DETAILED ACTION Notice of Pre-AIA or AIA Status The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA. T his action is responsive to the Application filed on June 12, 2023 . Claims 1-19 are pending in the case. Claims 1, 8, 13, and 19 are the independent claims. This action is non-final. Claim Interpretation The following is a quotation of 35 U.S.C. 112(f): (f) Element in Claim for a Combination. – An element in a claim for a combination may be expressed as a means or step for performing a specified function without the recital of structure, material, or acts in support thereof, and such claim shall be construed to cover the corresponding structure, material, or acts described in the specification and equivalents thereof. The following is a quotation of pre-AIA 35 U.S.C. 112, sixth paragraph: An element in a claim for a combination may be expressed as a means or step for performing a specified function without the recital of structure, material, or acts in support thereof, and such claim shall be construed to cover the corresponding structure, material, or acts described in the specification and equivalents thereof. The claims in this application are given their broadest reasonable interpretation using the plain meaning of the claim language in light of the specification as it would be understood by one of ordinary skill in the art. The broadest reasonable interpretation of a claim element (also commonly referred to as a claim limitation) is limited by the description in the specification when 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph, is invoked. As explained in MPEP § 2181, subsection I, claim limitations that meet the following three-prong test will be interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph: (A) the claim limitation uses the term “means” or “step” or a term used as a substitute for “means” that is a generic placeholder (also called a nonce term or a non-structural term having no specific structural meaning) for performing the claimed function; (B) the term “means” or “step” or the generic placeholder is modified by functional language, typically, but not always linked by the transition word “for” (e.g., “means for”) or another linking word or phrase, such as “configured to” or “so that”; and (C) the term “means” or “step” or the generic placeholder is not modified by sufficient structure, material, or acts for performing the claimed function. Use of the word “means” (or “step”) in a claim with functional language creates a rebuttable presumption that the claim limitation is to be treated in accordance with 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph. The presumption that the claim limitation is interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph, is rebutted when the claim limitation recites sufficient structure, material, or acts to entirely perform the recited function. Absence of the word “means” (or “step”) in a claim creates a rebuttable presumption that the claim limitation is not to be treated in accordance with 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph. The presumption that the claim limitation is not interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph, is rebutted when the claim limitation recites function without reciting sufficient structure, material or acts to entirely perform the recited function. Claim limitations in this application that use the word “means” (or “step”) are being interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph, except as otherwise indicated in an Office action. Conversely, claim limitations in this application that do not use the word “means” (or “step”) are not being interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph, except as otherwise indicated in an Office action. This application includes one or more claim limitations that do not use the word “means,” but are nonetheless being interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph, because the claim limitation(s) uses a generic placeholder that is coupled with functional language without reciting sufficient structure to perform the recited function and the generic placeholder is not preceded by a structural modifier. Such claim limitation(s) is/are: dynamic model establishment module, offline deep learning prediction model establishment module, online deep learning prediction model establishment module, deep learning correction model establishment module, and self-correction module in claims 7 and 8; and end subdevice, edge subdevice, and cloud subdevice in claim 13. Because this/these claim limitation(s) is/are being interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph, it/they is/are being interpreted to cover the corresponding structure described in the specification as performing the claimed function, and equivalents thereof. If applicant does not intend to have this/these limitation(s) interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph, applicant may: (1) amend the claim limitation(s) to avoid it/them being interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph (e.g., by reciting sufficient structure to perform the claimed function); or (2) present a sufficient showing that the claim limitation(s) recite(s) sufficient structure to perform the claimed function so as to avoid it/them being interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph. Claim Rejections - 35 USC § 112 The following is a quotation of 35 U.S.C. 112(b): (b ) CONCLUSION.