Estimation device for the amount of oil coke deposits in a turbocharger

Abstract

An estimating device for an oil coke deposition amount in a turbocharger, which estimates an oil coke amount deposited inside a turbocharger installed in a vehicle and comprises an executing device and a storing device, wherein: the storage device stores a neural network into which driving state variables indicating a driving state of the vehicle are input and from which an internal temperature of the turbocharger is output, the neural network being learned using a measured value of the internal temperature and measured values ​​of the driving state variables measured at a measuring time of the internal temperature as teacher data; and the execution device executes an interior temperature estimation process for calculating an output of the neural network as an estimated value of the interior temperature using the measured values ​​of the driving state quantities as an input, and a deposit amount estimation process for calculating an increase amount of the deposit amount based on the estimated value of the interior temperature and for calculating an estimated value of the deposit amount as a value integrating the increase amount.

Description

BACKGROUND OF THE INVENTION1. Field of the InventionThe present invention relates to an oil coke deposition amount estimating device in a turbocharger.2. Description of the Related ArtIn JP 2009-270524 A, it is mentioned that oil contained in blowby gas flowing into a compressor of a turbocharger together with intake air cokes by undergoing a temperature rise due to compression of the intake air in the compressor, and is fixed to the inside of the compressor.Further, US 2019 / 0 325 671 A1 discloses an apparatus for machine learning an amount of unburned fuel using a neural network, in which correlation functions showing correlations between parameters and an amount of unburned fuel are found for parameters related to the operation of an internal combustion engine, and the parameters having strong degrees of correlation with the amount of unburned fuel discharged from the engine are selected from the parameters based on the correlation functions. The amount of unburned fuel is learned by the neural network from the selected parameters and the amount of unburned fuel.In addition, WO 2013 / 080,360 A1 discloses an internal combustion engine having a supercharger, comprising: a blow-by gas path that introduces blow-by gas to a position on the upstream side of a compressor in an intake path; an oil separator installed halfway along the blow-by gas path; a blow-by gas bypass path that bypasses the oil separator; and a switching valve that can select between an oil-collecting passage shape and a non-oil-collecting passage shape. When the operating conditions are such that the accumulation of deposits inside the compressor poses a problem, the switching valve is controlled so that the non-oil-collecting passage shape is selected.Summary of the InventionInside a turbocharger there is oil which is used, inter alia, for lubricating a bearing journal. Also, when the inside of the turbocharger is heated to a high temperature by exhaust gases during the operation of the engine, this oil inside the turbocharger cokes and deposits on a wall surface of an oil passage, a journal part, etc. As the deposition of coked oil, that is, so-called oil coke, proceeds, the oil flow and rotation of a turbine shaft may be obstructed.At present, the only way to check the condition of oil coke deposit inside a turbocharger is to disassemble the turbocharger. Most vehicles, however, are operated without periodic disassembly and maintenance of the turbocharger. In these cases, the turbocharger needs to be designed so that the oil coke deposition amount remains in such a range that does not cause any problem throughout the life of the vehicle. Such a configuration would be based on the assumption of common operating conditions and operating times of vehicles. Meanwhile, even vehicles developed as home vehicles can be used in various forms including car sharing. In such cases, the operating conditions and operating times between the vehicles vary widely. This makes it difficult to define the common operating conditions and operating times and thus design a turbocharger as described above.A coke deposition amount estimating device in a turbocharger that solves the above problem corresponds to a device that estimates an oil coke amount deposited inside a turbocharger installed in a vehicle, and includes an execution device and a storage device. The storage device stores a neural network to which driving state quantities indicating a driving state of the vehicle are input, and from which an internal temperature of the turbocharger is output. The neural network is learned using, as teacher data, a measurement value of the internal temperature and measurement values of the driving state amounts measured at the measurement time of the internal temperature. The execution device executes an internal temperature estimation process for calculating an output of the neural network as an estimation value of the internal temperature using the measurement values of the driving state amounts as an input, and a deposition amount estimation process for calculating an increase amount of the deposition amount based on the estimation value of the internal temperature and calculating an estimation value of the deposition amount as a value integrating the increase amount.The amount of oil coke that is newly generated and deposited inside a turbocharger depends on the internal temperature of the turbocharger. The internal temperature of the turbocharger changes with the running state of the vehicle. There are a large number of driving state quantities that affect the internal temperature of the turbocharger, and the relationship between each driving state quantity and the internal temperature is complicated.In view of this, the neural network stored in the storage device of the above-described