Process model error correction method and system

Abstract

A method is provided for a virtual sensor system. The method may include establishing a first process model indicative of interrelationships between a plurality of input parameters and a plurality of output parameters and establishing a second process model indicative of interrelationships between at least the plurality of input parameters and modeling errors of the first process model. The method may also include operating the first process model to generate values of the plurality of output parameters and simultaneously operating the second model to generate estimated deviations between the values of the plurality of output parameters and desired values of the plurality of output parameters. Further, the method may include compensating the values of the plurality of output parameters with the estimated deviations to generate the desired values of the plurality of output parameters.

Description

TECHNICAL FIELD

This disclosure relates generally to process modeling techniques and, more particularly, to methods and computer systems for process modeling error correction.

BACKGROUND

Predictive modeling refers to generating a model from a given set of data records of both input parameters and output parameters and predicting actual output parameters corresponding to actual input parameters based on the model. Predictive models may be built by using various methods from data for many different families of models, such as decision trees, decision lists, linear equations, and neural networks.

The data records used to build a model are known as training data records. In certain situations, the training data records may be unable to cover the entire input space of the input parameters or the training data records may be discrete such that uniform relationships represented by a single predictive model between input parameters and output parameters may be unavailable across the entire input space and/or output space.

Techniques such as boosting and/or bagging may be used to divide the input space and/or output space by applying a large number of mathematical models. Each mathematical model may only cover a part of the input space and/or output space. For example, U.S. Pat. No. 6,546,379 (the '379 patent) issued to Hong et al. on Apr. 8, 2003, discloses a cascade boosting method for boosting predictive models for resolving the interpretability problem of previous boosting methods and mitigating the fragmentation problem when applied to decision trees.

However, such conventional techniques, while involving a large number of models, may cause coarse transitions from the large number of models. These coarse transitions may reduce the accuracy of the overall predictive model and may also cause confusion for the users of the overall predictive model.

Methods and systems consistent with certain features of the disclosed systems are directed to solving one or more of the problems set forth above.

SUMMARY OF THE INVENTION

One aspect of the present disclosure includes a method for a virtual sensor system. The method may include establishing a first process model indicative of interrelationships between a plurality of input parameters and a plurality of output parameters and establishing a second process model indicative of interrelationships between at least the plurality of input parameters and modeling errors of the first process model. The method may also include operating the first process model to generate values of the plurality of output parameters and simultaneously operating the second model to generate estimated deviations between the values of the plurality of output parameters and desired values of the plurality of output parameters. Further, the method may include compensating the values of the plurality of output parameters with the estimated deviations to generate the desired values of the plurality of output parameters.

Another aspect of the present disclosure includes a computer system. The computer system may include a database configured to store information relevant to a virtual sensor system and a processor. The processor may be configured to operate a first process model to generate values of a plurality of output parameters and to simultaneously operate a second model to generate estimated deviations between the values of the plurality of output parameters and desired values of the plurality of output parameters. The processor may also be configured to compensate the values of the plurality of output parameters with the estimated deviations to generate the desired values of the plurality of output parameters.

Another aspect of the present disclosure includes a work machine. The work machine may include a power source configured to provide power to the work machine, a control system configured to control the power source, and a virtual sensor system. The virtual sensor system may include a first process model and a second process model. Further, the virtual sensor system may be configured to operate the first process model to generate values of a plurality of sensing parameters and to simultaneously operate the second model to generate estimated deviations between the values of the plurality of sensing parameters and desired values of the plurality of sensing parameters. The virtual sensor system may also be configured to compensate the values of the plurality of sensing parameters with the estimated deviations to generate the desired values of the plurality of sensing parameters. The control system may control the power source based upon the desired values of the plurality of sensing parameters.

Another aspect of the present disclosure includes a computer-readable medium for use on a computer system. The computer-readable medium may include computer-executable instructions for performing a method. The method may include establishing a first process model indicative of interrelationships between a plurality of input parameters and a plurality of output parameters and establishing a second process model indicative of interrelationships between at least the plurality of input parameters and modeling errors of the first process model. The method may also include operating the first process model to generate values of the plurality of output parameters and simultaneously operating the second model to generate estimated deviations between the values of the plurality of output parameters and desired values of the plurality of output parameters. Further, the method may include compensating the values of the plurality of output parameters with the estimated deviations to generate the desired values of the plurality of output parameters.

BRIEF DESCRIPTION OF THE DRAWINGS

FIG. 1 illustrates an exemplary work machine in which features and principles consistent with certain disclosed embodiments may be incorporated;

FIG. 2 illustrates a block diagram of an exemplary virtual sensor system consistent with certain disclosed embodiments;

FIG. 3 illustrates a logical block diagram of an exemplary virtual sensor system consistent with certain disclosed embodiments;

FIG. 4 illustrates a flowchart diagram of an exemplary virtual sensor process model generation and optimization process consistent with certain disclosed embodiments;

FIG. 5A illustrates a logical block diagram of an exemplary virtual sensor system in serial arrangement consistent with certain disclosed embodiments;

FIG. 5B illustrates a logical block diagram of another exemplary virtual sensor system in parallel arrangement consistent with certain disclosed embodiments; and

FIG. 6 shows a flowchart diagram of an exemplary virtual sensor system creation and operation process consistent with certain disclosed embodiments.

