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 .
This action is in response to the applicant’s communication filed on 7/10/2026
Claims 1-8, and 10-20 are pending
Claim 9 is canceled
Response to Arguments
Applicant’s arguments with respect to claim(s) 1 and 12 have been considered but are moot because the new ground of rejection does not rely on any reference applied in the prior rejection of record for any teaching or matter specifically challenged in the argument.
Claim Rejections - 35 USC § 103
In the event the determination of the status of the application as subject to AIA 35 U.S.C. 102 and 103 (or as subject to pre-AIA 35 U.S.C. 102 and 103) is incorrect, any correction of the statutory basis (i.e., changing from AIA to pre-AIA ) for the rejection will not be considered a new ground of rejection if the prior art relied upon, and the rationale supporting the rejection, would be the same under either status.
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 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.
Claim(s) 1-2, 10-12, and 14-15 is/are rejected under 35 U.S.C. 103 as being unpatentable over Haesendonckx et al. USPGPUB 2012/0226376 A1 (hereinafter Haesendonckx) in view of Zhao et al. USPGPUB 2021/0389737 A1 (hereinafter Zhao), and further in view of Yagi et al. USPGPUB 2024/0042665 A1 (hereinafter Yagi).
Regarding claim 1, Haesendonckx teaches a method for minimizing a deviation of a physical parameter of a blow-molded container from a target value (Par. [0017] “with the use of a simulation model, based on the measured parameters characterizing the blow molding process, at least one property of the finished blow molded container is computed and is compared to a desired value and that, based upon an eventual deviation between the desired value and the actual value, the parameter influencing the blow molding process is changed in such a way that an eventually remaining deviation is minimized”), the method comprising:
determining a physical parameter of a container assigned to a machine parameter value of a blow molding machine (Par. [0064] “The measuring device 41 measures the wall thickness or a wall thickness distribution in the area of the blow molded container” … “The temperature profile typically concerns the temperature distribution of the preform 1 in its longitudinal direction and/or the temperature distribution between the inner or outer limitations of the preform walls.”);
based on the physical parameter and the target value, determining a change in the machine parameter (Par. [0064] “Taking into consideration the information which has been
taken in, the simulation model 48 makes available the correction values 57 for the temperature profile; Par. [0010] - [0011]” A significant parameter for predetermining the resulting material distribution is the heat distribution realized prior to blow molding in the preforms.” … “through the wall of the preform, a suitable temperature profile from the outside to the inside is predetermined. Basically, it can be assumed that areas of the preform with a lower temperature lead to thicker wall areas of the blown container, and that the warmer areas of the preform are stretched to a greater extent when carrying out the blow molding deformation and, as a result, lead to thinner wall areas of the blow molded container.”);
determining an optimal machine parameter value for achieving a minimum deviation from the target value of the physical parameter of a blow-molded container (Par. [0017] “the parameter influencing the blow molding process is changed in such a way that an eventually remaining deviation is minimized”; Par. [0059] “the simulation model has one or more model outlets 50 which facilitate influencing the regulating behavior. In accordance with an embodiment, the regulating characteristic is changed by means of one of the regulators 42, 43, 45 through the model outlet 50.”; Par. [0061] “regulation of a container manufacture can take place, for example, on the basis of a predetermined pressure pattern for the blow pressure. If in a comparison of the measured values with the values generated by the simulation model, deviations in at least one of the measured parameters are recognized, for example, for each of the product cycles, the starting point for supplying the pre-blowing pressure can be changed and/or it is possible to increase or decrease the speed of the stretching process in a suitable manner”…”It is also being considered to carry out an adjustment or the temperature profile and/or the heating power.”); and
controlling the blow molding machine based on the obtained optimal machine parameter value (Par. [0059] “regulating characteristic is changed by means of one of the regulators 42, 43, 45 through the model outlet 50”).