estimation device has learned the relationships between the driving state amounts and the internal temperature in advance. Using such a neural network, the internal temperature of the turbocharger can be estimated from the driving state quantities of the vehicle. In addition, the increase amount of the oil coke deposition amount can be obtained from the estimated internal temperature, and further, the deposition amount can be obtained as a value integrating this increase amount. Thus, this estimating device can estimate accurately the amount of oil coke deposited inside the turbocharger.The execution device of the above-described estimation device may be configured to execute an instruction process for instructing restriction of engine output when the estimated value of the deposition amount exceeds a predetermined threshold value. Limiting engine output may mitigate the increase in the internal temperature of the turbocharger and thereby the subsequent oil coke deposition. Thus, the possibility of excessive deposition of oil coke beyond an allowable range can be reduced.The execution device of the above-described estimation device may be configured to execute a command process for instructing a notification of a user of the vehicle that the turbocharger is in a maintenance-requiring state when the estimation value of the deposition amount exceeds a predetermined threshold. In this case, at the time when the oil coke deposit has reached a certain amount, a notification is issued to the user recommending maintenance. Thus, maintenance at an appropriate time before the oil coke deposit exceeds the allowable range can be recommended.A certain time is required before changes in the driving state amounts are reflected in the internal temperature. Therefore, it is desirable that time-series data of the driving state amounts be used as the input to the neural network.Main running state quantities of a vehicle that affect the internal temperature of the turbocharger include a vehicle speed, an engine speed, an accelerator operation amount, a fuel injection amount, a boost pressure, an intake air flow rate, an outside air temperature, and an intake air temperature. Therefore, it is desirable to include one or more of these quantities in the driving state quantities used as the input to the neural network.BRIEF DESCRIPTION OF THE DRAWINGSFeatures, advantages, and technical and industrial significance of exemplary embodiments of the invention will be described below with reference to the accompanying drawings, in which like numerals denote like elements, and in which: FIG. 1 is a diagram schematically showing the configuration of an embodiment of an estimation device of an oil coke deposition amount in a turbocharger; FIG. 2 is a block diagram showing relationships between processes involved in estimation of the oil coke deposition amount performed in the estimation device; FIG. 3 is a diagram schematically showing the configuration of a neural network used for estimating the oil coke deposition amount in the estimation device; FIG. 4 is a flowchart of an internal temperature estimation process executed by the estimation device; FIG. 5 is a flowchart of a deposit amount estimation process and a command process executed by the estimation device; and FIG. 6 is a graph showing relationships among a wall surface temperature of an oil passage of the turbocharger, a holding time of the wall surface temperature, and the oil coke deposition amount.Detailed Description of EmbodimentsAn embodiment of an oil coke deposition amount estimating apparatus in a turbocharger will be described in detail below with reference to FIGS. 1 to 6.Configuration of the TurbochargerFirst, the configuration of a turbocharger 20 for which the estimation device of the embodiment estimates an oil coke deposition amount will be described with reference to FIG. 1. As shown in FIG. 1, the turbocharger 20 is installed in an engine 10. The engine 10 is mounted on a vehicle. The engine 10 is provided with an intake passage 11 and an exhaust passage 12. Further, in the engine 10, an oil pump 13 is installed, which operates in response to rotation of the engine 10.The turbocharger 20 includes a turbine housing 21 installed in the exhaust passage 12 of the engine 10 and a compressor housing 22 installed in the intake passage 11 of the engine 10. The turbine housing 21 and the compressor housing 22 are coupled to each other via a bearing housing 23. Inside the turbine housing 21, a turbine wheel 24 is installed, which rotates when blown by exhaust gas flowing through the exhaust passage 12. Inside the compressor housing 22, a compressor wheel 25 is installed which compresses intake air flowing through the intake passage 11 upon rotation thereof. A turbine shaft 26 coupling the turbine wheel 24 and the compressor wheel 25 to each other is passed through the bearing housing 23. The turbine shaft 26 is rotatably mounted on the bearing housing 23 via a slide bearing 27. An oil passage 28 is formed inside the bearing housing 23 and corresponds to a passage for oil to flow through the slide bearing 27. A part of the oil discharged from the oil pump 13 is guided to the oil passage 28.Configuration of Estimation DeviceNext, the configuration of the estimation device of the embodiment will be described. The vehicle provided with the turbocharger 20 is equipped with an electronic control unit 30 for controlling the engine. The electronic control unit 30 includes an arithmetic processing circuit 31 that executes various processes for machine control, and a memory 32 that stores programs and data for machine control. To the electronic control unit 30, detection signals of state quantities indicative of a running state of the vehicle including a vehicle speed V, an engine speed NE, an accelerator operation amount ACC, a boost pressure PB, an intake air flow rate GA, an outside air temperature TH1, and an intake air temperature TH2 are input. In the configuration of this embodiment, the electronic control unit 30 corresponds