DETAILED DESCRIPTION

Reference will now be made in detail to exemplary embodiments, which are illustrated in the accompanying drawings. Wherever possible, the same reference numbers will be used throughout the drawings to refer to the same or like parts.

FIG. 1 illustrates an exemplary work machine 100 in which features and principles consistent with certain disclosed embodiments may be incorporated. Work machine 100 may include a virtual sensor system 130, which may be implemented by using one or more process models with error correction capabilities. Work machine 100 may refer to any type of fixed or mobile machine that performs some type of operation associated with a particular industry, such as mining, construction, farming, transportation, etc. and operates between or within work environments (e.g., construction site, mine site, power plants and generators, on-highway applications, etc.). Non-limiting examples of mobile machines include commercial machines, such as trucks, cranes, earth moving vehicles, mining vehicles, backhoes, material handling equipment, farming equipment, marine vessels, aircraft, and any type of movable machine that operates in a work environment. Work machine 100 may also include any type of commercial vehicle such as cars, vans, and other vehicles. Although, as shown in FIG. 1, work machine 100 is an earth handling type work machine, it is contemplated that work machine 100 may be any type of work machine.

Work machine 100 may also include an engine 110, an engine control module (ECM) 120, physical sensors 140 and 142, and a data link 150. Engine 110 may include any appropriate type of engine or power source that generates power for work machine 100, such as an internal combustion engine or fuel cell generator. ECM 120 may include any appropriate type of engine control system configured to perform engine control functions such that engine 110 may operate properly. ECM 120 may include any number of devices, such as microprocessors or microcontrollers, memory modules, communication devices, input/output devices, storages devices, etc., to perform such control functions. Further, ECM 120 may also control other systems of work machine 100, such as transmission systems, and/or hydraulics systems, etc. Computer software instructions may be stored in or loaded to ECM 120. ECM 120 may execute the computer software instructions to perform various control functions and processes.

ECM 120 may be coupled to data link 150 to receive data from and send data to other components, such as engine 110, physical sensors 140 and 142, virtual sensor system 130, and/or any other components (not shown) of work machine 100. Data link 150 may include any appropriate type of data communication medium, such as cable, wires, wireless radio, and/or laser, etc. Physical sensor 140 may include one or more sensors provided for measuring certain parameters of work machine operating environment. For example, physical sensor 140 may include emission sensors for measuring emissions of work machine 100, such as Nitrogen Oxides (NO_x), Sulfur Dioxide (SO₂), Carbon Monoxide (CO), total reduced Sulfur (TRS), etc. In particular, NO_x emission sensing and reduction may be important to normal operation of engine 110. Physical sensor 142, on the other hand, may include any appropriate sensors that are used inside engine 110 or other work machine components (not show) to provide various measured parameters about engine 110 or other components, such as temperature, speed, etc.

Virtual sensor system 130 may include any appropriate type of control system having one or more process models. The process models may be trained to generate values of sensing parameters based on a plurality of measured parameters. The sensing parameters may refer to those measurement parameters that are directly measured by a particular physical sensor. For example, a physical NO_x emission sensor may measure the NO_x emission level of work machine 100 and provide values of NO_x emission level, the sensing parameter, to other components, such as ECM 120. Sensing parameters, however, may also include any output parameters that may be measured indirectly by physical sensors and/or calculated based on readings of physical sensors.

On the other hand, the measured parameters may refer to any parameters relevant to the sensing parameters and indicative of the state of a component or components of work machine 100, such as engine 110. For example, for the sensing parameter NO_x emission level, measured parameters may include various parameters such as compression ratios, turbocharger efficiency, aftercooler characteristics, temperature values, pressure values, ambient conditions, fuel rates, and engine speeds, etc.

Further, virtual sensor system 130 may be configured as a separate control system or, alternatively, may coincide with other control systems such as ECM 120. FIG. 2 shows an exemplary functional block diagram of virtual sensor system 130.

As shown in FIG. 2, virtual sensor system 120 may include a processor 202, a memory module 204, a database 206, an I/O interface 208, a network interface 210, and a storage 212. Other components, however, may also be included.

Processor 202 may include any appropriate type of general purpose microprocessor, digital signal processor, or microcontroller. Processor 202 may be configured as a separate processor module dedicated to controlling engine 110. Alternatively, processor 202 may be configured as a shared processor module for performing other functions unrelated to virtual sensors.

Memory module 204 may include one or more memory devices including, but not limited to, a ROM, a flash memory, a dynamic RAM, and a static RAM. Memory module 204 may be configured to store information used by processor 202. Database 206 may include any type of appropriate database containing information on characteristics of measured parameters, sensing parameters, mathematical models, and/or any other control information.