Haesendonckx does not explicitly teach based on an iteration process, determining an optimal machine parameter value for achieving a minimum deviation from the target value of the physical parameter of a blow-molded container, the iteration process comprising:
a first iteration step for determining a deviation from the target value of the physical parameter of a blow-molded container based on the change in the machine parameter value; and
a second iteration step for determining an adjusted change in the machine parameter value based on the determined deviation of the physical parameter of a blow-molded container from the target value,
wherein the adjusted change in the machine parameter value determined by the second iteration step is subsequently provided as an input to the first iteration step, in which a new deviation from the target value of the physical parameter is determined based on the adjusted change in the machine parameter value;
obtaining the optimal machine parameter value, wherein obtaining the optimal machine parameter value includes determining an optimal adjusted change based on the minimum deviation from the target value of the physical parameter from a set of deviations from the target value of the physical parameter.
However, Zhao teaches based on an iteration process (Par. [0006], “the optimal target value obtained in each step as feedback, thus judging the current optimized direction and gradually iterating to the optimal process parameter.”; Par. [0055], “uses an iterative gradient estimation method to calculate the gradient direction of the current point as a parameter adjustment direction, thus optimizing the parameters constantly”) determining an optimal machine parameter value for achieving a minimum deviation from the target value of the physical parameter of a blow-molded container (Par. [0057], “determining an optimized quality index Q (e.g., product weight, warping, and the like), a quality index target value Qtarget and a permissible error δ of an optimization objective”; Formula (1); Par. [0058], “J(X) denotes the difference between the current quality index value and the target value … X consists of n parameters x1 ,x2, …, xn, such as, packing pressure, melt temperature” – Zhao optimizes molding process parameters to minimize the difference between a product quality value and its target value.), the iteration process comprising:
a first iteration step for determining a deviation from the target value of the physical parameter of a blow-molded container based on the change in the machine parameter value (Par. [0060], “each parameter is disturbed respectively … to obtain n+1 sets of process parameters and the corresponding quality values after perturbation, thereby obtaining the corresponding optimal target value J (a difference value between the corresponding quality value and quality target value)” – Zhao changes the process parameters and determines the corresponding difference between the resulting quality value and the quality target value.); and
a second iteration step for determining an adjusted change in the machine parameter value based on the determined deviation of the physical parameter of a blow-molded container from the target value (Par. [0027], “further, the gradient value of the current parameter sample at X0 can be calculated by the following Formula”; Par. [0029], “The reverse direction of the gradient calculated by the Formula … is an optimized direction of the parameter; a next iteration point is produced by the following Formula, and the obtained new process parameter will be closer to the preset target value”; Par. [0030], “where, αo is a step size of parameter adjustment; X0 is a current parameter sample; Xupdate0 is the updated current parameter sample; and ∇J(X0) is a gradient value.” – Zhao uses the gradient calculated from the target difference values to determine the direction and step for updating the process parameters.),
wherein the adjusted change in the machine parameter value determined by the second iteration step is subsequently provided as an input to the first iteration step (Par. [0032], “on the contrary, it cannot satisfy the termination conditions, update the parameter sample in step (3), and return to step (4) to perform the next optimal computation.”; Par. [0033], “A newly obtained … is used to substitute the point whose target value is the worst in the primary gradient computation matrix, and to be a current parameter sample” – the updated process parameter is used as a current parameter sample in the next optimization computation.), in which a new deviation from the target value of the physical parameter is determined based on the adjusted change in the machine parameter value (Par. [0030], “After obtaining the updated parameter Xupdate0, the updated parameter needs to be de-normalized (as shown in Formula (2)), so that the parameter returns to a physical parameter value (Xdenorm), which facilitates technicians to obtain a target value J(Xupdate0 ), J(Xupdate0)=Q(Xupdate0)−Qtarget at Xupdate0 through experimental operation.”; Par. [0031], “if a simulation method is used to calculate the quality index value or target value, the Xupdate0 obtained can be directly used for judgment by programming.”; Par. [0033], “A newly obtained (Xupdate0, J(Xupdate0)) is used to substitute the point whose target value is the worst in the primary gradient computation matrix, and to be a current parameter sample” – Zhao determines a new target difference value corresponding to the updated process parameter.);
obtaining the optimal machine parameter value (Par. [0032], “if |Q2 −Qtarget|<δ, the optimization objective obtained has satisfied termination conditions, and the optimization progress stops; Xupdate0 is the optimal parameter, and after physical transformation, the actual optimal parameter information can be output”).