to the estimation device.FIG. 2 is an overview of processes executed by the electronic control unit 30. Each of the processes illustrated in FIG. 2 is a process implemented when the arithmetic processing circuit 31 reads and executes a program stored in the memory 32.The arithmetic processing circuit 31 performs a required output determination process F 1 for determining a required output PE* corresponding to a required value of an engine output and an operation amount determination process F 2 for determining an engine operation amount based on the required output PE*. In the required output determination process F 1, the value of the required output PE* is determined based on the engine speed NE, the accelerator operation amount ACC, and so on. In the operation amount determination process F 2, the engine operation amounts including a throttle opening TA, a fuel injection amount QINJ, and an ignition timing AOP are determined based on the required output PE*, the engine speed NE, and so on. In the operation amount determination process F 2, the value of each engine operation amount is determined so that an engine output corresponding to the value of the required output PE* can be generated at the current engine speed NE.Here, the arithmetic processing circuit 31 performs an internal temperature estimation process F 3 for estimating the internal temperature of the turbocharger 20, and a deposition amount estimation process F 4 for estimating an amount of oil coke deposited inside the turbocharger 20. In the internal temperature estimation process F3, an estimated value of the internal temperature of the turbocharger 20 is calculated based on the running state amounts of the vehicle. In the deposition amount estimation process F 4, an estimation value of the oil coke deposition amount is calculated based on the estimation value of the internal temperature obtained in the internal temperature estimation process F 3. Further, the arithmetic processing circuit 31 executes a command process F 5 based on the estimated value of the oil coke deposition amount to instruct that a warning lamp 33 is lighted and that the engine output is restricted. In this embodiment, the arithmetic processing circuit 31 that executes the internal temperature estimation process F 3, the deposition amount estimation process F 4, and the command process F 5 corresponds to the execution device.In this embodiment, in the internal temperature estimation process F 3, wall surface temperatures at three portions of the oil passage 28 are respectively estimated as the internal temperature of the turbocharger 20. In the deposition amount estimation process F 4, the amount of the oil coke deposited on each of the three portions is estimated. These three portions correspond to the portions P 1, P 2, and P 3 illustrated in FIG. 1. These portions P 1, P 2, and P 3 are portions of the oil passage 28 provided inside the turbocharger 20 where the wall surface temperature tends to become high, and the deposition of oil coke tends to cause problems.Configuration of Neural NetworkIn the internal temperature estimation process F 3, the arithmetic processing circuit 31 calculates estimated temperatures t 1, t 2, and t 3 corresponding to estimation values of the wall surface temperatures at the portions P 1, P 2, and P 3, respectively, using a neural network 34 stored in the memory 32.The neural network 34 includes an input layer having N nodes, an intermediate layer having M nodes, and an output layer having three nodes. The symbol "i" in the following description means any integer that is not less than one and not more than N. The symbol "j" in the following description means any integer that is not less than one and not more than M.In FIG. 3, input values of the respective nodes in the input layer are denoted by X[1], X[2],..., X[N]. The input values X[ 1] to X[N] of the respective nodes are driving state amounts that are state amounts indicating a driving state of the vehicle. As the running state quantities which constitute the input values X[1] to X[N], time-series data of each of the vehicle speed V, the engine speed NE, the accelerator pedal operation amount ACC, the fuel injection amount QINJ, the boost pressure PB, and the intake air flow rate GA, and instantaneous value data of the outside air temperature TH1 and the intake air temperature TH2 are used. Here, the time series data refers to a series of values arranged in chronological order and representing a predetermined number of the latest values among measured values of each driving state quantity acquired in a predetermined sampling cycle.In FIG. 3, input values of the respective nodes in the intermediate layer are denoted by U[ 1], U[ 2],..., U[M], and output values of the respective nodes in the intermediate layer are denoted by Z[ 1], Z[ 2],..., Z[M]. The input value U[j] of each node in the intermediate layer is calculated as a sum of values corresponding to the input values X[ 1], X[ 2],..., X[N] of the input layer, each multiplied by a weight Wij. The output values Z[ 1], Z[ 2],..., Z[M] of the respective nodes in the intermediate layer are each calculated as a return value of an activation function F having the input value U[j] of this node as a parameter. In this embodiment, a sigmoid function is used as the activation function F.In FIG. 3, input values of the three nodes in the output layer are denoted by Y[ 1], Y[ 2], and Y[ 3]. As Y[ 1], a sum of values corresponding to the output values Z[j] of the respective nodes in the intermediate layer is input each multiplied by a weight Vj 1. as Y[ 2], a sum of values corresponding to the output values Z[j] of the respective nodes in the intermediate layer is input each multiplied by a weight Vj 2. As Y[3], a sum of values corresponding to the output values Z[j] of the respective nodes in the intermediate layer is input, each multiplied by a weight Vj 3. In this neural network 34, the input values Y[ 1], Y[ 2], and Y[ 3] of the respective nodes in the output layer are directly used as output