Further, I/O interface 208 may also be connected to data link 150 to obtain data from various sensors or other components (e.g., physical sensors 140 and 142) and/or to transmit data to these components and to ECM 120. Network interface 210 may include any appropriate type of network device capable of communicating with other computer systems based on one or more communication protocols. Storage 212 may include any appropriate type of mass storage provided to store any type of information that processor 202 may need to operate. For example, storage 212 may include one or more hard disk devices, optical disk devices, or other storage devices to provide storage space.

As explained above, virtual sensor system 130 may include one or more process models with error correction capabilities to provide values of certain sensing parameters to ECM 120. FIG. 3 shows a logical block diagram of an exemplary virtual sensor system 130 with one process model.

As shown in FIG. 3, a virtual sensor process model 304 may be established to build interrelationships between input parameters 302 (e.g., measured parameters) and output parameters 306 (e.g., sensing parameters). After virtual sensor process model 304 is established, values of input parameters 302 may be provided to virtual sensor process model 304 to generate values of output parameters 306 based on the given values of input parameters 302 and the interrelationships between input parameters 302 and output parameters 306 established by the virtual sensor process model 304.

In certain embodiments, virtual sensor system 130 may provide levels of NO_x emitted from an exhaust system (not shown) of work machine 100. Input parameters 302 may include any appropriate type of data associated with NO_x emission levels. For example, input parameters 302 may include parameters that control operations of various response characteristics of engine 110 and/or parameters that are associated with conditions corresponding to the operations of engine 110. For example, input parameters 302 may include fuel injection timing, compression ratios, turbocharger efficiency, aftercooler characteristics, temperature values (e.g., intake manifold temperature), pressure values (e.g., intake manifold pressure), ambient conditions (e.g., ambient humidity), fuel rates, and engine speeds, etc. Other parameters, however, may also be included. Input parameters 302 may be measured by certain physical sensors, such as physical sensor 142, or created by other control systems such as ECM 120. Virtual sensor system 130 may obtain values of input parameters 302 via an input 310 coupled to data link 150.

On the other hand, output parameters 306 may correspond to sensing parameters. For example, output parameters 306 of a NO_x virtual sensor may include NO_x emission level, and/or any other types of output parameters used by NO_x virtual sensing application. Output parameters 306 (e.g., NO_x emission level) may be sent to ECM 120 via output 320 coupled to data link 150.

After virtual sensor process model 304 is established, values of input parameters 302 may be provided to virtual sensor process model 304 to generate values of output parameters 306 based on the given values of input parameters 302 and the interrelationships between input parameters 302 and output parameters 306 established by the virtual sensor process model 304. For example, virtual sensor system 130 may include a NO_x virtual sensor to provide levels of NO_x emitted from an exhaust system (not shown) of work machine 100.

Virtual sensor process model 304 may include any appropriate type of mathematical or physical model indicating interrelationships between input parameters 302 and output parameters 306. For example, virtual sensor process model 304 may be a neural network based mathematical model that is trained to capture interrelationships between input parameters 302 and output parameters 306. Other types of mathematic models, such as fuzzy logic models, linear system models, and/or non-linear system models, etc., may also be used. Virtual sensor process model 304 may be trained and validated using data records collected from a particular engine application for which virtual sensor process model 304 is established. That is, virtual sensor process model 304 may be established according to particular rules corresponding to a particular type of application using the data records, and the interrelationships of virtual sensor process model 304 may also be verified by using part of the data records.

After virtual sensor process model 304 is trained and validated, virtual sensor process model 304 may be optimized to define a desired input space of input parameters 302 and/or a desired distribution of output parameters 306. The validated or optimized virtual sensor process model 304 may be used to produce corresponding values of output parameters 306 when provided with a set of values of input parameters 102. In the above example, virtual sensor process model 304 may be used to produce NO_x emission level based on measured parameters, such as ambient humidity, intake manifold pressure, intake manifold temperature, fuel rate, and engine speed, etc.

The establishment and operations of virtual sensor process model 304 may be carried out by processor 202 based on computer programs stored on and/or loaded to virtual sensor system 130. Alternatively, the establishment of virtual sensor process model 304 may be realized by other computer systems, such as ECM 120 or a separate general purpose computer configured to create process models.

Processor 202 may perform a virtual sensor process model generation and optimization process to generate and optimize virtual sensor process model 304. FIG. 4 shows an exemplary model generation and optimization process performed by processor 202.

As shown in FIG. 4, at the beginning of the model generation and optimization process, processor 202 may obtain data records associated with input parameters 302 and output parameters 306 (step 402). The data records may include information characterizing engine operations and emission levels including NO_x emission levels. Physical sensor 140, such as physical NO_x emission sensors, may be provided to generate data records on output parameters 306 (e.g., sensing parameters such as NO_x levels). ECM 120 and/or physical sensor 142 may provide data records on input parameters 302 (e.g., measured parameters, such as intake manifold temperature, intake manifold pressure, ambient humidity, fuel rates, and engine speeds, etc.). Further, the data records may include both input parameters and output parameters and may be collected based on various engines or based on a single test engine, under various predetermined operational conditions.

