INTRODUCTION

The measurement and registry of the different values of the machines structural and functional parameters can be assigned to diagnosis variables (sometimes also called parameters of diagnosis).When sensors connected permanently in the system are controlled to diagnose and the values are automatically registered and kept in a historical data, not always this fact that seems simple, implies to count on suitable methods and means of this measurement procedure, which is known like continuous monitoring (on line) (Silva et al., 2001). In opposite case, the measurements and registry of the data made every certain time interval, which is known like discreet monitoring (off line), is not discarded now.

Camacho et al. (2007), stated the following “The methods of detection and diagnosis of faults based on historical data are the most applied in the industries of processes, because they are easy to be implemented, do not need a mathematical model of the process and require little a priori knowledge of the process and the faults”. That criterion is shared by the authors of this research, but it is valid whenever there is correct and reliable previous information, not always available.

Regarding the previous exposition reiterated by Camacho et al. (2007), it is important to indicate that “In the methods based on historical data, the detection and diagnosis of faults are made by means of the processing a great volume of data. Due to this, several ways exist in which the data can be changed and presented like a priori knowledge for the diagnosis system”. The expert systems, the methods based on fuzzy logic and the qualitative analyses of tendency are examples of methods of qualitative extraction. For the case of methods of quantitative extraction, in Literature, neuronal networks and statistical techniques, appear widely used. Statistical techniques such as: the Square Analysis of Principal Components (PCA), Partial Minimums (PLS), the Discriminate Analysis of Fisher (FDA), the Analysis of Variables Canonicals (CVA) and the Analysis of Independent Components (ICA) have been applied in the detection and diagnosis of faults previously. These methods as used in the diagnosis, must be used together to determine the faults (Cigolini, 2009).

Specialized literature in data analysis, considers that the evaluation of the behavior of variables should be made through control graphs. For example, Escalante (2003), defines to the graphs of control like tools that show the performance of a process with respect to the time, being its objective to evaluate, to control and to improve its processes. In them, the process is evaluated as stable if the data are within the control limits and unstable if they are outside such.

The authors of this work maintain the most concrete criterion referring that the control graphs evaluate the behavior of the variable with two criteria: good (stable) and critic (unstable), and they do not only analyze the fluctuation of the variable data within the control limits. This asseveration facilitates a better interpretation of the phenomenon under study and allows objectively reaching improving proposals.

Escalante (2003) stated there are three methods of analyses to know the tendency an event: the method of visual examination of the graph, of the line of tendency and the rule of the filters. The three methods only determine if the tendency of an event is bullish or bearish, they do not look for the causes of the line of tendency abrupt rupture, therefore it is inferred that the forecast actions before indicated (logistic and of organizational assurance), are outside the problem.

“The more used means to detect the tendency of a series are based on the application of filters to the data. A filter is not more than a mathematical function that applied to the values of the series produces a new series with certain characteristics. Among those filters we found the mobile means”¹. “There are other procedures to extract the tendency, as adjustment of polynomials, smoothed by means of exponential functions, etc. A filter class, that is particularly useful to eliminate the tendency, is based on applying differences to the series until turning it stationary. A difference of first order is obtained reducing two contiguous values”. These methods only allow obtaining new smoother series from the initial data, but they do not let to know the reason of the variation.

From the study of the methods referred by the authors previously mentioned, Batista y Urquiza, (2008), expose three methods for the analysis of the values of diagnosis variables that are within the technique of tendency analyses and that for their objectivity deserve to be analyzed:

First: comparative state of the value of the variables with respect to the warning levels or alarm that have been established, according to recommendations of the manufacturer, well-known and studied norms that adjusts to the conditions of work of the evaluated machines, or according to established own norms for each one of the machines from its particular conditions of operation.

Second: Comparative method (MC). Through the reason of growth of value of the variables, as it is declared in ISO 2372-1974 norm.

Third: Rapidity of change of value of the variables (RCVV). This method consists in calculating the average rate of change (Ri - j) of the variable value between two different moments of time i and j (time interval of work), that is the time worked by the machine between two consecutive measurements.