Haesendonckx and Zhao are analogous art because they contain functional similarities. They both determine a deviation between a quality parameter and a target value and adjust a process parameter based on the deviation.
Therefore, at the time of effective filing date, it would have been obvious to a person of ordinary skill in the art to modify the above blow molding method, as taught by Haesendonckx, and incorporate an iterative process parameter optimization, as taught by Zhao.
One of ordinary skill in the art would have been motivated to improve the efficiency of determining optimal process parameters, as suggested by Zhao (Par. [0006]).
Haesendonckx and Zhao do not explicitly teach wherein obtaining the optimal machine parameter value includes determining an optimal adjusted change based on the minimum deviation from the target value of the physical parameter from a set of deviations from the target value of the physical parameter.
However, Yagi teaches wherein obtaining the optimal machine parameter value includes determining an optimal adjusted change based on the minimum deviation from the target value of the physical parameter from a set of deviations from the target value of the physical parameter (Par. [0149], “calculating the predicted value of the objective variable using the trained regression model with respect to all combinations of explanatory variables configured by combining the candidate values of each parameter, and extracting a combination of explanatory variables having a smallest deviation between the predicted value and the target value”; Par. [0152], “The training and optimization system 4 calculates the weight of the molded article using the trained regression model and the standardization information on the training database for each combination of parameters”; Par. [0153], “the training and optimization system 4 generates the molding condition closest to the target weight as the optimal condition by calculating absolute values of differences between the predicted value and the target value and sorting the absolute values of the differences in ascending order” – When applied to the adjusted process parameters taught by Zhao, Yagi’s selection of the molding condition having the smallest deviation determines the adjusted process parameter corresponding to the minimum deviation as the optimal adjusted change.).
Haesendonckx, Zhao, and Yagi are analogous art because they contain functional similarities. They determine a deviation between a quality parameter and a target value and use the deviation in determining a molding process parameter.
Therefore, at the time of effective filing date, it would have been obvious to a person of ordinary skill in the art to modify the above blow molding method, as taught by Haesendonckx and Zhao, and incorporate selecting the molding condition corresponding to the smallest deviation from the target value, as taught by Yagi.
One of ordinary skill in the art would have been motivated to improve the determination of an optimal molding condition that most closely achieves the target value, as suggested by Yagi (Par. [0149] – [0153]).
Regarding claim 2, the combination of Haesendonckx, Zhao, and Yagi teaches all the limitations of the base claims as outlined above.
Haesendonckx further teaches wherein a predictive model is used to determine the deviation from the target value of the physical parameter of a blow-molded container (Par. [0017] “with the use of a simulation model, based on the measured parameters characterizing the blow molding process, at least one property of the finished blow molded container is computed and is compared to a desired value and that, based upon an eventual deviation between the desired value and the actual value”).
Regarding claim 10, the combination of Haesendonckx, Zhao, and Yagi teaches all the limitations of the base claims as outlined above.
Haesendonckx further teaches wherein the physical parameter of the blow-molded container comprises a wall thickness, a variable characteristic of the wall thickness, a bottom thickness, a variable characteristic of the bottom thickness, and/or a molecular orientation (Par. [0054] “The immediate input value for the wall thickness regulator 42 is the regulating difference between a predetermined wall thickness and the measured wall thickness” – Haesendonckx teaches the wall thickness alternative.).
