Automation of diesel fuel consumption accounting on a special self-propelled rolling stock of Russian Railways

The article provides the results of a study to assess the effectiveness of automation of diesel fuel consumption accounting for a certain type of special self-propelled rolling stock of the company Russian Railways. The main current provisions of the fuel consumption assessment are considered. It is proposed to supplement the provisions on the adjustment of fuel consumption standards by identification methods. The identification method includes monitoring actual values, performance indicators, rolling stock mileage and fuel consumption forecasts. The analysis showed that the efficiency of fuel consumption automation can be as high as 30%. The effect is achieved by fulfilling the forecast fuel consumption standards and eliminating unauthorized fuel overruns.


Introduction
The current task for the territorial branches of the company Russian Railways is to increase fuel efficiency on a special self-propelled rolling stock (SSPS) One of the ways to increase fuel efficiency the SSPS diesel engines are to automate fuel consumption accounting.
Basic automation equipment includes fuel consumption control sensors installed in the diesel fuel system. The data measured by fuel consumption sensors is monitored through electronic microprocessor systems, which are installed in the cockpit of SSPS drivers.
Processing of fuel consumption data for the SSPS, rationing of fuel for travel and efficiency assessment is carried out at stationary fuel release points or in the fuel departments of the territorial branches of The Russian Railways.
To date, the scientific literature has considered several major areas of fuel efficiency improvement.
The first direction is to include works that define strategic approaches to control the consumption of diesel fuel [1]. The work of this area includes research on both hardware and software to account for fuel consumption.
The second direction includes studies on the impact on fuel consumption and the level of operating costs of locomotives [2]. At the same time, an increase in the age of the locomotive fleet is considered as one of the main factors influencing. The study found that fuel consumption could increase to 2.75% per year. This area includes research to improve the efficiency of diesel engines by disabling part of the cylinders. Improvements in performance can range from 4 to 30%.
The third significant area includes research on fuel regulation [3], as well as on the impact of diesel engines on pollution [4 -10]. These works examine the operational and emission characteristics of an engine powered by a mixture of diesel and organic matter.
The fourth area includes research on the use of hydrogen gas in diesel engines using onboard batteries. The total efficiency of the power plant can be as high as 30% [11]. This area includes, for example, research on the operation of a diesel engine with a two-fuel mode of operation (Common Rai), equipped with an electronic control unit [12,13]. .
As part of the implementation of all designated areas of research, automated diesel consumption systems are provided. The purpose of the study was to assess the impact of these systems on fuel consumption savings during the normal functioning of the SSPS. Accurate measurement of the actual fuel consumption allows to make adequate rationing, to reduce its consumption, including unsanctioned fuel consumption.

Research methods
According to the regulations of the Russian Federation with an indirect method of static measurements and an indirect method of measurements based on hydrostatic principle, the error of sensors to measure diesel fuel should be no more than 0.65 %. According to literature analysis, sensors with similar accuracy can be attributed to sensors described in the works of the [14,15], such as KRAL Volumeter sensors. According to the information provided in the [14] sensors, with a sensor error of 0.5%, the error of the entire automated accounting system can reach more than 10%. Thus, with a sensor error of 0.65%, the total error of automated fuel consumption measurement should be expected at 13%. These calculations suggest that the formation of standards for diesel fuel consumption of the SSPS on the basis of its actual measurements can be made with the same error.
The second objective of the study was to assess the effectiveness of automation of diesel fuel consumption and the use of the method of identifying fuel overruns, including unauthorized. This paper presents the results of the analysis of fuel consumption accounting for the type: MLT-6 -motorvehicle lift loading transport. This type of motor ized spree is widely used in the activities of the "Infrastructure Direction" at the Russian Railway's subsidiaries.
Fuel consumption data were obtained from studies conducted in two phases: -with the equipment of the SSPS with fuel consumption automation devices (2016); -with automation devices unstaffed by the SSPS (2019). Fuel consumption studies were assessed during the normal operation of the SSPS in its regular regimes: working mode, idling mode, transport mode. Automation devices used "KVARTA" kits developed by Electromechanics (http://www.elmeh.ru) that were equipped with SSPS in 2016. According to the developer, the "KVARTA" complex allows to assess the accuracy of fuel consumption with a margin of error of no more than 0.67%.