The data records may also be collected from experiments designed for collecting such data. Alternatively, the data records may be generated artificially by other related processes, such as other emission modeling or analysis processes. The data records may also include training data used to build virtual sensor process model 304 and testing data used to validate virtual sensor process model 304. In addition, the data records may also include simulation data used to observe and optimize virtual sensor process model 304.

The data records may reflect characteristics of input parameters 302 and output parameters 306, such as statistic distributions, normal ranges, and/or precision tolerances, etc. Once the data records are obtained (step 402), processor 202 may pre-process the data records to clean up the data records for obvious errors and to eliminate redundancies (step 404). Processor 202 may remove approximately identical data records and/or remove data records that are out of a reasonable range in order to be meaningful for model generation and optimization. After the data records have been pre-processed, processor 202 may select proper input parameters by analyzing the data records (step 406).

The data records may be associated with many input variables, such as variables corresponding to fuel injection timing, compression ratios, turbocharger efficiency, aftercooler characteristics, various temperature parameters, various pressure parameters, various ambient conditions, fuel rates, and engine speeds, etc. The number of input variables may be greater than the number of a particular set of input parameters 302 used for virtual sensor process model 304, that is, input parameters 302 may be a subset of the input variables. For example, input parameter 302 may include intake manifold temperature, intake manifold pressure, ambient humidity, fuel rate, and engine speed, etc., of the input variables.

A large number of input variables may significantly increase computational time during generation and operations of the mathematical models. The number of the input variables may need to be reduced to create mathematical models within practical computational time limits. Additionally, in certain situations, the number of input variables in the data records may exceed the number of the data records and lead to sparse data scenarios. Some of the extra input variables may have to be omitted in certain mathematical models such that practical mathematical models may be created based on reduced variable number.

Processor 202 may select input parameters 302 from the input variables according to predetermined criteria. For example, processor 202 may choose input parameters 302 by experimentation and/or expert opinions. Alternatively, in certain embodiments, processor 202 may select input parameters based on a mahalanobis distance between a normal data set and an abnormal data set of the data records. The normal data set and abnormal data set may be defined by processor 202 using any appropriate method. For example, the normal data set may include characteristic data associated with input parameters 302 that produce desired output parameters. On the other hand, the abnormal data set may include any characteristic data that may be out of tolerance or may need to be avoided. The normal data set and abnormal data set may be predefined by processor 202.

Mahalanobis distance may refer to a mathematical representation that may be used to measure data profiles based on correlations between parameters in a data set. Mahalanobis distance differs from Euclidean distance in that mahalanobis distance takes into account the correlations of the data set. Mahalanobis distance of a data set X (e.g., a multivariate vector) may be represented asMD_i=(X_i−μ_x)Σ⁻¹(X_i−μ_x)′  (1)where μ_x is the mean of X and Σ⁻¹ is an inverse variance-covariance matrix of X. MD_i weights the distance of a data point X_i from its mean μ_x such that observations that are on the same multivariate normal density contour will have the same distance. Such observations may be used to identify and select correlated parameters from separate data groups having different variances.

Processor 202 may select input parameter 302 as a desired subset of input variables such that the mahalanobis distance between the normal data set and the abnormal data set is maximized or optimized. A genetic algorithm may be used by processor 202 to search input variables for the desired subset with the purpose of maximizing the mahalanobis distance. Processor 202 may select a candidate subset of the input variables based on a predetermined criteria and calculate a mahalanobis distance MD_normal of the normal data set and a mahalanobis distance MD_abnormal of the abnormal data set. Processor 202 may also calculate the mahalanobis distance between the normal data set and the abnormal data (i.e., the deviation of the mahalanobis distance MD_x=MD_normal−MD_abnormal). Other types of deviations, however, may also be used.

Processor 202 may select the candidate subset of input variables if the genetic algorithm converges (i.e., the genetic algorithm finds the maximized or optimized mahalanobis distance between the normal data set and the abnormal data set corresponding to the candidate subset). If the genetic algorithm does not converge, a different candidate subset of input variables may be created for further searching. This searching process may continue until the genetic algorithm converges and a desired subset of input variables (e.g., input parameters 302) is selected.

Optionally, mahalanobis distance may also be used to reduce the number of data records by choosing a part of data records that achieve a desired mahalanobis distance, as explained above.

After selecting input parameters 302 (e.g., intake manifold temperature, intake manifold pressure, ambient humidity, fuel rate, and engine speed, etc.), processor 202 may generate virtual sensor process model 304 to build interrelationships between input parameters 302 and output parameters 306 (step 408). In certain embodiments, virtual sensor process model 304 may correspond to a computational model, such as, for example, a computational model built on any appropriate type of neural network. The type of neural network computational model that may be used may include back propagation, feed forward models, cascaded neural networks, and/or hybrid neural networks, etc. Particular type or structures of the neural network used may depend on particular applications. Other types of computational models, such as linear system or non-linear system models, etc., may also be used.