Regarding claim 11, the combination of Haesendonckx, Zhao, and Yagi teaches all the limitations of the base claims as outlined above.
Haesendonckx further teaches wherein, in addition to determining the physical parameter of the container assigned to the machine parameter value of the blow molding machine, the method further comprises determining a disturbance variable, wherein the disturbance variable is an environmental condition and/or a property of a preform (Par. [0024] “A simulation which is close to reality can be carried out by evaluating at least one parameter characterizing the blow molding process, selected from the group including preform temperature, temperature distribution in the preform, blow pressure sequence, stretching sequence, material properties, material distribution in the preform, material properties in the blow molded container, container contour, ambient parameters” – the preform temperature, temperature distribution, and material distribution are properties of a preform, and the ambient parameters correspond to environmental conditions.).
Regarding claim 12, Haesendonckx teaches a blow molding machine for producing containers (Par. [0002] “device for blow molding containers”), comprising:
a sensor device (Par. [0002] “at least one sensor for measuring at least one parameter”); and
a control apparatus (Par. [0002] “sensor is connected to the control device”),
wherein the sensor device is configured to determine a physical parameter of a container (Par. [0012] “measurement of an actual wall thickness in the area of the blown containers can be effected by means of so-called wall thickness sensors”) assigned to a machine parameter value of the blow molding machine and to pass the machine parameter value and the physical parameter to the control apparatus (Par. [0001], “at least one parameter characterizing the blow molding process is measured and supplied to a control device, which acts on at least one adjusting element for changing at least one of the parameters which influence the blow molding process “; Par. [0059] “At least one of the measurement values supplied by the sensors 41, 44, 46 is supplied to a simulation model 48” – simulation model is used by the control device),
wherein the control apparatus is configured to:
based on the physical parameter and a target value, determine a change in the machine parameter value (Par. [0019] “based on the measured parameters characterizing the blow molding process, at least one property of the finished blow molded container is computed and is compared to a desired value and that, based upon an eventual deviation between the desired value and the actual value, the parameter influencing the blow molding process is changeable by the control element”);
determine an optimal machine parameter value for achieving a minimum deviation from the target value of the physical parameter (Par. [0017] “the parameter influencing the blow molding process is changed in such a way that an eventually remaining deviation is minimized”; Par. [0059] “the simulation model has one or more model outlets 50 which facilitate influencing the regulating behavior. In accordance with an embodiment, the regulating characteristic is changed by means of one of the regulators 42, 43, 45 through the model outlet 50.”; Par. [0061] “regulation of a container manufacture can take place, for example, on the basis of a predetermined pressure pattern for the blow pressure. If in a comparison of the measured values with the values generated by the simulation model, deviations in at least one of the measured parameters are recognized, for example, for each of the product cycles, the starting point for supplying the pre-blowing pressure can be changed and/or it is possible to increase or decrease the speed of the stretching process in a suitable manner”…”It is also being considered to carry out an adjustment or the temperature profile and/or the heating power.”); and
control the blow molding machine based on the obtained optimal machine parameter value (Par. [0059] “regulating characteristic is changed by means of one of the regulators 42, 43, 45 through the model outlet 50”).
Haesendonckx does not explicitly teach based on an iteration process, determine an optimal machine parameter value for achieving a minimum deviation from the target value of the physical parameter, wherein the iteration process comprises:
a first iteration step for determining a deviation from the target value of the physical parameter of a blow-molded container based on the change in the machine parameter value;
a second iteration step for determining an adjusted change in the machine parameter value based on the determined deviation of the physical parameter of a blow-molded container from the target value, and
wherein the adjusted change in the machine parameter determined by the second iteration step is subsequently provided as an input to the first iteration step, in which a new deviation from the target value of the physical parameter is determined based on the adjusted change in the machine parameter value; and
obtaining the optimal machine parameter value, wherein obtaining the optimal machine parameter value includes determining an optimal adjusted change based on the minimum deviation from the target value of the physical parameter from a set of deviations from the target value of the physical parameter.