—The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the inventor or a joint inventor regards as the invention. The following is a quotation of 35 U.S.C. 112 (pre-AIA), second paragraph: The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the appl icant regards as his invention. Claims 2, 8, and 14 are rejected under 35 U.S.C. 112(b) or 35 U.S.C. 112 (pre-AIA), second paragraph, as being indefinite for failing to particularly point out and distinctly claim the subject matter which the inventor or a joint inventor (or for applications subject to pre-AIA 35 U.S.C. 112, the applicant), regards as the invention. Claims 2, 8, and 14 each recite, on lines 18-19, 17, and 19, respectively, “the online training algorithm.” This limitation lacks antecedent basis. Claim Rejections – 35 USC § 101 35 U.S.C. 101 reads as follows: Whoever invents or discovers any new and useful process, machine, manufacture, or composition of matter, or any new and useful improvement thereof, may obtain a patent therefor, subject to the conditions and requirements of this title. Claim 19 is rejected under 35 U.S.C. 101 because the claimed invention is directed to non-statutory subject matter. The claim does not fall within at least one of the four categories of patent eligible subject matter because the claims are directed to transitory signals ; i.e., the claim recites a computer-readable storage medium, but does not specify that it is non-transitory. Under the broadest reasonable interpretation, the claimed computer-readable storage medium encompasses transitory forms of signal transmission (see MPEP 2106.03.II) and therefore is not a machine, an article of manufacture, a process, or a composition of matter as contemplated ty 35 U.S.C. 101. Examiner respectfully suggests amending the claims to clarify that the claimed invention does not encompass transitory signals, e.g. by reciting a non-transitory computer readable storage medium . Claim Rejections – 35 USC § 103 The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action: A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made. The factual inquiries set forth in Graham v. John Deere Co. , 383 U.S. 1, 148 USPQ 459 (1966), that are applied for establishing a background for determining obviousness under 35 U.S.C. 103 are summarized as follows: 1. Determining the scope and contents of the prior art. 2. Ascertaining the differences between the prior art and the claims at issue. 3. Resolving the level of ordinary skill in the pertinent art. 4. Considering objective evidence present in the application indicating obviousness or nonobviousness. This application currently names joint inventors. In considering patentability of the claims under pre-AIA 35 U.S.C. 103(a), the examiner presumes that the subject matter of the various claims was commonly owned at the time any inventions covered therein were made absent any evidence to the contrary. Applicant is advised of the obligation under 37 CFR 1.56 to point out the inventor and invention dates of each claim that was not commonly owned at the time a later invention was made in order for the examiner to consider the applicability of pre-AIA 35 U.S.C. 103(c) and potential pre-AIA 35 U.S.C. 102€, (f) or (g) prior art under pre-AIA 35 U.S.C. 103(a). Claim s 1, 7, 13, and 19 are rejected under 35 U.S.C. 103 as being unpatentable over Modi et al (US 20190179271 A1) in view of Jiang et al. (US 20220159639 A1) . With respect to claims 1 and 7, Modi teaches an adaptive deep learning-based intelligent prediction apparatus for a complex industrial system, wherein the apparatus comprises a dynamic model establishment module, an offline deep learning prediction model establishment module, an online deep learning prediction model establishment module, a deep learning correction model establishment module, and a self-correction module configured to perform a method ( e.g. paragraphs 0093-0094, computer including processors and memory/storage including software instructions and data used to implement described invention, where the memory may include off line model, process model flowsheet, first principles model, rule engine, other models and data, etc. ) ; and the adaptive deep learning-based intelligent prediction method for a complex industrial system, wherein the method comprises: establishing a dynamic model for a complex industrial system ( e.g. paragraph 0032, plant-wide process model/source model ; i.e. an overall model for the system which is not deployed online; paragraph 0044, Fig. 5 step 520, selecting process/source model for industrial, chemical, or other such plant; selected source model is a first-principles model encompassing wide plant process scope developed and configured for offline use, such as for design or rating of the plant operations ) ; establishing an offline deep learning prediction model by using the dynamic model ( e.g. paragraph 0034, converting source model into unit process (operation-centric) model of selected modeled operating unit ; i.e. a version of the process model prior to calibration which is not deployed online; paragraph 0046-0047, Fig. 5 step 540, selecting modeled operating unit of interest and generating offline version of unit process model for the selected modeled operating unit by converting the source model into a model centric to only the selected modeled operating unit; paragraph 0051, Fig. 5 step 580, saving the generated unit process model; referred to as the offline version ) ; establishing an online deep learning prediction model by using the offline deep learning prediction model ( e.g. paragraphs 0035-0038, calibrating unit process model, reconciling modeled flow such that the unit process model can perform at steady-state using measurements collected by physical instruments of the physical operating unit, and estimating compositions of feed streams entering physical operating unit to update the calibration dataset and again reconcile the unit process model; once unit process model is calibrated, it is deployed online by storing the calibrated process model and dynamically executing it online with real-time plant data; paragraph 0052, Fig. 6A, calibrating the unit process model to function online to drive real-time decision making at the plant ) ; establishing a deep learning correction model ( e.g. paragraph 0082, rule engine generated for predictive insight and prescriptive guidance, referring to model(s) utilized to predict events including discrepancies between plant measurements taken by physical instruments and first principles model predictions and imminent undesirable operating events ) ; and correcting a model by using the deep learning correction model ( e.g. paragraph 0082, based on a predicted event, rule engine issuing alert of the