values of these nodes. Symbols Y[ 1], Y[ 2], and Y[ 3] represent the estimated temperature t 1 in the section P 1, the estimated temperature t 2 in the section P 2, and the estimated temperature t 3 in the section P 3, respectively.learning of the neural networkNext, a method of generating such a neural network 34, that is, learning the neural network 34, will be described. The learning of the neural network 34 is performed using a computer for learning.For learning of the neural network 34, teacher data is generated using a vehicle for learning in which temperature sensors are installed at the portions P 1 to P 3 of the turbocharger 20. In the vehicle for learning, in order to generate the teacher data, the driving state amounts used as an input to the neural network 34 and the temperatures at the portions P 1 to P 3 are measured. This measurement is performed in various running states, and a large number of data sets are generated, which respectively combine measurement values of the temperatures at the sections P 1 to P 3 and measurement values of the running state quantities measured at the time of measurement of these temperatures.The learning of the neural network 34 is performed using the teacher data composed of a large number of data sets thus generated. Specifically, first, the measured values of the driving state quantities in the data set are input to the input layer of the neural network 34 as the values X[ 1] to X[N]. Then, the values of the weights Wij, Vj1, Vj2 and Vj3 are corrected using an error feedback method, so that errors between the values Y[1] to Y[3] output from the neural network 34 in response to this input and the measurement values of the temperatures at the portions P1 to P3 become smaller. This process for correcting the weights Wij, Vj1, Vj2 and Vj3 is repeated until the errors become less than or equal to a predetermined value. When the errors become less than or equal to the predetermined value, it is determined that the learning of the neural network 34 has been completed. The memory 32 of the electronic control unit 30 of each vehicle stores the thus learned neural network 34, that is, a learned network.Internal Temperature Estimation ProcessHereinafter, the details of the internal temperature estimation process F 3 will be described with reference to FIG. 4. In the internal temperature estimation process F 3, the estimated temperatures t 1 to t 3 at the respective portions P 1 to P 3 are calculated using the neural network 34.FIG. 4 is a flowchart of a processing routine involved in the internal temperature estimation process F 3. The process of this routine is repeatedly executed by the arithmetic processing circuit 31 at a cycle of a predetermined time T during the operation of the engine 10.When the process of this routine is started, first, at step S 100, measured values of the driving state amounts to be input to the neural network 34 are read. Specifically, measured values of time series data of the vehicle speed V, the engine speed NE, the accelerator operation amount ACC, the fuel injection amount QINJ, the boost pressure PB, and the intake air flow rate GA, and measured values of instantaneous value data of the outside air temperature TH1 and the intake air temperature TH2 are read.Subsequently, in step S 110, the measurement values of the driving state amounts read in step S 100 are set as the values of the input values X[ 1] to X[N] of the respective nodes in the input layer of the neural network 34. In the next step S 120, values of Y[ 1] to Y[ 3] corresponding to the output of the neural network 34 are calculated. The values Y[ 1], Y[ 2], and Y[ 3] calculated here respectively represent the estimated temperatures t 1, t 2, and t 3 at the portions P 1 to P 3 at the time of measuring the driving state quantities input to the neural network 34.In this embodiment, a range of values that the wall surface temperature of the oil passage 28 can take at the portions P 1 to P 3 during the operation of the engine 10 is divided into a plurality of temperature ranges. In the following description, these temperature ranges are denoted by R[ 1], R[ 2],..., R[L] in the order of increasing temperature. The symbol "L" here represents the number of temperature ranges. Further, in this embodiment, counters indicating the number of times of calculation of the estimated temperature for the respective temperature ranges R[ 1] to R[L] are set for each of the sections P 1 to P 3. In the following description, counters indicating the number of times of calculation of the estimated temperature t 1 in the respective temperature ranges R[ 1], R[ 2],..., R[L] for the section P 1 are referred to as counters C 1[ 1], C 1[L]. Similarly, counters indicating the number of times of calculation of the estimated temperature t 2 in the respective temperature ranges R[ 1], R[ 2],..., R[L] for the section P 2 are referred to as counters C 2[ 1], C 2[L]. Further, counters indicating the number of times of calculation of the estimated temperature t 3 in the respective temperature ranges R[ 1], R[ 2],..., R[L] for the section P 3 are referred to as counters C 3[ 1], C 3[L].In step S130, the following values are incremented. In step S 130, the value of the counter C 1[ 1] for the temperature range R[ 1] including a calculated value of the estimated temperature t 1 is increased. Further, in step S 130, both the value of the counter C 2[ 1] for the temperature range R[ 1] including a calculated value of the estimated temperature t 2 and the value of the counter C 3[ 1] for the temperature range R[ 1] including a calculated value of the estimated temperature t 3 are increased. Thereafter, the process of this routine is ended in the current period.Deposit Amount Estimation Process and Command ProcessHereinafter, the details of the deposition amount estimation process F 4 and the command process F 5 will be described with reference to FIGS. 5 and 6. FIG. 5 is a flowchart of a processing routine involved in the deposit amount estimation process F 4 and the