From the analysis made to Batista y Urquiza’ methods (2008), it is possible to conclude: the first presents the disadvantage to evaluate the state of the variable in good or critical, in good while the value of the variable is within the limits defined by the manufacturer and in critic when this value has exceeded the limit. The second one was elaborated to evaluate the amplitude of the vibrations (as important variable for the object of study of the present work) and to use it in the evaluation of other variables, where it is necessary to elaborate its rules again.

The third one, while calculating the rapidity of changing of the variable value, provides better information and evaluation of its state. Nevertheless, to know the RCVV, the time the machine has worked between two consecutive measurements must be place in the denominator. Such data is sometimes difficult to be obtained. In addition, in their interpretation rules, expressions such as relatively great and small are used. Due to this, the authors of the proposal state that the effectiveness of the analysis of cases, when applying this method, depends on the experience of the specialist and the way the mathematical instruments of estimation are utilized.

Through the techniques of Mining of Data, it is known that in many occasions the datum does not directly reveal its relation with a particular phenomenon, but often the datum hides the necessary and effective information to make a correct technical diagnosis and interpretation of the obtained result, which makes necessary to design technologies that allow extracting this information to be used (Larose, 2005).

In addition to the previous authors, some others have studied the conditions of monitoring or the data or variable values of the variables like method to evaluate the state of an internal combustion engine or to know the presence the faults in this equipment. These authors are:

Porteiro et al. (2011), who outline that the condition of the industrial equipment monitoring in general, and individual of diesel engines, is very important to assure the production and to reduce the costs in all the industrial facilities.the diesel engines, is very important to assure the production and to reduce the costs in all the industrial facilities.

Li et al. (2012), who state that the normal operation of marine diesel engines assures the completion and the effectiveness of a trip. Any fault can generate significant economic losses and to severe accidents. It is therefore crucial, to supervise the conditions of the engine in a reliable and opportune way to prevent its bad operation.

Figlus et al. (2014), refer that a change in the technical specification of mechanical components of internal combustion engines cannot be detected by the systems of diagnosis on board installed in vehicles. In similar cases, the measures and the analyses of the vibrant acoustics signals that are registered can be useful. The authors of this work consider that, because the systems of diagnosis register several variables, and that context makes it difficult to determine the relation between the presence of the failures and the variables of diagnosis.

Jinming et al. (2012), tpresent a new method that relates, through the empirical mode of decomposition technique (EMD), the signals of the vibration of the butt surface produced by the separation of the valve with the failures of the diesel engine.

Albarbar (2013) outlines the acoustic signals emitted by the diesel engines convey useful indicators on their conditions of operation and state of functioning. The author proposes to use the technique continuous wavelet transform (CWT) to determine the relation between the variations of the speed and the load of the engine, with the fault detection during the injection of the engine and failures related to the lubrication.

Kateris et al. (2014) analyze that recently, the investigations have been centered in the putting into practice the analysis of the vibration signals for the diagnosis of the functioning state of the systems. In the article, the authors elaborate a diagnosis system based on neuronal networks that relates the vibrations to the failures to detect the failures in the bearings of rotatory machines.

Authors such as: Kim et al. (2013) and Deng y Zhao (2014) refer the importance of the diagnosis in real time. Deng y Zhao (2014) approach in his article the difficulty to extract the useful characteristics of the vibration signals and propose an integral method for the extraction of the failure characteristics based on the kurtosis and the energy of Teager. In addition, Zhang et al. (2014) based on the estimation of the density of Kernel and on the divergence of Kullback-Leibler, propose a statistical method for the diagnosis of failures.

Due to before exposed and to the importance of having a coDue to the above and to the importance of having a correct evaluation of the variables, the authors of the work, after studying the different methods of Analyses of Tenden- cy, statistical, series of time, the graphs of control and the analysis the condition monitoring, conclude the necessity of having a procedure or method to find, in the available data, the necessary, significant and useful information to know the causes of the changes in the values of the variable and simul taneously, this knowledge is used in the technical diagnosis of the engines of internal combustion (MCI) Diesel engine of generator sets destined to the generation of electricity in the country, being this central objective of the work.