event as well as suggestions for corrective actions, such as informing the user that the first principles model is no longer properly calibrated for the current asset operation or that drifts in physical instruments may be occurring indicating the need for re-calibration of the physical instruments ) ; wherein the online deep learning prediction model is configured to predict a parameter of the complex industrial system in real time ( e.g. paragraph 0038, online unit process model associated/linked to real-time plant data and solves the online unit process model to compute values for KPIs of interest; performing real-time monitoring and predictive analytics of operations of the modeled operating unit; based on the monitoring and predictive analytics, instrumentation, control and operation computer providing monitoring and predictive analytics to personnel to take actions to optimize or otherwise control operations of the model operating unit or other operating units of the plant ) . Modi does not explicitly disclose that the deep learning correction model is based on a structure that is the same as a structure of the online deep learning prediction model and correcting the online deep learning prediction model by using the deep learning correction model. However, Jiang teaches that the deep learning correction model is based on a structure that is the same as a structure of the online deep learning prediction model and correcting the online deep learning prediction model by using the deep learning correction model ( e.g. paragraphs 0143-0144, Fig. 7, long-term predictor comprises training/evaluation model; historical data used by training/evaluation model to generate updated weights/parameters which are communicated to second RNN; generation RNN of predictive model 704 and evaluation RNN of training/evaluation model 701 have the same architecture; paragraph 0154, in the case of using data collected through use of real time control system, the training/edge model 701 is deployed in the long-term predictor for fine-tuning the previously generated model parameters on the fly; the evaluation RNN has the same architecture as the generation RNN; paragraph 0162, in embodiment where model originally trained offline, training that takes place in training/evaluation model 701 is for fine-tuning; i.e. as shown in Fig. 7, a training/evaluation model/RNN is deployed (analogous to a correction model), where this model has the same architecture as the prediction model ) . In addition, assuming arguendo that Modi does not teach that the models are deep learning models, Jian teaches that the models are deep learning models ( e.g. models as cited above, including those trained offline and then used in a real-time/online environment, having a deep learning architecture such as an RNN architecture; paragraph 0147, indicating that the RNN comprises a plurality of layers of RNN cells where the last layer is connected to an output layer, etc.). Accordingly, it would have been obvious to one of ordinary skill in the art before the effective filing date of the invention having the teachings of Modi and Jiang in front of him to have modified the teachings of Modi (directed to dynamic construction and online deployment of operation-centric first-principles model for predictive analytics ), to incorporate the teachings of Jiang (directed to edge-intelligence for stability guaranteed real-time control systems ) to include the capability to base the correction model and the online prediction model on the same architecture . One of ordinary skill would have been motivated to perform such a modification in order to provide a real-time control system which can be used to enable flexible operation of equipment and facilitate intelligent automation in a number of different fields while mitigating various deficiencies existing systems as described in Jiang ( paragraph 000 1 -000 8 ) . With respect to claim 19, Modi in view of Jiang teaches a computer-readable storage medium storing a computer program, wherein the computer program is executed by a processor to implement the method according to claim 1 ( e.g. Modi paragraphs 0093-0094, computer including processors and memory/storage including software instructions and data used to implement described invention, where the memory may include off line model, process model flowsheet, first principles model, rule engine, other models and data, etc. ) . With respect to claim 13, Modi in view of Jiang teaches a n adaptive deep learning-based intelligent prediction device for a complex industrial system to implement the method according to claim 1, wherein the device comprises an end subdevice, an edge subdevice, and a cloud subdevice ( e.g. paragraphs 0031-0032, Fig. 1, system 100 configured to build, calibrate, and deploy online unit process model, including one or more computers/servers to execute model convertor, model calibrator, and deployment engine as well as centralized data store and data server and distributed control system communicatively coupled to instrument devices that physically measure and control operating unit at plant; paragraphs 0091-0093, Fig. 9 illustrating processing environment in which invention is implemented, including client devices 50 and server computers 60, where the internal structure of these computers includes software instructions 92 and data 94, including for implementing present invention such as offline model, process model flowsheet, data historian, first-principles model, hydraulic model, rule engine, etc. when executed by CPU 84; Examiner notes that each respective subdevice appears to be recited as a component of an overall device with no additional structure and, therefore, may include any combination of hardware and/or software ) ; the end subdevice is configured to collect input data and output data of the complex industrial system ( e.g. paragraph 0035, Fig. 1, dataset creator retrieving plant data from real-time plant historian (located at centralized data store 130 or another location on the plant network; paragraph 0038, real-time plant data collected in real-time