The neural network computational model (i.e., virtual sensor process model 304) may be trained by using selected data records. For example, the neural network computational model may include a relationship between output parameters 306 (e.g., NO_x emission level, etc.) and input parameters 302 (e.g., intake manifold temperature, intake manifold pressure, ambient humidity, fuel rate, and engine speed, etc.). The neural network computational model may be evaluated by predetermined criteria to determine whether the training is completed. The criteria may include desired ranges of accuracy, time, and/or number of training iterations, etc.

After the neural network has been trained (i.e., the computational model has initially been established based on the predetermined criteria), processor 202 may statistically validate the computational model (step 410). Statistical validation may refer to an analyzing process to compare outputs of the neural network computational model with actual or expected outputs to determine the accuracy of the computational model. Part of the data records may be reserved for use in the validation process.

Alternatively, processor 202 may also generate simulation or validation data for use in the validation process. This may be performed either independently of a validation sample or in conjunction with the sample. Statistical distributions of inputs may be determined from the data records used for modeling. A statistical simulation, such as Latin Hypercube simulation, may be used to generate hypothetical input data records. These input data records are processed by the computational model, resulting in one or more distributions of output characteristics. The distributions of the output characteristics from the computational model may be compared to distributions of output characteristics observed in a population. Statistical quality tests may be performed on the output distributions of the computational model and the observed output distributions to ensure model integrity.

Once trained and validated, virtual sensor process model 304 may be used to predict values of output parameters 306 when provided with values of input parameters 302. Further, processor 202 may optimize virtual sensor process model 304 by determining desired distributions of input parameters 302 based on relationships between input parameters 302 and desired distributions of output parameters 306 (step 412).

Processor 202 may analyze the relationships between desired distributions of input parameters 302 and desired distributions of output parameters 306 based on particular applications. For example, processor 202 may select desired ranges for output parameters 306 (e.g., NO_x emission level that is desired or within certain predetermined range). Processor 202 may then run a simulation of the computational model to find a desired statistic distribution for an individual input parameter (e.g., one of intake manifold temperature, intake manifold pressure, ambient humidity, fuel rate, and engine speed, etc.). That is, processor 202 may separately determine a distribution (e.g., mean, standard variation, etc.) of the individual input parameter corresponding to the normal ranges of output parameters 306. After determining respective distributions for all individual input parameters, processor 202 may combine the desired distributions for all the individual input parameters to determine desired distributions and characteristics for overall input parameters 302.

Alternatively, processor 202 may identify desired distributions of input parameters 302 simultaneously to maximize the possibility of obtaining desired outcomes. In certain embodiments, processor 202 may simultaneously determine desired distributions of input parameters 302 based on zeta statistic. Zeta statistic may indicate a relationship between input parameters, their value ranges, and desired outcomes. Zeta statistic may be represented as

$\zeta =\sum _1^j\sum _1^iS_{\mathrm{ij}}\left(\frac{{\sigma }_i}{{\stackrel{\_}{x}}_i}\right)\left(\frac{{\stackrel{\_}{x}}_j}{{\sigma }_j}\right),$ where x_i represents the mean or expected value of an ith input; x_j represents the mean or expected value of a jth outcome; σ_i represents the standard deviation of the ith input; σ_j represents the standard deviation of the jth outcome; and is |S_ij| represents the partial derivative or sensitivity of the jth outcome to the ith input.

Under certain circumstances, x_i may be less than or equal to zero. A value of 3σ_i may be added to x_i to correct such problematic condition. If, however, x_i is still equal zero even after adding the value of 3σ_i, processor 202 may determine that σ_i may be also zero and that the process model under optimization may be undesired. In certain embodiments, processor 202 may set a minimum threshold for σ_i to ensure reliability of process models. Under certain other circumstances, σ_j may be equal to zero. Processor 202 may then determine that the model under optimization may be insufficient to reflect output parameters within a certain range of uncertainty. Processor 202 may assign an indefinite large number to ζ.

Processor 202 may identify a desired distribution of input parameters 302 such that the zeta statistic of the neural network computational model (i.e., virtual sensor process model 304) is maximized or optimized. An appropriate type of genetic algorithm may be used by processor 202 to search the desired distribution of input parameters 302 with the purpose of maximizing the zeta statistic. Processor 202 may select a candidate set values of input parameters 302 with predetermined search ranges and run a simulation of virtual sensor process model 304 to calculate the zeta statistic parameters based on input parameters 302, output parameters 306, and the neural network computational model. Processor 202 may obtain x_i and or σ_i by analyzing the candidate set values of input parameters 302, and obtain x_j and σ_j by analyzing the outcomes of the simulation. Further, processor 202 may obtain |S_ij| from the trained neural network as an indication of the impact of the ith input on the jth outcome.