This analysis has allowed the authors the search of a the- oretical procedure that allows objectively evaluating the state of a diagnosis symptom. That is also a source of information to be used in the process of equipment diagnosis and thus to evaluate correctly their technical state to propose correction measures and forecast to improve the motors operation

METHODS

In order to develop the investigation whose results are exposed in this article, authors used theoretical methods of investigation as analysis and synthesis and historical - logical for the study of the object through the time and to develop a logical analysis.

The techniques used in the harvesting of the data were participant observation, official documents of the company: technical files, internal registries of maintenances and documents.

Mathematical development of the formula to evaluate the variation of the value of the diagnosis variable

The formula to evaluate the variation of the value of the diagnosis variable will be used in a case study where the state of the diagnosis variables registered in fixed diesel internal combustion engines is going to evaluated.

The formula to evaluate the variation of the value of the diagnosis variable will be used in a case study where the state of the diagnosis variables registered in fixed diesel internal combustion engines is going to evaluated.

In practical life it is not only of interest to know how a function in an interval varies, but also it is of interest to determine the instantaneous variation of that function in a point within that interval, and the causes that provoke it. This variation or increase is known if differential calculus is applied to the function. This method is the one used to obtain the formula set out in the work, to evaluate the values of the diagnosis variables regarding the limits defined by the manufacturer in diesel internal combustion engines of the generator sets.

The study begins with the assumption that between two consecutive measurements there will always be a linear function so that, knowing X and Y values in the moment “n-1”, the value of the function at moment “n”, can be calculated given by expression 1:

(1)

Where:

y^n-1 Value of the variable and at the moment n-1;

x^n-1 Value of variable X at the moment n-1.;

k Constant.

If the increase of the function is calculated now, the expression 2 will be obtained, in which variable “x” can take values in the closed interval [a, b], where “a” and “b” are the superior and inferior value limits of the variable, respectively:

(2)

Where:

Average variation of the function or increase of the function.

variation of the value of variable X.

From the expression 2, it is clear that the increase or decrease of the variable is given by the constant “k”, which can be cleared as it is in expression 3:

(3)

The calculation of the variation of the value, that in ahead will be denoted by “Vv “, is expressed through expression 3, where the difference the images of expression 1 can be repre- sented by (Yn-Yn-1). The difference of the value of the dominion given by the difference between the limits of the interval; in this case, they represent the superior and inferior limit defined values for the variable analyzed.

Definitively, expression 3 takes the form:

Where:

V_v -variation of the value of the variable (dimensionless); Yn -value of the variable at moment “n”;

Y_{n -1} -value of the variable at the moment “n-1”;

SL -superior limit of the variable studied;

(4)

IL -inferior limit of the variable studiedy.

The variation of the value of the variable can be found between 0 and -1 when the values of the variable studied are decreasing and it takes values between 0 and 1 when the vari- able is increasing.

Procedure to evaluate the variation of the value of the diagnosis variables in internal combustion engines of the generator sets

Steps:

1. To make the calculation of the variation of the value of the variable by equation 4

2. To apply the following rules for the evaluation of the variable:

Rules

1. If VV = 0 there is no change in the state of the variable. Stable State [E].

2. If VV is equal to 1 there is a significant change of the state of the variable, serious symptom for the operation of the equipment. Critical State [C].

If VV is between 0 and 1 there is a gradual change of its state; the probability of the existence of a failure increases as the To elaborate a graphic of the normal value and the variation of the value of the variable.

RESULTS AND DISCUSSION

Example of calculation of the variation of the value of the variable

In the work, the Table 1 of random form for a period of ten days calculation is in the table of the variation of the value for a variable (fuel pressure) of an internal combustion engine of a generator set. The variation was calculated for the 8 public variables of 22 MCI of the existing 32 in a location of electricity generation, representing 61.1 % of the total. The values took within a period de10 months through of Regime of Control Book of the Engine in Operation, instituted in the company.