by instrumentation, control and operation computer 175 from physical instruments of the plant network 120 and written to the real-time plant historian for retrieval by other plant computers ) ; the edge subdevice is configured to predict a parameter of the complex industrial system in real time by using an online deep learning prediction model ( e.g. paragraph 0038, deployment engine 160 associates online unit process model to link with the retrieved real-time plant data and solves the online unit process model to compute values for KPIs of interest, which can be used for real-time monitoring and predictive analytics of the operations of the modeled operating unit ) ; and the cloud subdevice is configured to implement and tune a deep learning correction model and correct the online deep learning prediction model by using the deep learning correction model ( e.g. paragraph 0082, generating rule engine which is a combination of models in conjunction with domain specific logic; based on a predicted event, rule engine issuing alert of the event as well as suggestions for corrective actions, such as informing the user that the first principles model is no longer properly calibrated for the current asset operation or that drifts in physical instruments may be occurring indicating the need for re-calibration of the physical instruments paragraph 0084, rule engine composed of models which generate alert scores consumed by rule engine; alert models receive model inputs including first principles model predictions, plant tag data, etc., to generate alert scores; paragraph 0088, tuning parameters of rule engine ) . Jiang further teaches the cloud subdevice is configured to train a deep learning correction model and correct the online deep learning prediction model by using the deep learning correction model ( e.g. paragraphs 0143-0144, Fig. 7, long-term predictor comprises training/evaluation model; historical data used by training/evaluation model to generate updated weights/parameters which are communicated to second RNN; paragraph 0154, in the case of using data collected through use of real time control system, the training/edge model 701 is deployed in the long-term predictor for fine-tuning the previously generated model parameters on the fly; paragraph 0158-0159, calculating average loss between predicted command signals and label matrix, using in gradient descent algorithm when modifying model parameters; calculating mean square error between output results and corresponding actual command in label matrix; updating the weights of the RNN in the direction of the negative gradient of the loss; paragraph 0162, in embodiment where model originally trained offline, training that takes place in training/evaluation model 701 is for fine-tuning; i.e. as shown in Fig. 7, a training/evaluation model/RNN trained and deployed (analogous to a correction model), where this model is used to correct the prediction model in a real-time operating environment ). In addition, assuming arguendo that Modi does not teach that the models are deep learning models, Jian teaches that the models are deep learning models ( e.g. models as cited above, including those trained offline and then used in a real-time/online environment, having a deep learning architecture such as an RNN architecture; paragraph 0147, indicating that the RNN comprises a plurality of layers of RNN cells where the last layer is connected to an output layer, etc.). Accordingly, it would have been obvious to one of ordinary skill in the art before the effective filing date of the invention having the teachings of Modi and Jiang in front of him to have modified the teachings of Modi (directed to dynamic construction and online deployment of operation-centric first-principles model for predictive analytics), to incorporate the teachings of Jiang (directed to edge-intelligence for stability guaranteed real-time control systems ) to include the capability to train the correction model and use it to correct the online prediction model , where each of the offline model, online model, and correction model may be implemented using deep learning architectures . One of ordinary skill would have been motivated to perform such a modification in order to provide a real-time control system which can be used to enable flexible operation of equipment and facilitate intelligent automation in a number of different fields while mitigating various deficiencies existing systems as described in Jiang (paragraph 000 1 -000 8 ). Claim s 2, 3, 8, 9, 14 , and 15 are rejected under 35 U.S.C. 103 as being unpatentable over Modi in view of Jiang , further in view of Liu et al. (US 20170091615 A1) , further in view of Hatakeyama et al. (US 20240005217 A1) . With respect to claims 2 , 8 , and 14 , Modi in view of Jiang teaches all of the limitations of claims 1 , 7 , and 14 as previously discussed, and Jiang further teaches wherein a dynamic model is established for the complex industrial system, which comprises/ the establishing a dynamic model for a complex industrial system comprises: the establishing (by the offline deep learning prediction model establishment module) an offline deep learning prediction model by using the dynamic model comprises: establishing the offline deep learning prediction model by using a long-short term memory (LSTM) network ( e.g. paragraph 0142, long term predictor comprises an RNN based on LSTM architecture; paragraph 0154, RNN of long - term predictor trained offline ; paragraph 0162, model is generated offline ; Fig. 7, showing long-term predictor 515 including a predictive model 704 having a generation RNN; i.e. an initial version of the predic tion/predictive model is generated offline using an LSTM architecture ); the establishing (by the online deep learning prediction model establishment module) an online deep learning prediction model by using the offline deep learning prediction model comprises: establishing the online deep learning prediction model by using an LSTM network, wherein an input of a single neuron, a neuron quantity, a cell node quantity, and a network