Processor 202 may select the candidate set of input parameters 302 if the genetic algorithm converges (i.e., the genetic algorithm finds the maximized or optimized zeta statistic of virtual sensor process model 304 corresponding to the candidate set of input parameters 302). If the genetic algorithm does not converge, a different candidate set values of input parameters 302 may be created by the genetic algorithm for further searching. This searching process may continue until the genetic algorithm converges and a desired set of input parameters 302 is identified. Processor 202 may further determine desired distributions (e.g., mean and standard deviations) of input parameters 302 based on the desired input parameter set. Once the desired distributions are determined, processor 202 may define a valid input space that may include any input parameter within the desired distributions (step 414).

In one embodiment, statistical distributions of certain input parameters may be impossible or impractical to control. For example, an input parameter may be associated with a physical attribute of a device, such as a dimensional attribute of an engine part, or the input parameter may be associated with a constant variable within virtual sensor process model 304 itself. These input parameters may be used in the zeta statistic calculations to search or identify desired distributions for other input parameters corresponding to constant values and/or statistical distributions of these input parameters.

Further, optionally, more than one virtual sensor process model may be established. Multiple established virtual sensor process models may be simulated by using any appropriate type of simulation method, such as statistical simulation. Output parameters 306 based on simulation of these multiple virtual sensor process models may be compared to select a most-fit virtual sensor process model based on predetermined criteria, such as smallest variance with outputs from corresponding physical sensors, etc. The selected most-fit virtual sensor process model 304 may be deployed in virtual sensor applications.

As explained above, after virtual sensor process model 304 is trained, validated, and optimized, virtual sensor process model 304 may then be used by virtual sensor system 130 to predict output parameters 306. Further, ECM 120 and virtual sensor system 130 may provide control functions to relevant components of work machine 100. For example, ECM 120 may control engine 110 according to NO_x emission level provided by virtual sensor system 130, and, in particular, by virtual sensor process model 304.

However, under certain circumstances, such as unrepresentative data records, discrete data records, and/or complexity of the process model, deviations between output parameters 306 of virtual sensor process model 304 and output parameters of a physical sensor that is modeled by virtual sensor process model 304 may exist. Alternatively, the deviations may exist between values of output parameters 306 of virtual sensor process model 304 and corresponding desired output parameters predetermined by other software programs or users. These deviations may be referred as modeling errors and may need to be corrected for accuracy. Other modeling errors, however, may also be used. FIG. 5A shows an exemplary serial arrangement of virtual sensor system 130 for correcting modeling errors.

As shown in FIG. 5A, an error compensation process model 502 may be used to compensate certain modeling errors. Virtual sensor process model 304 and error compensation process model 502 may be arranged in series. That is, output 504 from virtual sensor process model 304 may be provided to error compensation process model 502, together with input parameters 302, as inputs. Outputs 506 from error compensation process model 502 and output 504 from virtual sensor process model 304 may be both provided to logic 508 to generate output parameters 306. Logic 508 may include any appropriate type of computer software program and/or hardware logic device. Logic 508 may use output 506 to compensate output 504 and to generate more accurate output parameters 306.

The arrangement of virtual sensor process model 304 and error compensation process model 502 may also be in parallel. FIG. 5B shows an exemplary parallel arrangement of virtual sensor system 130 to correct modeling errors. As shown in FIG. 5B, error compensation process model 502 may be provided with input parameters 302, without output 504 from virtual sensor process model 304 as input. However, same as the serial arrangement, outputs 506 from error compensation process model 502 and output 504 from virtual sensor process model 304 may be both provided to logic 508. Logic 508 may then use output 506 to compensate output 504 to generate more accurate output parameters 306.

The creations and operations of both the serial arrangement and the parallel arrangement of virtual sensor system 130 may be carried out by processor 202 via executing certain computer software programs. FIG. 6 shows an exemplary creation and operation process performed by processor 202.

As shown in FIG. 6, processor 202 may create virtual sensor process model 304, as explained above (step 602). After virtual sensor process model 304 is created, processor 202 may operate virtual process model 304 to generate data records (step 604). The data records may include data for input parameters 302, output parameters 306, and/or output 504, etc. Data records of output parameters 306 may include values from virtual sensor process model 304, values from an actual physical sensor that is modeled by virtual sensor process model 304, and/or predetermined desired values of output parameters 306. Modeling errors (e.g., deviations between the values of output parameters 306 or output 504 from virtual sensor process model 304 and actual and/or desired values of output parameters 306), may be derived from the data records and may also become a part of the data records. For example, deviations may be calculated by comparing output 504 and corresponding actual and/or desired output parameters 306. After sufficient data records are generated, the data records may be used for training, simulation, validation, and/or optimization purposes.