However, Zhao teaches based on an iteration process (Par. [0006], “the optimal target value obtained in each step as feedback, thus judging the current optimized direction and gradually iterating to the optimal process parameter.”; Par. [0055], “uses an iterative gradient estimation method to calculate the gradient direction of the current point as a parameter adjustment direction, thus optimizing the parameters constantly”), determine an optimal machine parameter value for achieving a minimum deviation from the target value of the physical parameter (Par. [0057], “determining an optimized quality index Q (e.g., product weight, warping, and the like), a quality index target value Qtarget and a permissible error δ of an optimization objective”; Formula (1); Par. [0058], “J(X) denotes the difference between the current quality index value and the target value … X consists of n parameters x1 ,x2, …, xn, such as, packing pressure, melt temperature” – Zhao optimizes molding process parameters to minimize the difference between a product quality value and its target value), wherein the iteration process comprises:
a first iteration step for determining a deviation from the target value of the physical parameter of a blow-molded container based on the change in the machine parameter value (Par. [0060], “each parameter is disturbed respectively … to obtain n+1 sets of process parameters and the corresponding quality values after perturbation, thereby obtaining the corresponding optimal target value J (a difference value between the corresponding quality value and quality target value)” – Zhao changes the process parameters and determines the corresponding difference between the resulting quality value and the quality target value.);
a second iteration step for determining an adjusted change in the machine parameter value based on the determined deviation of the physical parameter of a blow-molded container from the target value (Par. [0027], “further, the gradient value of the current parameter sample at X0 can be calculated by the following Formula”; Par. [0029], “The reverse direction of the gradient calculated by the Formula … is an optimized direction of the parameter; a next iteration point is produced by the following Formula, and the obtained new process parameter will be closer to the preset target value”; Par. [0030], “where, αo is a step size of parameter adjustment; X0 is a current parameter sample; Xupdate0 is the updated current parameter sample; and ∇J(X0) is a gradient value.” – Zhao uses the gradient calculated from the target difference values to determine the direction and step for updating the process parameters.), and
wherein the adjusted change in the machine parameter determined by the second iteration step is subsequently provided as an input to the first iteration step (Par. [0032], “on the contrary, it cannot satisfy the termination conditions, update the parameter sample in step (3), and return to step (4) to perform the next optimal computation.”; Par. [0033], “A newly obtained … is used to substitute the point whose target value is the worst in the primary gradient computation matrix, and to be a current parameter sample” – the updated process parameter is used as a current parameter sample in the next optimization computation), in which a new deviation from the target value of the physical parameter is determined based on the adjusted change in the machine parameter value (Par. [0030], “After obtaining the updated parameter Xupdate0, the updated parameter needs to be de-normalized (as shown in Formula (2)), so that the parameter returns to a physical parameter value (Xdenorm), which facilitates technicians to obtain a target value J(Xupdate0 ), J(Xupdate0)=Q(Xupdate0)−Qtarget at Xupdate0 through experimental operation.”; Par. [0031], “if a simulation method is used to calculate the quality index value or target value, the Xupdate0 obtained can be directly used for judgment by programming.”; Par. [0033], “A newly obtained (Xupdate0, J(Xupdate0)) is used to substitute the point whose target value is the worst in the primary gradient computation matrix, and to be a current parameter sample” – Zhao determines a new target difference value corresponding to the updated process parameter.); and
obtaining the optimal machine parameter value (Par. [0032], “if |Q2 −Qtarget|<δ, the optimization objective obtained has satisfied termination conditions, and the optimization progress stops; Xupdate0 is the optimal parameter, and after physical transformation, the actual optimal parameter information can be output”),
Haesendonckx and Zhao are analogous art because they contain functional similarities. They both determine a deviation between a quality parameter and a target value and adjust a process parameter based on the deviation.