layer quantity of the online deep learning prediction model are all the same as those of the offline deep learning prediction model ( e.g. paragraph 0120, real-time control system with predictive model for predicting command messages; paragraph 0126, Fig. 5 showing system/technique for use in real-time control system, including predictive models; paragraph 0129, predictor includes long-term predictor; paragraph 0142, long term predictor comprises an RNN based on LSTM architecture; paragraph 0144, Fig. 7, long-term predictor comprises predictive model 704 comprising a generation RNN; paragraph 0154, optionally training data collected through use of real time control system and performing fine-tuning on the fly; paragraph 0162, performing fine-tuning; i.e. the offline-generated prediction model using the LSTM architecture (long term predictor) is then deployed in a real-time setting (i.e. analogous to an online setting), such that the deployed version of the prediction model is an online version of the offline-trained prediction model and therefore has the same architecture as the offline-trained prediction model, i.e. a same LSTM architecture, and therefore a same input of a single neuron, neuron quantity, cell node quantity, and network layer quantity are used in both the offline-trained version of the model and the online (i.e. real-time) version of the model ); using weight and bias parameters of each layer of the offline deep learning prediction model as initial values of weight and bias parameters of the corresponding layer of the online deep learning prediction model ( e.g. as cited above, the offline-trained model is deployed in an online/real-time environment; since is the same/a copy of the model trained offline, the online/real-time version initially includes the same weight and bias parameters of each layer as the offline version ); and correcting weight and bias parameters of a last layer of the online deep learning prediction model online by using the online training algorithm based on an error between the labels and an output of the online deep learning prediction model ( e.g. paragraphs 0071-0072, determining average loss based on difference between predicted information signal and label matrix; updating predictive machine learning model based on average loss; paragraph 0160, predictive model 704 measures the prediction accuracy of the generation RNN by calculating error between predicted commands and labels ); the establishing (by the deep learning correction model establishment module) a deep learning correction model based on a structure that is the same as a structure of the online deep learning prediction model comprises: establishing the deep learning correction model by using an LSTM network, wherein an input of a single neuron, a neuron quantity, a cell node quantity, and a network layer quantity of the deep learning correction model are all the same as those of the online deep learning prediction model ( e.g. paragraph 0142, long term predictor comprises an RNN based on LSTM architecture; paragraphs 0143-0144, Fig. 7, long-term predictor comprises training/evaluation model 701 comprising an evaluation RNN; historical data used by training/evaluation model to generate updated weights/parameters which are communicated to second RNN; generation RNN of predictive model 704 and evaluation RNN have the same architecture; paragraph 0154, in the case of using data collected through use of real time control system, the training/edge model 701 is deployed in the long-term predictor for fine-tuning the previously generated model parameters on the fly; the evaluation RNN has the same architecture as the generational RNN ; paragraph 0162, in embodiment where model originally trained offline, training that takes place in training/evaluation model 701 is for fine-tuning ; i.e. as shown in Fig. 7, a training/evaluation model/RNN is also deployed (analogous to a correction model), where this model has the same LSTM architecture as the prediction model, and therefore a same input of a single neuron, neuron quantity, cell node quantity, and network layer quantity ) ; correcting weight and bias parameters of each layer of the deep learning correction model in real time by using a training algorithm based on an error between the labels and an output of the deep learning correction model ( e.g. paragraph 0126, indicating that the system/technique is for use in a real-time control system; paragraph 0127-0129, commands transmitted in timeslots, including to the predictor 511 which includes the long-term predictor 515; paragraph 0131, providing both predicted command and actual command for each timeslot; paragraph 0132, predictor 515 providing new prediction every time a new packet is received; paragraph 0158-0159, calculating average loss between predicted command signals and label matrix, using in gradient descent algorithm when modifying model parameters; calculating mean square error between output results and corresponding actual command in label matrix; updating the weights of the RNN in the direction of the negative gradient of the loss; paragraph 0160, training/evaluation model 701 measures the prediction accuracy of the evaluation RNN by calculating error between predicted commands and labels ); and the correcting (self-correction module) the online deep learning prediction model by using the deep learning correction model comprises: when a preset condition is met, replacing weight and bias parameters of each layer of the online deep learning prediction model with weight and bias parameters of the corresponding layer of the deep learning correction model ( e.g. paragraph 0160, the performance of the training/evaluation model 701 is compared to the performance of the predictive model 704; once the performance of the training/evaluation model outperforms (has a better performance metric) the predictive model, the parameters of the predictive model are updated; gradient of the evaluation RNN 702 is communicated to the predictive model 704 to generate updated weights/parameters 707 for use in making future