command process F 5. The series of processes shown in FIG. 5 is executed by the arithmetic processing circuit 31 every time the vehicle travels a predetermined distance D.When the process of this routine is started, first, in step S 200, the values of the counters C 1[ 1] to C 1[L], C 2[ 1] TO C 2[L], and C 3[ 1] to C 3[L] are read. Subsequently, in step S 210, the values of increase amounts Δ 1 to Δ 3 of the amounts of oil coke deposited at the respective portions P 1 to P 3 during a period from the last execution to the current execution of this routine are calculated. The increase amounts Δ 1 to Δ 3 are calculated as values that respectively satisfy the relationships of Expressions (1) to (3). The symbols SC[ 1] to SC[L] in Expressions (1) to (3) represent coking rates set for the respective temperature ranges. The value of the coking rate SC[ 1] represents an amount of oil coke that deposits on the wall surface of the oil passage 28 when a state in which the wall surface temperature of the oil passage 28 has become a temperature within the corresponding temperature range R[ 1] is maintained for the predetermined time T. [Expression 1]FIG. 6 shows relationships among the temperature of the wall surface of the oil passage 28, a holding time of the temperature, and the amount of oil coke deposited on the wall surface of the oil passage 28 when the wall surface of the oil passage 28 is maintained at a constant temperature. Oil coke is produced when oil is heated above a certain temperature. In the following description, a temperature at a lower limit of a range of the oil temperature at which oil coke is generated is referred to as a coking start temperature Tx. When the wall surface temperature is in a range lower than the coking start temperature Tx, the oil coke deposition amount is zero regardless of the holding time. On the other hand, in a range of the wall surface temperature not lower than the coking start temperature Tx, the ratio of the deposition amount to the holding time, that is, the coking rate, increases as the wall surface temperature increases. The value of the coking rate SC[ 1] in each temperature range R[ 1] is set based on these relationships. Thus, in the temperature ranges on the lower temperature side to the coking start temperature Tx, the value of the coking rate is set to be zero. In the temperature ranges on the higher temperature side to the coking start temperature Tx, a value larger than the value of the coking rate is set for a temperature range farther on the high temperature side.In the next step S 220, the values of estimated coke deposition amounts M 1 to M 3 corresponding to estimation values of the oil coke amounts deposited on the portions P 1 to P 3 respectively are updated based on the increase amounts Δ 1 to Δ 3 calculated in step S 210. Here, the estimated coke deposition amounts M 1 to M 3 are each updated such that a sum of a value before the update and a corresponding amount of the increase amounts Δ 1 to Δ 3 added thereto constitutes a value after the update.In the inside temperature estimation process F 3, it is assumed that values that accurately reflect the wall surface temperatures at the portions P 1 to P 3 are calculated as the values of the estimated temperatures t 1 to t 3. It is also assumed that the wall surface temperatures at the portions P 1 to P 3 are kept constant during a period from the time of calculation of the estimated temperatures t 1 to t 3 by the processing routine of FIG. 4 to the next execution of this processing routine. In this case, the amount of oil coke deposited during this period at the portion P 1 increases by an amount corresponding to the value of the coking rate SC[ 1] in the temperature range R[ 1] including the value of the estimated temperature t 1. Similarly, the amount of oil coke deposited on the portion P 2 increases by an amount corresponding to the value of the coking rate SC[ 1] in the temperature range R[ 1] including the value of the estimated temperature t 2. Further, the amount of oil coke deposited on the portion P 3 increases by an amount corresponding to the value of the coking rate SC[ 1] in the temperature range R[ 1] including the value of the estimated temperature t 3.The increase amount Δ 1 is obtained, for each of the values of the estimated temperature t 1 calculated during a period in which the vehicle travels the predetermined distance D, as a value that integrates the coking rate SC[ 1] in the temperature range R[ 1] including the calculated value of this estimated temperature t 1. The increase amounts Δ 2 and Δ 3 are obtained in the same manner. Values each integrating the corresponding one of the increase amounts Δ 1 to Δ 3 obtained each time the vehicle travels the predetermined distance D are calculated as the values of the estimated coke deposition amounts M 1 to M 3. In this manner, in this embodiment, the estimated coke deposition amounts M 1 to M 3 corresponding to estimation values of the oil coke amounts deposited on the portions P 1 to P 3, respectively, are calculated as values integrating the amounts by which the oil coke amounts deposited on the portions P 1 to P 3 increase during each predetermined time T and obtained from the calculated values of the estimated temperatures t 1 to t 3. Thus, in this embodiment, the estimated coke deposition amounts M 1 to M 3 are calculated substantially by calculating the increase amounts of the oil coke deposition amounts based on the calculated values of the estimated temperatures t 1 to t 3 and then integrating these increase amounts.When the estimated coke deposition amounts M 1 to M 3 are thus calculated, the process proceeds to step S 230. In step S 230, the values of the counters C 1[ 1] to C 1[L], C 2[ 1] TO C 2[L], and C 3[ 1] to C 3[L] are reset to zero, and then the process proceeds to step S 240.When the process proceeds to step S 240, it is determined in step S 240 whether or not one or more of the estimated coke deposition amounts M 1 to M 3 at the respective portions P 1 to P 3 is / are