Statistical analysis and identification models for assessing technological performance were used to improve the analysis methods for assessing the efficiency of diesel fuel consumption at the SSPS processes [16]. Diesel consumption efficiency can only be assessed by comparing actual fuel consumption with the established fuel consumption norms for this type of SSPS and for specific operating conditions. Obviously, the positive efficiency will depend both on the accuracy of the measurement of actual fuel consumption and on the adequacy of the established consumption norms. According to the regulatory document "The method of planning and rationing fuel consumption for special rolling stock in the Russian Railways, No. 2464p approved in 2007, the concept of "fuel consumption rate" is an estimated amount of fuel, necessary to carry out the planned amount of work according to the established technological modes. -hour in total mode. According to the above document, the fuel consumption rate for MLT-6 is determined by the formula: where is wm b -fuel consumption rate for working mode, kg/h; im b fuel consumption rate the SSPS for idling mode.
The methodology does not provide for the rationing of the specific fuel consumption of the SSPS taking into account the transport mode. This may reduce the reliability of the Fuel Consumption (Norm), kg indicator. The disadvantages of the methodology considered should include fixed as static norms of specific fuel consumption, the absence for a number of SSPS, including for MLT-6, the consumption norms for the transport mode.
The identified shortcomings of the above rationing approach can be partially corrected by the use of "Methods of rationing, planning and analysis of diesel fuel use on the SSPS based on onboard systems № 3060p, approved by Russian Railways in 2015.
In the methodology of the rate of specific fuel costs N i b in the i mode of operation, the SSPS is determined by statistics for the base period according to the formula: where F j B F j H -actual fuel consumption, kg and running time, hour, respectively in imode based on the results of k trips is the fuel consumption rate for the transport mode.
The disadvantages of the methods considered include the limited number and types of factors used for rationing. Only opening hours in SSPS modes are used as factors and do not use, for example, Run, km. As a result of the above limitation, multi-factor simulation indicators of diesel consumption are not used. On the basis of previous studies, the authors proposed improvements in the method of rationing and planning of diesel fuel consumption on the SSPS by identification methods. The main provisions and stages of the methodology: 1 where ξ is the stochastic component of F j B , formed by factors that are not taken into account. A significant advantage of the technique over the known is the possibility of using in general non-linear types of operators.
3) Operator A parameters are being identified. In a private case, when linear it is related to э Х factors, the operator of A is advisable to present in the form of a linear multifactorial equation of regression of the species: where 0 a is free member of the regression equation; ... ...
The quality of the model is considered excellent if MAPE is 10%. If 10% of MAPE is 20%, the quality of the model is considered good.

Experimental data and results
The analysis of fuel consumption monitoring results at the MLT-6 was based on measurements taken by Electromechanics company in 2016. The data were obtained from 40 visits over a period of two months (table 1). The research has monitored the actual fuel costs -Fuel Consumption (Fact), kg. and only one operating indicator -Run, km per trip. The actual consumption was compared to the fuel consumption rate -Fuel Consumption (Norma), kg. According to the data provided, graphs of actual and normative fuel consumption were constructed in conjunction with Run (Fig. 1). Visual analysis of the graphs presented in Figure 1 shows a steady excess of the normalized indicators of Fuel Consumption (Norma), kg over actually measured -Fuel Consumption (Fact), kg. Statistical processing of the data showed that the average normative fuel consumption per trip is (51.3 kg) and is 28% higher than the actual (39.9 kg).
The effect is achieved by automating fuel consumption accounting for the SSPS.  (5), (6). Figure 2 shows that there is a high relationship between Run and Fuel Consumption (Norma) and Fuel Consumption (Fact) in the 2016 studies -R correlation ratios of more than 0,82.