predictions; paragraph 0161, weights of the generation RNN 705 are updated based on the received gradient ); wherein historical data input into the deep learning correction model is more than historical data input into the online deep learning prediction model ( e.g. Fig. 7, showing the training/evaluation model 701 receiving historical signals, where the predictive model 704 does not appear to receive historical signals; paragraph 0143, historical data 703 used by training/evaluation model to generate updated weights/parameters ). Accordingly, it would have been obvious to one of ordinary skill in the art before the effective filing date of the invention having the teachings of Modi and Jiang in front of him to have modified the teachings of Modi (directed to dynamic construction and online deployment of operation-centric first-principles model for predictive analytics), to incorporate the teachings of Jiang (directed to edge-intelligence for stability guaranteed real-time control systems ) to include the capability to , after training the offline model, use it in a real-time/online environment, such that the offline model and online model have the same structure, weights, parameters, and initial values, and to further implement the correction model with a same structure/architecture, where each of the models is implemented with an LSTM architecture, and to further train the models, including by calculating error/loss between labels and outputs of the models and updating weights and parameters based on the error/loss, and to further utilize the correction model to replace the weight and parameters of the online model based on a preset condition such as respective accuracy of the models, where the correction model receives as input more historical data than the online model . One of ordinary skill would have been motivated to perform such a modification in order to provide a real-time control system which can be used to enable flexible operation of equipment and facilitate intelligent automation in a number of different fields while mitigating various deficiencies existing systems as described in Jiang (paragraph 000 1 -000 8 ). Modi and Jiang do not explicitly disclose determining a neuron quantity, a cell node quantity, and a network layer of the LSTM network, and weight and bias parameters of each layer by using an offline training algorithm based on an error. However, Liu teaches determining a neuron quantity, a cell node quantity, and a network layer quantity of the LSTM network, and weight and bias parameters of each layer by using an offline training algorithm based on an error ( e.g. paragraph 0010, selecting artificial neural network model including multiple layers, initializing the ANN by defining a number of nodes included in each layer, activation function for use in each neuron cell node in each layer, a number of bias nodes to be included in each layer, and training the network to develop optimal set of weights and bias node values; paragraph 0042, number of hidden layers in network different for different datasets ; paragraph 0050, training is the process of determining proper weights and bias values applied to the various inputs at activation nodes in the network; paragraph 0051, training process continues until set of weights and bias node values is found that minimizes cost; paragraph 0053, selecting number of hidden layers to be included in model, number of nodes to be included in each layer, and sets of weights and bias values; applying iterative process of determining proper weights and bias values; paragraph 0055, once model is initialize, training process continues by computing gradients associated with determined weights and bias values, such as by using backpropagation to determine error attributed to each layer (for each individual node in the layer); paragraph 0056, performing optimization on all generated gradients, selecting optimum set of weights and bias values that is an acceptable set of parameters for the neural network model and best fits the data (time series); paragraph 0068, training algorithm used to find weights that minimize overall error measure; paragraph 0069, error term measuring contribution of each node to errors in generated output values, to find optimized combination of weights and bias values for the neural network; i.e. the structural details of the network can be determined, including neuron, cell node, and layer quantities, and weight and bias parameters for each layer, including by using a training algorithm based on an error ). Accordingly, it would have been obvious to one of ordinary skill in the art before the effective filing date of the invention having the teachings of Modi , Jiang, and Liu in front of him to have modified the teachings of Modi (directed to dynamic construction and online deployment of operation-centric first-principles model for predictive analytics) and Jiang (directed to edge-intelligence for stability guaranteed real-time control systems ), to incorporate the teachings of Liu (directed to predicting power plant operational parameters utilizing artificial neural network deep learning ) to include the capability to determine the structural details of the model (such as the offline model of Modi, which may be based on LSTM architecture as taught by Jiang), including neuron, cell node, and layer quantities, and weight and bias parameters for each layer, including by using a training algorithm based on an error . One of ordinary skill would have been motivated to perform such a modification in order to accurately predict future values of time series data associated with a plant, such that plant personnel are able to schedule future events in a cost-efficient, timely manner, using a neural network which is able to follow trends in the time series data without overfitting as described in Liu ( abstract ). Modi and Jiang do not explicitly disclose determining (by the dynamic model establishment module) an input variable and an output variable of the dynamic model, wherein the output variable is a predicted variable; using the input variable of the dynamic