equal to or larger than a predetermined warning threshold α. If the determination result is affirmative (Yes), the process proceeds to step S 250. In step S 250, a command signal instructing that the warning lamp 33 is lit is output, and then the current processing of the routine is ended. On the other hand, when the determination result in step S 240 is negative (No), the process proceeds to step S 260.When the process proceeds to step S 260, it is determined in step S 260 whether or not one or more of the estimated coke deposition amounts M 1 to M 3 at the respective portions P 1 to P 3 is / are equal to or greater than a predetermined output restriction threshold value β. A value smaller than the warning threshold α is set as the output restriction threshold β. When the determination result in step S 260 is affirmative (Yes), the process proceeds to step S 270. In step S 270, a command signal instructing that the machine output is restricted is output, and then the current processing of the routine is ended. On the other hand, when the determination result in step S 260 is negative (No), the current processing of the routine is directly ended.The values of the estimated coke deposition amounts M 1 to M 3 are stored and held in the memory 32 even when the electronic control unit 30 is not in operation. When an oil coke deposited inside the turbocharger 20 is removed by maintenance or the turbocharger 20 is replaced with a new one, the values of the estimated coke deposition amounts M 1 to M 3 stored in the storage 32 are each reset to zero.In determining the required output PE* in the required output determination process F 1, the electronic control unit 30 sets a maximum value in a setting range of the value of the required output PE* to a smaller value when a command signal for restricting the output is output as compared to when the command signal is not output. In this embodiment, the output of the engine 10 is thus restricted.In this embodiment, the processes from step S 200 to step S 230 of FIG. 5 are processes corresponding to the deposition amount estimation process F 4. The processes from step S 240 to step S 270 of FIG. 5 are processes corresponding to the command process F 5.Operation and Effects of EmbodimentThe operation and effects of the embodiment will be described.The amounts of oil coke deposited on the portions P 1 to P 3 of the oil passage 28 provided inside the turbocharger 20 depend on the wall surface temperatures at the portions P 1 to P 3. The wall surface temperatures at the portions P 1 to P 3 change with the running state of the vehicle. There are a large number of driving state quantities that affect the wall surface temperature, and the relationship between each driving state quantity and the inside temperature is complicated. In view of this, in the embodiment, the relationships between the traveling state amounts and the wall surface temperatures at the neural network 34-shaped portions P 1 to P 3 are learned by machine learning. Using this neural network 34, the estimated temperatures t 1 to t 3 corresponding to estimation values of the wall surface temperatures at the portions P 1 to P 3 are calculated from the measurement values of the driving state amounts. Thus, the estimated temperatures t 1 to t 3 are calculated as values that accurately reflect the wall surface temperatures at the portions P 1 to P 3.Further, the increase amounts of the oil coke quantities deposited on the portions P 1 to P 3 during the predetermined time T are obtained from the calculated values of the estimated temperatures t 1 to t 3. In the embodiment, the estimated coke deposition amounts M 1 to M 3 corresponding to estimation values of the oil coke amounts deposited on the respective portions P 1 to P 3 are calculated as values integrating the increase amounts of the oil coke deposition amounts obtained from the calculated values of the estimated temperatures t 1 to t 3. Thus, the values of the estimated coke deposition amounts M 1 to M 3 are calculated as values that accurately reflect the actual oil coke amounts deposited on the respective portions P 1 to P 3.The oil coke deposition amount estimating device in a turbocharger of the above embodiment can achieve the following effects:(1) In the embodiment, the relationships between the driving state amounts and the wall surface temperatures at the portions P 1 to P 3 are learned in the form of the neural network 34. In the internal temperature estimation process F 3, using this neural network 34, the estimated temperatures t 1 to t 3 corresponding to estimation values of the wall surface temperatures at the portions P 1 to P 3 are calculated from the measurement values of the driving state amounts. Further, in the embodiment, in the deposition amount estimation process F 4, values integrating the increase amounts obtained from the calculation result of the estimated temperatures t 1 to t 3 are calculated as the values of the estimated coke deposition amounts M 1 to M 3 corresponding to estimation values of the oil coke amounts deposited at the portions P 1 to P 3, respectively. Thus, the amounts of oil coke deposited inside the turbocharger 20 can be accurately estimated.(2) In the embodiment, a command signal instructing that the output of the engine 10 is restricted is output when one of the estimated coke deposition amounts M 1 to M 3 becomes equal to or larger than the output restriction threshold value β. When the output of the engine 10 becomes high, the temperature of the exhaust gas flowing into the turbine wheel 24 of the turbocharger 20 becomes high, and also the wall surface temperatures at the portions P 1 to P 3 become high. The deposition of oil coke on the wall surface of the oil passage 28 is further promoted as the wall surface temperature is higher. Therefore, restricting the output of the machine 10 can mitigate the increase in wall surface temperature and thereby reduce the likelihood of further deposition of oil coke. In practice, however, restricting the discharge of the engine 10 in a state where the deposition of oil coke has not yet advanced so far as to require restricting the discharge would be unpleasant for the user of the vehicle. In this regard, the amounts of oil coke deposited on the portions P 1 to P 3 can be accurately estimated in the embodiment. Therefore, a restriction can be made on the output of the machine 10 for reducing the possibility of depositing oil coke at an appropriate time.