An Electronic Control Unit (ECU) adapted to monitor physical states and a method thereof

2. Applicants: a. Name: Bosch Global Software Technologies Private Limited

Nationality: INDIA

Address: 123, Industrial Layout, Hosur Road, Koramangala,

Bangalore - 560095, Karnataka, India b. Name: Robert Bosch GmbH

Nationality: GERMANY

Address: Feuerbach, Stuttgart, Germany

Complete Specification:

The following specification describes and ascertains the nature of this invention and the manner in which it is to be performed Field of the invention

[0001] The present disclosure relates to the field of computation and prognostics of real-world system. In particular, the present disclosure proposes an Electronic Control Unit (ECU) adapted to monitor physical states and a method thereof.

Background of the invention

[0002] The Fourth Industrial Revolution (or Industry 4.0) is transforming and traditional industrial and automotive practices, using modem technologies for prognostics of various machines employed in the set-up. In modern day industrial systems, prognostics and monitoring of machine’s performance has become of utmost importance to predict maintenance and prevent an impending breakdown. Machines as complex as an automobile employ numerous physical systems that need monitoring and diagnosis at regular intervals. Prognostics of these physical systems involves predicting future state of the physical systems by solving stiff ordinary differential equations.

[0003] Integration methods for stiff systems are particularly challenging and are subject to intense research over the past. Running such solvers on an ECU is very challenging because of the compute and memory restrictions on embedded hardware. Broadly, there are two basic methods by which ODE systems can be integrated, firstly we can either use explicit methods or implicit methods, further they can either be run using fixed timesteps or variable timesteps. The explicit methods are used only for non- stiff ODEs but for stiff systems, implicit methods are used. The advantage of implicit methods is the fact that higher time step sizes can be used but it is more compute intensive. For stiff systems, implicit methods are the required choice because explicit methods will require an unfeasibly small time-step. For systems with widely varying dynamics, it may also be required to use variable step solvers. These solvers adjust the time-step sizes while keeping the local error under a user specified tolerance. This means that the solver automatically selects the highest possible time step at a certain point in the simulation. This ensures that the simulation finishes the computation as soon as possible.

[0004] For ECU execution, running an implicit solver poses a variety of challenges. One of the main challenges is to keep computational requirements low and also constant per-compute step. This is needed because on ECUs functions are usually deployed on set rasters (50ms, 100ms etc..), so if a function takes a variable amount of compute resources every time, then we need to allocate resources to the function based on the worst case execution time. Otherwise, we risk deadline resets in case the compute cost is too high on any particular step. Hence there is a need for a framework to keep the per-step compute cost constant and the whole processing efficient and accurate. [0005] Patent application US2018257664 AA titled “Distributed monitoring and control of a vehicle” discloses a distributed system for monitoring and control of a vehicle that includes a supervisory controller with a first computer readable storage media for monitoring and storing a plurality of operational parameters regarding a physical system of the vehicle. The supervisory controller communicates with a server via two different communications networks. Method steps are provided for characterizing and predicting functional details of a system state of the physical system using the model parameters and at least one operational parameter of the physical system, and for using values obtained by the server regarding a plurality of different vehicles in order to improve the monitoring and control of the vehicle. A method is also provided to determine and report any operational parameters miss a corresponding performance target. A method is also provided for changing the storage or transmission of operational parameters based on their relative importance.

Brief description of the accompanying drawings

[0006] An embodiment of the invention is described with reference to the following accompanying drawings:

[0007] Figure 1 depicts an Electronic Control Unit (ECU) adapted to monitor physical state of a machine;

[0008] Figure 2 illustrates method steps (200) to monitor physical state of a machine. Detailed description of the drawings

[0009] Figure 1 depicts an Electronic Control Unit (ECU (103)) adapted to monitor physical state of a machine. A machine encompasses many physical systems that resides in an industrial or automotive set-up. In an exemplary embodiment of the present invention depicted in figure 1, the physical system is depicted as the oxidation catalyst (104) inside an exhaust gas treatment system (100) in a vehicle whose physical state is monitored by the ECU (103). The exhaust gas after-treatment system (100) comprises the diesel oxidation catalyst (104), a diesel particulate filter (106), an inlet temperature sensor (102), temperature sensor (105), and at least a diesel particulate filter (106) among other components such as a delta pressure sensor (107) connected across the diesel particulate filter (106) of the exhaust gas system known to a person skilled in the art. The components are mounted inside an exhaust gas pipe (101) downstream of an exhaust manifold of the internal combustion engine. The sensors are in communication with the ECU (103).

[0010] The ECU (103) is an embedded system that is implemented as any or a combination of: one or more microchips or integrated circuits interconnected using a parent board, hardwired logic, software stored by a memory device and executed by a microprocessor, firmware, an application specific integrated circuit (ASIC), and/or a field programmable gate array (FPGA). In Industrial or automotive systems, ECU (103) is basically a logic circuitry implemented as one or more microcontrollers, digital signal processors, central processing units, state machines, logic circuitries, and/or any component that operates on input from sensors based on operational instructions to get a meaningful result.

[0011] The machine (vehicle in the exemplary example) comprises a plurality of sensors which include but are not limited to air-fuel ratio meter, engine speed sensor, throttle position sensor, crank position sensor, cam position sensor, knock sensor, Manifold Absolute Pressure or MAP Sensor, Mass Air Flow or MAF Sensor, Oxygen or Lambda sensor, fuel pressure sensor, vehicle speed sensor and the like. The plurality of sensors measure a plurality of parameters and send the information as a plurality of inputs to the ECU (103). Hence, the plurality of inputs comprise at least one or more from the group of engine operating parameters and vehicle operating parameters.