model as an input of the LSTM network, using output data of the dynamic model as labels, the error is between the labels and an output of the offline deep learning prediction model; However, Hakateyama teaches determining (by the dynamic model establishment module) an input variable and an output variable of the dynamic model, wherein the output variable is a predicted variable ( e.g. paragraph 0050, unlabeled training examples used to train trained teacher model; paragraph 0051, generating labels for input unlabeled training examples using trained teacher model, and outputs labels which correspond to respective predictions of the teacher model for the input unlabeled training examples; i.e. determining inputs and outputs of the model, such as a teacher model, where the output is a prediction based on the input ); using the input variable of the dynamic model as an input of the LSTM network ( e.g. paragraph 0050, unlabeled training examples used to train the student models; i.e. the same inputs are provided to a second/student model in addition to the teacher model ), using output data of the dynamic model as labels ( e.g. paragraph 0051, generating labels for input unlabeled training examples using trained teacher model, and outputs labels which correspond to respective predictions of the teacher model for the input unlabeled training examples; i.e. the outputs of the teacher model are used as labels ), the error is between the labels and an output of the offline deep learning prediction model ( e.g. paragraph 0053, performing prediction using student model and calculating error between prediction by the student model and prediction by the teacher model, and outputting the error ); Accordingly, it would have been obvious to one of ordinary skill in the art before the effective filing date of the invention having the teachings of Modi , Jiang, Liu, and Hakateyama in front of him to have modified the teachings of Modi (directed to dynamic construction and online deployment of operation-centric first-principles model for predictive analytics) , Jiang (directed to edge-intelligence for stability guaranteed real-time control systems ) , and Liu (directed to predicting power plant operational parameters utilizing artificial neural network deep learning ), to incorporate the teachings of Hakateyama (directed to improving training accuracy of machine learning models ) to include the capability to implement the dynamic model as a teacher model and the offline model as a student model, where input and output data/predictions of the teacher model are determined, and the output data/predictions of the teacher model are used as labels for training the student/offline model, where the student/offline model is trained using the same input data as the teacher model, and the error is determined based on a difference between the labels output from the teacher/dynamic model and the outputs of the student/online model . One of ordinary skill would have been motivated to perform such a modification in order to significantly improve the accuracy of a model as described in Hakateyama ( paragraph 0007 ). With respect to claim s 3, 9, and 15, Modi in view of Jiang, further in view of Liu, further in view Hakateyama of teaches all of the limitations of claim s 2, 8, and 14 as previously discussed, and Liu further teaches wherein the correcting weight and bias parameters of a last layer of the online deep learning prediction model online is specifically correcting some weight parameters and some bias parameters of the last layer of the online deep learning prediction model online ( e.g. paragraph 0010, training the network to develop optimal set of weights and bias node values; paragraph 0050, training is the process of determining proper weights and bias values applied to the various inputs at activation nodes in the network; paragraph 0051, training process continues until set of weights and bias node values is found that minimizes cost; paragraph 0053, selecting sets of weights and bias values; applying iterative process of determining proper weights and bias values; paragraph 0055, once model is initialize, training process continues by computing gradients associated with determined weights and bias values , such as by using backpropagation to determine error attributed to each layer (for each individual node in the layer) ; paragraph 0056, performing optimization on all generated gradients, selecting optimum set of weights and bias values that is an acceptable set of parameters for the neural network model and best fits the data (time series); paragraph 0068, training algorithm used to find weights that minimize overall error measure; paragraph 0069, error term measuring contribution of each node to errors in generated output values, to find optimized combination of weights and bias values for the neural network; i.e. the optimal weight and bias values are found for each node in each layer (i.e. including the last layer) such that at least some weight parameters and at least some bias parameters are corrected for the last layer of the network ) . As previously cited, Jiang teaches that the correction of the parameters for the online deep learning model can be performed online. Accordingly, it would have been obvious to one of ordinary skill in the art before the effective filing date of the invention having the teachings of Modi , Jiang, Hakateyama, and Liu in front of him to have modified the teachings of Modi (directed to dynamic construction and online deployment of operation-centric first-principles model for predictive analytics) , Hakateyama (directed to improving training accuracy of machine learning models ) , and Jiang (directed to edge-intelligence for stability guaranteed real-time control systems ), to incorporate the teachings of Liu (directed to predicting power plant operational parameters utilizing artificial neural network deep learning ) to include the capability to determine the structural details of the model (such as the offline model of Modi, which may be based on LSTM architecture as taught by Jiang), including