After the data records are available, processor 202 may create error compensation process model 502 using the data records (step 606). Error compensation process model 502 may be trained, validated, and/or optimized by using any appropriate method. For example, error compensation process model 502 may be trained by using the same process that may be used by virtual sensor process model 304. In the serial arrangement, data records of output 504 from error compensation process model 502 may be included in inputs to error compensation process model 502. On the other hand, in the parallel arrangement, data records of output 504 may be used to derive data records on modeling errors but may be unavailable for training error compensation process model 502. However, the amount of computation in the parallel arrangement may be significantly reduced and/or the speed of calculation in the parallel arrangement may be increased.

After error compensation process model 502 is created, processor 202 may operate both virtual sensor process model 304 and error compensation process model 502 (step 608). In one embodiment, both process models may be operated simultaneously. The operations may generate output 504 from virtual sensor process model 304 and output 506 from error compensation process model 502. Processor 202 may also provide output 504 and output 506 to logic 508. Processor 202 may also operate logic 508 to compensate output 504 with output 506 (i.e., model error compensation) to generate more accurate output parameters 306 (step 610).

Processor 202 may also present the compensated output parameters 306 to other control systems, such as ECM 120 (step 612). For example, processor 202 or virtual sensor system 130 may provide output parameters 306, such as NO_x emission level to ECM 120. ECM 120 may obtain output parameters 306 (e.g., NO_x emission level) via data link 150. After ECM 120 obtains the NO_x emission level from virtual sensor system 130, ECM 120 may control engine 110 and/or other components of work machine 100 based on the NO_x emission level (step 508). For example, ECM 120 may perform certain emission enhancing or minimization processes.

INDUSTRIAL APPLICABILITY

The disclosed methods and systems may provide efficient and accurate virtual sensor process models that can cover entire range of input and/or output spaces. Such technology may be used in a wide range of virtual sensors, such as sensors for engines, structures, environments, and materials, etc. In particular, the disclosed systems and methods provide practical solutions when process models are difficult to build using other techniques due to computational complexities and limitations of available data records.

The disclosed methods and systems may be used in combination with other process modeling techniques to significantly increase speed, accuracy, practicality, and/or flexibility. Other applications involving process modeling may also benefit from the disclosed methods and systems. Conventional techniques such as boosting and bagging may be replaced and/or improved by using the disclosed methods and systems.

The disclosed systems and methods may also be used by work machine manufacturers to reduce cost and increase reliability by replacing costly or failure-prone physical sensors. Reliability and flexibility may also be improved by the disclosed virtual sensor system. The disclosed virtual sensor techniques may be used to provide a wide range of parameters in components such as emission, engine, transmission, navigation, and/or control, etc. Further, parts of the disclosed system or steps of the disclosed method may also be used by computer system providers to facilitate or integrate other process models. For example, certain computer design software manufactures may integrate the disclosed methods and systems to improve the performance of various computer-based design software programs.

Other embodiments, features, aspects, and principles of the disclosed exemplary systems will be apparent to those skilled in the art and may be implemented in various environments and systems.

Claims

1. A computer-implemented method for a virtual sensor system, comprising: establishing a first process model indicative of interrelationships between a plurality of input parameters and a plurality of output parameters;establishing a second process model indicative of interrelationships between at least the plurality of input parameters and modeling errors of the first process model;operating the first process model to generate values of the plurality of output parameters based on the interrelationships between the plurality of input parameters and the plurality of output parameters;simultaneously operating the second model to generate estimated deviations between the values of the plurality of output parameters and desired values of the plurality of output parameters based on the interrelationships between at least the plurality of input parameters and the modeling errors of the first process model;compensating the values of the plurality of output parameters from the first process model with the estimated deviations from the second process model to generate the desired values of the plurality of output parameters; andproviding the desired values of the plurality of output parameters to a control system to control a machine.

2. The method according to claim 1, wherein the second process model is indicative of interrelationships between the modeling errors and a combination of the plurality of input parameters and the values of the plurality of output parameters from the first process model.

3. The method according to claim 2, wherein the simultaneously operating includes: obtaining the values of the plurality of output parameters from the first process model;providing the values of the plurality of output parameters and the plurality of input parameters to the second process model; andgenerating, by the second process model, the estimated deviations based on the interrelationships between the modeling errors and the combination of the plurality of input parameters and the values of the plurality of output parameters from the first process model.

4. The method according to claim 1, wherein establishing the first process model includes: obtaining data records associated with one or more input variables and the plurality of output parameters;selecting the plurality of input parameters from the one or more input variables;generating a computational model indicative of the interrelationships between the plurality of input parameters and the plurality of output parameters;determining desired statistical distributions of the plurality of input parameters of the computational model; andrecalibrating the plurality of input parameters based on the desired statistical distributions to define a desired input space.

5. The method according to claim 4, wherein selecting further includes: pre-processing the data records; andusing a genetic algorithm to select the plurality of input parameters from the one or more input variables based on a mahalanobis distance between a normal data set and an abnormal data set of the data records.

6. The method according to claim 4, wherein generating further includes: creating a neural network computational model;training the neural network computational model using the data records; andvalidating the neural network computation model using the data records.