Based on the proposed identification method (5)- (7) it is possible to adjust the Fuel Consumption (Norma) indicator based on the construction of a two-factor model based on Fuel Consumption (Fact) and Run (Figure 2 b). In Figure 1, the adjusted expense is presented as Predicted 1f Fuel Consumption, kg. MAPE -estimates of this indicator is 15%, which indicates the good quality of the model. Research to test the effectiveness of automation tools to account for fuel consumption has been continued in 2019. By the time of the research, the automation equipment on the IPT-6 was dismantled. Fuel consumption and operating modes were monitored according to the AU-12 route sheets for the period of two months of 2019 (table.2). It was found that when the route sheets were filled, the actual figures were filled only as inteccly, which did not meet the requirements of the methodology and reduced the accuracy of monitoring. Fuel consumption values by type of work are defined as constant values for each trip поездки:  Grafics in Figure 3 according to Table 2 shows that the database's inflated values can be  Thus, the lack of automation in the SSPS can lead to fuel overruns of an average of 18 per cent or more over the period. Reducing fuel consumption can be achieved by determining the exact fuel consumption standards in the database and by controlling whether it meets actual fuel consumption value. In order to accurately construct fuel consumption norms based on the proposed identification method based on the multidimensional species model (6), it was necessary to assess the relationship between Fuel Consumption (Fact) and the modes of operation of the researched MLT-6. Figure 4a presented histograms of fuel consumption IPT-6 in different modes, the type of regression relationships and the value of R correlation ratios between modes. In this image, the indicators of the resulting model form a surface that interpolates the main field of measurement points. Fuel Consumption (Fact) values representing "emissions" are released from the operational interconnection system and are located at a considerable distance from the formed surface.
In general, if you have more performance, such as environmental or temperature, it is possible to build more accurate models of Fuel Consumption (Norma) indicator.
The method of selection of indicators in this multidimensional model consists of several stages. Figure 5 a shows a dendogram built by cluster analysis. The dendogram is a generalized scheme of correlations of the indicators studied. On the graf, the closer the indicators are to zero, the higher the relationship. The dendogram axis is based on the difference between the unit and the R correlation ratio, i.e. K= (1 -R). For example, in Figure 5, Fuel Consumption (Fact) is most closely related to Hour Work. A projection on the axis of the diagram of combining variables of K = 0,4. The level of correlation is defined R = (1 -K) = 0,6. When selecting the indicators in a multidimensional model, you need to consider and eliminate the impact on the outcome of the multicollinearity property -the presence of linear links between the explaining variables.
To achieve this goal, only one is selected from the association of highly correlated indicators. On the presented dendogram to strongly correlated indicators (R = 0,85) are pairs: Hour in total ( totm H ), Fact ( wm H ) and Run -Fact ( trm H ). When forming a multidimensional model in the STATISTICA program, it is possible to eliminate multicollinearity as part of the standard procedure. Figure 5 b shows the results of identifying three parameters of the Fuel Consumption (Norma) model, using the performance indicators in the three modes provided in the traditional method: working mode idling mode, transport mode. The "Вeta" column presents paired correlation ratios, and the "B" column shows the identified parameters of the model's multidimensional equation. "Intercept" free member of the equation.The adequacy of the model is highmultiple correlation ratio of R=0.899. Fisher's criterion F =52 is more than table ( According to the parameters of the two-factor model in Figure 3, the implementation of Predicted 2f Fuel Consumption, kg, which is chosen as a normative one, is built. The values of this implementation allow: -to use it to assess fuel efficiency in travel when the working time and run of the SSPS are variationd; -to filter the "emissions" of fuel consumption in the database and to record unauthorized fuel consumption. In this example (figure 3), a correction of the fuel consumption indicator is made, which should be spent based on the MLT-6 modes of operation. In the period 16 Thus, only three trips can be detected fuel overruns of 565 kg. At an average monthly consumption of 1,800 kg, unauthorized fuel overruns are at least 30%. The error of the resulting model relative to Fuel Consumption (Fact) is defined as MAPE = 8%, which indicates.