[0012] The Electronic Control Unit (ECU (103)) adapted to monitor physical state of the machine is configured to: generate a Jacobian matrix based on the plurality of inputs and a plurality of values retrieved from mathematical model running inside the ECU; find a minimum set of independent columns using graph coloring technique for sparse Jacobian matrix; select at least one colour column for the sparse Jacobian matrix or at least one column for dense Jacobian Matrix at every time step; the ECU (103) selects the colour column or the column based on statistical analysis at every time step; compute the selected colour column of the Jacobian Matrix; solve a stiff ordinary differential equation in dependance of the Jacobian matrix to monitor a physical state of the machine. The functionality of the ECU (103) is explained better in accordance with figure 2 and method steps 200.

[0013] It should be understood at the outset that, although exemplary embodiments are illustrated in the figures and described below, the present disclosure should in no way be limited to the exemplary implementations and techniques illustrated in the drawings and described below.

[0014] Figure 2 illustrates method steps (200) to monitor physical state of a machine. The machine and the corresponding components of the physical system whose physical state is to be monitored have been elucidated in accordance with the exemplary embodiment in figure 1. The method is performed by the ECU (103) that receives a plurality of inputs from the plurality of afore-mentioned sensors. The inputs vary in accordance with the type of physical system being monitored. In the exemplary embodiment wherein the physical system to be monitored is the oxidation catalyst (104) of the exhaust gas system (100), the inputs comprise values from temperature sensor, pressure sensor and other values such as molar mass faction of the catalyst retrieved from mathematical model running inside the ECU. In alternate embodiment wherein the physical system being controlled is by using Model Predictive Control on chemical processes in an industrial setting. The input comprises flow rate measurement, pressure sensor, electrochemical sensors for gas concentration measurements etc.

[0015] Method step 201 comprises generating a Jacobian matrix by the ECU (103) based on the plurality of inputs received from the plurality of sensors and a plurality of values retrieved from mathematical models running inside the ECU. The plurality of inputs vary as per the physical system, for example the plurality of inputs for monitoring of a diesel oxidation catalyst (104) as physical system comprises temperature, pressure, molar mass of the catalyst and likewise.

[0016] Method step 202 comprises finding a minimum set of independent columns using graph coloring technique for a sparse Jacobian Matrix. A person skilled in the art would appreciate that Dense Jacobian matrices won’t require this method step.

[0017] Method step 203 comprises selecting at least one colour column for sparse Jacobian matrix or at least a column for dense Jacobian matrix at every time step. The selection of a colour column is based on a statistical method or any heuristic. For example, for an exemplary embodiment, the colour column with the highest average relative change of state at every time step is selected. Method step 204 comprises computing the selected colour column of the Jacobian Matrix. A person skilled in the art will appreciate that for dense Jacobians that do not require graph coloring technique we will not compute any colours but will compute at least one column at every timestep.

[0018] Method step 205 comprises solving a stiff ordinary differential equation by means of the ECU (103) in dependance of the Jacobian matrix to monitor a physical state of the machine.

[0019] The method steps are explained via an example in context of monitoring of a diesel oxidation catalyst (104). In this we used a solver which was particularly suitable to run with inexact Jacobians (Rosenbrock 2ns and 3rd order W solver) along with a heuristic where we only compute the colors of the matrix which have corresponding state variables which have on average changed the most from the previously computed timestep (method step 203 and method step 204).

[0020] As per method step 202 we have used the graph coloring technique to find the independent number of columns of the system. For example, assume 11 set of such independent columns were found. These of columns are codes with 11 different colours for the purposes of clarity. Conventionally this means that for computing the entire Jacobian, the number of function calls to the DOC model (compute intensive function which gives the derivatives of the ODE) is 11 at each time-step.

[0021] However, by using the heuristics (method step 203 and 204 ) as defined, we were able to reduce the number of function calls to 1. This is more than an order of magnitude savings in compute cost. Now as per the present invention, every step of the solver recomputes one color of the Jacobian at every time step in accordance with method step 204. Hence, the proposed method doesn’t compute the entire Jacobian matrix every time step. Instead, we compute only parts of this Jacobian every step. For example, we can compute the red colored columns of the matrix at timestep 1 , the yellow-colored columns of the matrix at time-step 2, the green colored columns of the matrix at time step 3 etc. depending upon the colour column with the highest average relative change of state at every time step is selected. This ensures that after a few time-steps, the entire Jacobian is recomputed, but was never fully re-computed at any particular step.

[0022] The proposed Electronic Control Unit (ECU (103)) adapted to monitor physical systems and the method thereof ensures constant compute cost per-step (as opposed to not computing anything in most of the steps and re-computing the entire Jacobian at a certain step where the values are very inaccurate, as done in the case of processor simulations). This is especially important in an embedded system like the ECU which runs on a fixed raster (1 ms or the like).

[0023] When running on a fixed raster, we need to ensure that the compute time taken by any particular function at every time- step is constant i.e. if we schedule a function to execute on a 1ms raster for example, we need to ensure that the time taken for computation is less than 1ms, otherwise we risk getting deadline resets on the computation (All computations on the ECU will stop). This is not the case in processors, which have operating system mechanisms to interrupt our executing function and context switch to other tasks based on schedulers.

[0024] A person skilled in the art will appreciate that while these method steps describes only a series of steps to accomplish the objectives, these methodologies may be implemented with adaptations and modification as per the physical system to be monitored. It must be understood that the embodiments explained in the above detailed description are only illustrative and do not limit the scope of this invention. Any modification to the Electronic Control Unit (ECU (103)) adapted to monitor physical systems and the method thereof are envisaged and form a part of this invention. The scope of this invention is limited only by the claims.