7. The method according to claim 4, wherein determining further includes: determining a candidate set of the input parameters with a maximum zeta statistic using a genetic algorithm; anddetermining the desired distributions of the input parameters based on the candidate set,wherein the zeta statistic ζ is represented by:

8. The method according to claim 1, wherein the plurality of output parameters include at least a NOx emission level.

9. The method according to claim 1, wherein the plurality of input parameters include at least intake manifold temperature, intake manifold pressure, ambient humidity, fuel rates, and engine speeds.

10. A computer system, comprising: a database configured to store information relevant to a virtual sensor system; anda processor configured to: operate a first process model to generate values of a plurality of output parameters based on the interrelationships between a plurality of input parameters and the plurality of output parameters;simultaneously operate a second model to generate estimated deviations between the values of the plurality of output parameters and desired values of the plurality of output parameters based on the interrelationships between at least the plurality of input parameters and modeling errors of the first process model;compensate the values of the plurality of output parameters from the first process model with the estimated deviations from the second process model to generate the desired values of the plurality of output parameters; andprovide the desired values of the plurality of output parameters to a control system to control a machine.

11. The computer system according to claim 10, wherein the processor is configured to: establish the first process model indicative of interrelationships between a plurality of input parameters and the plurality of output parameters; andestablish the second process model indicative of interrelationships between at least the plurality of input parameters and modeling errors of the first process model.

12. The computer system according to claim 11, wherein the second process model is indicative of interrelationships between the modeling errors and a combination of the plurality of input parameters and the values of the plurality of output parameters from the first process model.

13. The computer system according to claim 12, wherein, to simultaneously operate the second model, the processor is configured to: obtain the values of the plurality of output parameters from the first process model;provide the values of the plurality of output parameters and the plurality of input parameters to the second process model; andgenerate the estimated deviations based on the interrelationships between the modeling errors and the combination of the plurality of input parameters and the values of the plurality of output parameters from the first process model.

14. The computer system according to claim 11, wherein, to establish the first process model, the processor is configured to: obtain data records associated with one or more input variables and the plurality of output parameters;select the plurality of input parameters from the one or more input variables;generate a computational model indicative of the interrelationships between the plurality of input parameters and the plurality of output parameters;determine desired statistical distributions of the plurality of input parameters of the computational model; andrecalibrate the plurality of input parameters based on the desired statistical distributions to define a desired input space.

15. The computer system according to claim 14, wherein, to select the plurality of input parameters, the processor is further configured to: pre-process the data records; anduse a genetic algorithm to select the plurality of input parameters from the one or more input variables based on a mahalanobis distance between a normal data set and an abnormal data set of the data records.

16. The computer system according to claim 14, wherein, to generate the computational model, the processor is further configured to: create a neural network computational model;train the neural network computational model using the data records; andvalidate the neural network computation model using the data records.

17. The computer system according to claim 14, wherein, to determine the desired statistical distributions, the processor is further configured to: determine a candidate set of the input parameters with a maximum zeta statistic using a genetic algorithm; anddetermine the desired distributions of the input parameters based on the candidate set,wherein the zeta statistic ζ is represented by:

18. A computer-readable medium for use on a computer system, the computer-readable medium having computer-executable instructions for performing a method comprising: establishing a first process model indicative of interrelationships between a plurality of input parameters and a plurality of output parameters;establishing a second process model indicative of interrelationships between at least the plurality of input parameters and modeling errors of the first process model;operating the first process model to generate values of the plurality of output parameters based on the interrelationships between the plurality of input parameters and the plurality of output parameters;simultaneously operating the second model to generate estimated deviations between the values of the plurality of output parameters and desired values of the plurality of output parameters based on the interrelationships between at least the plurality of input parameters and the modeling errors of the first process model;compensating the values of the plurality of output parameters from the first process model with the estimated deviations from the second process model to generate the desired values of the plurality of output parameters; andproviding the desired values of the plurality of output parameters to a control system to control a machine.

19. The computer-readable medium according to claim 18, wherein the second process model is indicative of interrelationships between the modeling errors and a combination of the plurality of input parameters and the values of the plurality of output parameters from the first process model.

20. The computer-readable medium according to claim 19, wherein the simultaneously operating includes: obtaining the values of the plurality of output parameters from the first process model;providing the values of the plurality of output parameters and the plurality of input parameters to the second process model; andgenerating, by the second process model, the estimated deviations based on the interrelationships between the modeling errors and the combination of the plurality of input parameters and the values of the plurality of output parameters from the first process model.

21. The computer-readable medium according to claim 18, wherein establishing the first process model includes: obtaining data records associated with one or more input variables and the plurality of output parameters;selecting the plurality of input parameters from the one or more input variables;generating a computational model indicative of the interrelationships between the plurality of input parameters and the plurality of output parameters;determining desired statistical distributions of the plurality of input parameters of the computational model; andrecalibrating the plurality of input parameters based on the desired statistical distributions to define a desired input space.