In Table 1 the fall of pressure of gradual form is observed being more significant between days 8 and 9. By means of the comparative method, see Table 1, the values of the variable are always in the state Good for being within the limits defined by the manufacturer for the pressure of the fuel. They are: 0,38 inferior limit and 0,7 superior limit MPa under normal conditions of work of the equipment: speed 1800 min -1 and 75 % of load 1416 kW of power.

By means of the procedure to evaluate the variation of the value of the diagnosis variables, the values of the variable are in rule 1 and 3 in stable state and of analysis (Table 1)

TABLE 1.

Calculation of the variation of the value

The graphs of Figures 1 and 2 are made with the data of Table 1, which allow reaching the following valuations.

In Figure 1, the graphical analysis of the behavior of the values of the fuel pressure variable of an installed internal combustion engine in a location of electricity generation. There it is possible to observe through the comparative method how the values of the variable stay within the range of work defined by the manufacturer. That visual examination of the graphic allows concluding that the variation of the values of the variable of the internal combustion engine under normal conditions of work, previously described, does not enable to know if this variation is due to the appearance of a potential failure in the equipment fuel system because the values are between 0,38 and 0,7 MPa.

FIGURE 1.
Graphical analysis of the values of the variable by the MCI.

Figure 2 allows knowing how the variation of the values of the variable is within the limits from 0, 7 to 0, 38 MPa. For example, the variations between the eighth and ninth, and between the quarter and fifth days is of analysis because those variations are related to rule number 3, they take values from -0, 34 and -0, 22.

This visual and numerical examination of the values of the variable allows knowing that there is a disturbance in the process, in the structural and nonstructural elements related to the pressure fuel of the MCI. It means that a failure can be happening (a potential or a functional developed failure is identified) either in the fuel filter, the temperature or the fuel viscosity among others.

FIGURE 2.
Graphical analysis of the variation of the values of the variable.

The fulfillment of the procedure steps has allowed knowing that the variation of the values of a diagnosis variable, in this case, the pressure of the fuel of an internal combustion engine of the generator sets can be related to the appearance of a potential failure in the equipment or the system studied. This conclusion with the comparative method or the analysis of control graphs is not possible to be reached because both define that when the values of the variable are within the work range they do not indicate the appearance of a potential failure until they reach the value defined by the alarm manufacturer. In the case of study the alarm value is 0,38 MPa.

CONCLUSIONS

• A theoretical procedure has been designed and developed that enable to evaluate the state of a diagnosis variable (pressure of the fuel in this case) and to monitor its evolution. These analyses are not possible by any other method studied in the consulted bibliography. On the other hand the work allows inferring the kind of relationship existing between the state of the variable and the technical state of the equipment that is being analyzed.

• Due to its relevance and the technical and economical characteristics of the internal combustion engine of the generator sets, the developed procedure can be applied and generalized to similar equipment and apparatuses.

REFERENCES

ALBARBAR, A.: “An investigation into diesel engine air-borne acoustics using continuous wavelet transform”, Journal of Mechanical Science and Technology, 27(9): 2599-2604, 19 de septiembre de 2013, ISSN: 1738-494X, 1976-3824, DOI: 10.1007/s12206-013-0703-9.

BATISTA, R.C.; URQUIZA, S.S.: “El RCVV: Un enfoque diferente en el diagnóstico técnico.”, Ingeniería Mecánica, 3: 29–32, 2008, ISSN: 1815–5944.

CAMACHO, O.; PADILLA, D.; GOUVEIA, J.L.: “Diagnóstico de fallas utilizando técnicas estadísticas multivariantes”, Revista Técnica de la Facultad de Ingeniería Universidad del Zulia, 30(3): 253–262, 2007, ISSN: 0254-0770.

CIGOLINI, R.D. (ed.): Recent advances in maintenance and infrastructure management, no. solc. TS177 .R44 2009, Ed. Springer, London, 290 p., OCLC: ocn310400707, 2009, ISBN: 978-1-84882-488-1.

