance, reduce overhead costs and provide an additional guarantee of reliability.
The paper shows how some variables, such as soot and iron content, indicate the condition of Diesel engines.The analysis can be extended to include other variables.
The paper demonstrates that monitoring the condition of oil can increase the availability of equipment and improve fault prevention by allowing early intervention in degradation.It also demonstrates that the implementation of a condition monitoring based maintenance policy, using oil analysis has huge advantages for public transport companies, including lower downtime, higher availability and, consequently, better service to users.
Additionally, this kind of maintenance policy contributes to the rationalization of the size of a reserve fleet.
Finally, the methodology can be used for many other types of physical assets.

Figure 1 -
Figure 1 -Oil analysis results report -Bus

Figure 2 -
Figure 2 -Database in Excel

Figure 4 -
Figure 4 -Availability versus Need for buses -Systematic Preventive Maintenance

Table 3 -
Application of t-student test to iron content -Fe (ppm)

Table 5 -
Availability versus Need for buses -Systematic preventive maintenance