Therefore, at the time of effective filing date, it would have been obvious to a person of ordinary skill in the art to modify the above blow molding machine, as taught by Haesendonckx, and incorporate an iterative process parameter optimization, as taught by Zhao.
One of ordinary skill in the art would have been motivated to improve the efficiency of determining optimal process parameters, as suggested by Zhao (Par. [0006]).
Haesendonckx and Zhao do not explicitly teach wherein obtaining the optimal machine parameter value includes determining an optimal adjusted change based on the minimum deviation from the target value of the physical parameter from a set of deviations from the target value of the physical parameter.
However, Yagi teaches wherein obtaining the optimal machine parameter value includes determining an optimal adjusted change based on the minimum deviation from the target value of the physical parameter from a set of deviations from the target value of the physical parameter (Par. [0149], “calculating the predicted value of the objective variable using the trained regression model with respect to all combinations of explanatory variables configured by combining the candidate values of each parameter, and extracting a combination of explanatory variables having a smallest deviation between the predicted value and the target value”; Par. [0152], “The training and optimization system 4 calculates the weight of the molded article using the trained regression model and the standardization information on the training database for each combination of parameters”; Par. [0153], “the training and optimization system 4 generates the molding condition closest to the target weight as the optimal condition by calculating absolute values of differences between the predicted value and the target value and sorting the absolute values of the differences in ascending order” – When applied to the adjusted process parameters taught by Zhao, Yagi’s selection of the molding condition having the smallest deviation determines the adjusted process parameter corresponding to the minimum deviation as the optimal adjusted change.).
Haesendonckx, Zhao, and Yagi are analogous art because they contain functional similarities. They determine a deviation between a quality parameter and a target value and use the deviation in determining a molding process parameter.
Therefore, at the time of effective filing date, it would have been obvious to a person of ordinary skill in the art to modify the above blow molding machine, as taught by Haesendonckx and Zhao, and incorporate selecting the molding condition corresponding to the smallest deviation from the target value, as taught by Yagi.
One of ordinary skill in the art would have been motivated to improve the determination of an optimal molding condition that most closely achieves the target value, as suggested by Yagi (Par. [0149] – [0153]).
Regarding claim 14, the combination of Haesendonckx, Zhao, and Yagi teaches all the limitations of the base claims as outlined above.
Haesendonckx further teaches wherein the sensor device comprises a sensor configured to determine the physical parameter of the blow-molded container (Par. [0012] “measurement of an actual wall thickness in the area of the blown containers can be effected by means of so-called wall thickness sensors which operate for example optically, or with the use of sound waves.”).
Regarding claim 15, the combination of Haesendonckx, Zhao, and Yagi teaches all the limitations of the base claims as outlined above.
Haesendonckx further teaches wherein the sensor is configured to determine a wall thickness, a variable characteristic of the wall thickness, a bottom thickness, a variable characteristic of the bottom thickness, and/or a molecular orientation of a blow-molded container (Par. [0012] “measurement of an actual wall thickness in the area of the blown containers can be effected by means of so-called wall thickness sensors which operate for example optically, or with the use of sound waves.” – optical and sound wave operated wall thickness sensors can determine wall thickness and variable characteristics of both walls and bottoms).
Claim(s) 3-8, 13, 16-20 is/are rejected under 35 U.S.C. 103 as being unpatentable over Haesendonckx et al. USPGPUB 2012/0226376 A1 (hereinafter Haesendonckx) in view of Zhao et al. USPGPUB 2021/0389737 A1 (hereinafter Zhao) and Yagi et al. USPGPUB 2024/0042665 A1 (hereinafter Yagi), and further in view of Sun et al. CN 108830376 A (hereinafter Sun).
Regarding claim 3, the combination of Haesendonckx, Zhao, and Yagi teaches all the limitations of the base claims as outlined above.