(3) In the embodiment, when one of the estimated coke deposition amounts M 1 to M 3 becomes equal to or larger than the warning threshold α, a command signal instructing lighting of the warning lamp 33 is output. The user of the vehicle is informed that the turbocharger 20 requires maintenance via lighting of the warning lamp 33. In practice, issuing this notification in a state where the deposition of oil coke has not yet advanced so far as to require maintenance would be unpleasant for the user of the vehicle. In this regard, the amounts of oil coke deposited on the portions P 1 to P 3 can be accurately estimated in the embodiment. Therefore, this notification can be output at an appropriate time when the deposition of oil coke has advanced to an extent where maintenance of the turbocharger 20 is required.(4) The internal temperature of the turbocharger 20 is determined by a heat balance between a heat amount that the turbocharger 20 receives from the exhaust gas flowing inside the turbine housing 21 and a heat amount that dissipates a traveling wind hitting the turbocharger 20 from the turbocharger 20. Of these heat quantities, the heat quantity that the turbocharger 20 receives from the exhaust gas is determined by the temperature and the flow rate of the exhaust gas. Main running state quantities associated with the temperature and the flow rate of the exhaust gas include the engine speed NE, the accelerator operation amount ACC, the fuel injection amount QINJ, the boost pressure PB, the intake air flow rate GA, and the intake air temperature TH2. Meanwhile, the amount of heat that the traveling wind hitting the turbocharger 20 dissipates from the turbocharger 20 is determined by the flow rate and the temperature of the traveling wind hitting the turbocharger 20. The flow rate of the traveling wind hitting the turbocharger 20 becomes higher as the vehicle speed V increases. The temperature of the traveling wind hitting the turbocharger 20 corresponds to the outside air temperature TH 1. Thus, the vehicle speed V, the engine speed NE, the accelerator operation amount ACC, the fuel injection amount QINJ, the boost pressure PB, the intake air flow rate GA, the outside air temperature TH1, and the intake air temperature TH2 correspond to vehicle running state amounts having significant influences on the internal temperature of the turbocharger 20. In the embodiment, the vehicle speed V, the engine speed NE, the accelerator operation amount ACC, the fuel injection amount QINJ, the boost pressure PB, the intake air flow rate GA, the outside air temperature TH 1, and the intake air temperature TH 2, which are closely related to the inside temperature, are used as the vehicle running state amounts to be input to the neural network 34. Therefore, the neural network 34 may be configured as a model capable of accurately estimating the internal temperature of the turbocharger 20.(5) A certain time is required until changes in the vehicle running state amounts are reflected in the internal temperature of the turbocharger 20. Of the above-described driving state quantities to be input to the neural network 34, the vehicle speed V, the engine speed NE, the accelerator operation amount ACC, the fuel injection amount QINJ, the boost pressure PB, and the intake air flow rate GA correspond to driving state quantities that largely change during a travel of the vehicle. In the embodiment, for the vehicle speed V, the engine speed NE, the accelerator operation amount ACC, the fuel injection amount QINJ, the boost pressure PB, and the intake air flow rate GA, time series data of each of these quantities is used as the input to the neural network 34. Thus, the internal temperature of the turbocharger 20 can be estimated as a value reflecting a delay with which changes in the driving state amounts at the internal temperature are reflected.(6) The outside air temperature TH 1 and the intake air temperature TH 2 do not greatly change over a short time. Therefore, in obtaining time series data from each of the outside air temperature TH 1 and the intake air temperature TH 2, values in these time series data almost correspond to the same values. For this reason, using a single measurement value instead of time series data of each of the outside air temperature TH 1 and the intake air temperature TH 2 as the input to the neural network 34 has little influence on the estimation result of the inside temperature. On the other hand, when the number of values input to the neural network 34 increases, the structure of the neural network 34 becomes complicated accordingly, so that a longer time is required for learning and calculating the estimated temperatures t 1 to t 3. In the embodiment, for each of the outside air temperature TH 1 and the intake air temperature TH 2, a single measurement value is used as the input to the neural network 34 instead of time-series data, thereby avoiding unnecessary complication of the structure of the neural network 34.The embodiment can be implemented with the following changes applied thereto. The embodiment and the following modified examples may be implemented in combination within such a range that no technical inconsistency occurs.In the above embodiment, time series data of the vehicle speed V, the engine speed NE, the accelerator pedal operation amount ACC, the fuel injection amount QINJ, the boost pressure PB, and the intake air flow rate GA are input to the neural network 34, but a single measurement value of each of these may be input instead.In the above embodiment, the vehicle speed V, the engine speed NE, the