DENG, L.; ZHAO, R.: “Fault feature extraction of a rotor system based on local mean decomposition and Teager energy kurtosis”, Journal of Mechan- ical Science and Technology, 28(4): 1161-1169, 1 de mayo de 2014, ISSN: 1738-494X, 1976-3824, DOI: 10.1007/s12206-013-1149-9.

ESCALANTE, V.E.J.: Seis-sigma: metodología y técnica, Ed. Limusa-Noriega, México, 436 p., OCLC: 54022083, 2003, ISBN: 978-968-18-6391-3.

FIGLUS, T.; LIŠČÁK, Š.; WILK, A.; ŁAZARZ, B.: “Condition monitoring of engine timing system by using wavelet packet decomposition of a acoustic signal”, Journal of Mechanical Science and Technology, 28(5): 1663-1671, 20 de mayo de 2014, ISSN: 1738-494X, 1976-3824, DOI: 10.1007/s12206-014-0311-3.

JINMING, L.; ZHENZHEN, L.; KANGLE, W.: “Fault Diagnosis of Diesel Engine Based on EMD and TFD”, [en línea], En: Jin, D. y Lin, S. (eds.), Advances in Mechanical and Electronic Engineering, ser. Lecture Notes in Electrical Engineering, Ed. Springer Berlin Heidelberg, vol. 176, Berlin, Heidelberg, pp. 261-266, 2012, ISBN: 978-3-642-31506-0, Disponible en: http://link.springer.com/10.1007/978-3-642-31507-7, [Consulta: 3 de junio de 2016].

KATERIS, D.; MOSHOU, D.; PANTAZI, X.-E.; GRAVALOS, I.; SAWALHI, N.; LOUTRIDIS, S.: “A machine learning approach for the condition monitoring of rotating machinery”, Journal of Mechanical Science and Technology, 28(1): 61-71, 23 de enero de 2014, ISSN: 1738-494X, 1976- 3824, DOI: 10.1007/s12206-013-1102-y.

KIM, H.C.; CHO, M.G.; SHIM, J.: “Multi-stage diagnostic system for reciprocating compressor using DTW technique”, Journal of Mechanical Science and Technology, 27(8): 2293-2298, 19 de septiembre de 2013, ISSN: 1738-494X, 1976-3824, DOI: 10.1007/s12206-013-0612-y.

LAROSE, D.T.: Discovering knowledge in data: an introduction to data mining, no. solc. QA76.9.D343 L38 2005, Ed. Wiley-Interscience, Hoboken, N.J, 222 p., 2005, ISBN: 978-0-471-66657-8.

LI, Z.; YAN, X.; YUAN, C.; PENG, Z.: “Intelligent fault diagnosis method for marine diesel engines using instantaneous angular speed”, Journal of Mechanical Science and Technology, 26(8): 2413-2423, 23 de agosto de 2012, ISSN: 1738-494X, 1976-3824, DOI: 10.1007/s12206-012-0621-2.

PORTEIRO, J.; COLLAZO, J.; PATIÑO, D.; MÍGUEZ, J.L.: “Diesel engine condition monitoring using a multi-net neural network system with nonintrusive sensors”, Applied Thermal Engineering, 31(17–18): 4097-4105, diciembre de 2011, ISSN: 1359-4311, DOI: 10.1016/j.applthermaleng.2011.08.020.

SILVA, F. de la T.; MARÍN, E.P.; MESA, C.: “Técnica de monitorado continuo (on–line) para la evaluación del estado técnico de los turbogrupos de 64 y 100 MW.”, Ingeniería Mecánica, 2: 27–35, 2001, ISSN: 1815-5944.

ZHANG, F.; LIU, Y.; CHEN, C.; LI, Y.-F.; HUANG, H.-Z.: “Fault diagnosis of rotating machinery based on kernel density estimation and Kull- back-Leibler divergence”, Journal of Mechanical Science and Technology, 28(11): 4441-4454, 2014, ISSN: 1738-494X, 1976-3824, DOI: 10.1007/ s12206-014-1012-7.

Notes