neuron, cell node, and layer quantities, and weight and bias parameters for each layer, including by using a training algorithm based on an error . One of ordinary skill would have been motivated to perform such a modification in order to accurately predict future values of time series data associated with a plant, such that plant personnel are able to schedule future events in a cost-efficient, timely manner, using a neural network which is able to follow trends in the time series data without overfitting as described in Liu ( abstract ). Claim s 4, 10, and 16 are rejected under 35 U.S.C. 103 as being unpatentable over Modi in view of Jiang, further in view of Wang, W., Wang, D. Prediction of component concentrations in sodium aluminate liquor using stochastic configuration networks. Neural Comput & Applic 32, 13625–13638 (2020). https://doi.org/10.1007/s00521-020-04771-4 . (hereinafter Wang). With respect to claims 4, 10, and 16, Modi in view of Jiang teaches all of the limitations of claims 1, 7, and 13 as previously discussed , including Jiang’s teachings that the online model is an online deep learning prediction model (as previously cited). Modi and Jiang do not explicitly disclose wherein the complex industrial system is an alumina production system, and the model is configured to predict a caustic concentration detection error of the alumina production system in real time; and the caustic concentration detection error is a difference between a laboratory value of a caustic concentration and a caustic concentration measured by an online caustic concentration detection instrument. However, Wang teaches wherein the complex industrial system is an alumina production system ( e.g. page 13625, abstract, alumina production process ) , and the online deep learning prediction model is configured to predict a caustic concentration detection error of the alumina production system in real time ( e.g. page 13625, abstract, caustic hydroxide, alumina, and sodium carbonate, concentrations of which are respectively represented by C K , C A , and C C ; page 13626, second column, second paragraph, indicating that real-time detection of the concentrations is required; page 13627, first column, first paragraph, indicating that real-time detection of the concentrations has great significance; page 13628, second column, first full paragraph, device for measuring temperatures and conductivities of sodium aluminate liquor; after sampling and processing mechanism model for C K , C A is proposed, and the modelling error is compensated by an SCN model yK and yA is the artificial laboratory value of C K , C A , y Km and y Am is the output of the mechanism model, and e K and e A , the error between them, is the output of the SCN based compensation model ) ; and the caustic concentration detection error is a difference between a laboratory value of a caustic concentration and a caustic concentration measured by an online caustic concentration detection instrument ( e.g. page 13628, second column, first full paragraph, device for measuring temperatures and conductivities of sodium aluminate liquor; after sampling and processing mechanism model for C K , C A is proposed, and the modelling error is compensated by an SCN model yK and yA is the artificial laboratory value of C K , C A , y Km and y Am is the output of the mechanism model, and e K and e A , the error between them, is the output of the SCN based compensation model ) . Accordingly, it would have been obvious to one of ordinary skill in the art before the effective filing date of the invention having the teachings of Modi , Jiang, and Wang in front of him to have modified the teachings of Modi (directed to dynamic construction and online deployment of operation-centric first-principles model for predictive analytics) and Jiang (directed to edge-intelligence for stability guaranteed real-time control systems ), to incorporate the teachings of Wang (directed to prediction of component concentrations in sodium aluminate liquor using network-based models ) to include the capability to implement the system in an alumina production system, where the o nline model predicts a caustic concentration detection error of the alumina production system in real time and the error is a difference between a laboratory value of a caustic concentration and a caustic concentration measured by an online caustic concentration detection instrument . One of ordinary skill would have been motivated to perform such a modification in order to a provide for online estimate of component concentrations in sodium aluminate liquor, which is significant to implement process optimization and control of the alumina industry as described in Wang ( page 13625, second column, first paragraph ). Claim s 5, 6, 11, 12, 17 , and 18 are rejected under 35 U.S.C. 103 as being unpatentable over Modi in view of Jiang, further in view of Liu, further in view of Hakateyama, further in view of Wang . With respect to claims 5 , 11 , and 17 , Modi in view of Jiang, further in view of Liu, further in view of Hakateyama teaches all of the limitations of claims 2 , 8 , and 14 as previously discussed . Modi and Jiang do not explicitly disclose wherein the complex industrial system is an alumina production system, and the model is configured to predict a caustic concentration detection error of the alumina production system in real time; and the caustic concentration detection error is a difference between a laboratory value of a caustic concentration and a caustic concentration measured by an online caustic concentration detection instrument. However, Wang teaches wherein the complex industrial system is an alumina production system ( e.g. page 13625, abstract, alumina production process ), and the online deep learning prediction model is configured to predict a caustic concentration detection error of the alumina production system in real time ( e.g. page 13625, abstract, caustic hydroxide, alumina, and sodium carbonate, concentrations of which are respectively represented by C K , C A , and C C ; page 13626, second column, second paragraph,