accelerator operation amount ACC, the fuel injection amount QINJ, the boost pressure PB, the intake air flow rate GA, the outside air temperature TH1, and the intake air temperature TH2 are used as the running state amounts of the vehicle to be input to the neural network 34. One or more of these driving state quantities may be omitted in the input for the neural network 34, or driving state quantities of the vehicle other than these quantities may be added in the input for the neural network 34.In the above embodiment, the values of the estimated coke deposition amounts M1 to M3 are updated each time the vehicle travels the predetermined distance D. The values of the estimated coke deposition amounts M 1 to M 3 may be updated at a cycle other than this. For example, the values of the estimated coke deposition amounts M 1 to M 3 may be updated each time the estimated temperatures t 1 to t 3 are calculated. In this case, the estimated coke deposition amounts M 1 to M 3 are each updated such that a sum of a value before the update and the value of the coking rate SC[ 1] in the temperature range R[ 1] including the corresponding ones of the estimated temperatures t 1 to t 3 added thereto constitutes a value after the update.The user can be informed that maintenance of the turbocharger 20 is required via a method other than lighting up the warning lamp 33. For example, a command signal for the notification may be sent from the vehicle to the user's mobile terminal via a wide area network, and the notification may be made from the mobile terminal.If oil deteriorates, it tends to become coked. Therefore, the degree of deterioration of oil can be estimated from a route that the vehicle has traveled since an oil change, etc., and this estimation result can be reflected on the calculation result of the estimated coke deposition amounts M 1 to M 3. For example, based on the oil deterioration degree, the values of the coking rates SC[ 1] to SC[L] in the corresponding temperature ranges R[ 1] to R[L] may be calculated as values that become larger as the oil deterioration degree increases.In the above embodiment, the wall surface temperatures at the three portions P1 to P3 of the oil passage 28 are estimated as the internal temperature of the turbocharger 20, and the amounts of oil coke deposited at the portions P1 to P3 are estimated from these estimated values. The positions and the number of these portions at which the internal temperature of the turbocharger 20 and the oil coke deposition amounts are estimated may be changed as necessary.The upper limit value of the oil coke deposition amounts allowed inside the turbocharger 20 may vary between portions of the turbocharger 20. In these cases, different values of the warning threshold value α and the output restriction threshold value β should be set for different sections.In the above embodiment, the electronic control unit 30 installed in the vehicle performs estimation of the internal temperature and the coke deposition amounts. This estimation may be performed in a data center external to the vehicle. In this case, measurement values of travel state quantities are transmitted from the vehicle to the data center, and estimation of the internal temperature and the oil coke deposition amounts based on the transmitted measurement values is performed at the data center. Then, the estimation result of the oil coke deposition amounts or a command signal based on the estimation result is sent from the data center to the vehicle.In the above embodiment, the neural network 34 having only one intermediate layer is used, but the neural network 34 may be configured to have a plurality of intermediate layers.

Claims

An oil coke deposition amount estimation device in a turbocharger that estimates an oil coke amount deposited inside a turbocharger installed in a vehicle, and comprises an execution device and a storage device, wherein: the storage device stores a neural network to which driving state quantities indicating a driving state of the vehicle are input and from which an internal temperature of the turbocharger is output, the neural network being learned using an internal temperature measurement value and driving state quantity measurement values measured at a measurement time of the internal temperature as teacher data; and the execution device executes an internal temperature estimation process for calculating an output of the neural network as an estimation value of the internal temperature using the measurement values of the driving state amounts as an input, and a deposit amount estimation process for calculating an increase amount of the deposit amount based on the estimation value of the internal temperature and calculating an estimation value of the deposit amount as a value integrating the increase amount.The estimating device for an oil coke deposition amount in a turbocharger according to claim 1, wherein the executing device executes an instruction process for instructing restriction of an engine output when the estimated value of the deposition amount exceeds a predetermined threshold value.The estimating device for an oil coke deposition amount in a turbocharger according to claim 1, wherein the executing device executes a command process for instructing a notification of a user of the vehicle that the turbocharger is in a maintenance requiring state when the estimated value of the deposition amount exceeds a predetermined threshold value.The estimating device for an oil coke deposition amount in a turbocharger according to any one of claims 1 to 3, wherein time-series data of the driving state amounts is used as an input to the neural network.The estimating device for an oil coke deposition amount in a turbocharger according to any one of claims 1 to 4, wherein the driving state amounts include one or more of a vehicle speed, an engine speed, an accelerator operation amount, a fuel injection amount, a boost pressure, an intake air flow rate, an outside air temperature, and an intake air temperature.

Images