Haesendonckx, Zhao, and Yagi do not explicitly teach wherein the predictive model comprises a first neural network.
However, Sun teaches wherein the predictive model comprises a first neural network (Par. [0040] “The current value network group contains multiple value networks. Each value network is a neural network model. Each neural network model takes the current environment state as input and outputs a series of Q-values.”; Par. [0052], “The target value network has the same structure as the current value network in the decision machine. The content of the target value network is copied from the neural network model in the decision machine every N steps”; Par. [0053], “The target value network calculates the target value Ym”).
Haesendonckx, Zhao, Yagi, and Sun are analogous art because they contain functional similarities. They use computational models to determine an output based on input process or state information and use the determined output in selecting or adjusting a parameter.
Therefore, at the time of effective filing date, it would have been obvious to a person of ordinary skill in the art to modify the above blow molding method for iteratively determining an optimal machine parameter value, as taught by Haesendonckx, Zhao, and Yagi, and incorporate a neural network in the predictive model, as taught by Sun.
One of ordinary skill in the art would have been motivated to improve learning efficiency as suggested by Sun (Par. [0017]).
Regarding claim 4, the combination of Haesendonckx, Zhao, Yagi, and Sun teaches all the limitations of the base claims as outlined above.
Sun further teaches wherein the iteration process is based on a reinforcement learning model (Par. [0041], “reinforcement learning is used to learn the neural network model to be used in the next round”; Par. [0046] “Through reinforcement learning, we can use the model to automatically select the neural network model to be used in the next moment in order to obtain the maximum reward.”).
Regarding claim 5, the combination of Haesendonckx, Zhao, Yagi, and Sun teaches all the limitations of the base claims as outlined above.
Sun further teaches wherein the reinforcement learning model comprises a first component and a second component, and wherein by an interaction of the first component with the second component, the optimal machine parameter for minimizing the deviation of the physical parameter of the blow-molded container from the target value is obtained (Par. [0052], “The target value network has the same structure as the current value network in the decision machine”; Par. [0053], “The target value network calculates the target value Ym”; Par. [0040] “Each value network is a neural network model. Each neural network model takes the current environment state as input and outputs a series of Q-values”; Par. [0057], “The DQN error function update unit uses an error function Lm that minimizes the mean square objective value Ym and the Q value output by the decision machine” – the target value network corresponds to the first component and the selected current value network corresponds to the second component, which interact through the reinforcement learning process to determine the selected action.).
Regarding claim 6, the combination of Haesendonckx, Zhao, Yagi, and Sun teaches all the limitations of the base claims as outlined above.
Sun further teaches wherein the first component of the reinforcement learning model comprises the predictive model (Par. [0052], “The target value network has the same structure as the current value network in the decision machine.”; Par. [0053], “The target value network calculates the target value Ym” – the target value network corresponding to the first component performs the predictive value determination and corresponds to the predictive model.).
Regarding claim 7, the combination of Haesendonckx, Zhao, Yagi, and Sun teaches all the limitations of the base claims as outlined above.
Sun further teaches wherein the second component of the reinforcement learning model comprises a third neural network (Par. [0040], “The current value network group contains multiple value networks. Each value network is a neural network model.”; Par. [0078] “third convolutional neural network model contains three convolutional layers and two fully connected output layers.” – the selected current value network corresponding to the second component can comprise the third neural network.).
Regarding claim 8, the combination of Haesendonckx, Zhao, Yagi, and Sun teaches all the limitations of the base claims as outlined above.
Sun further teaches wherein the first iteration step is performed by the first neural network (Par. [0052], “The target value network has the same structure as the current value network in the decision machine”; Par. [0053], “The target value network calculates the target value Ym” – the target value network corresponding to the predictive model and first neural network determines the target-related value used in the first iteration step.), and the second iteration step is performed by the third neural network (Par. [0027], “Based on the current environmental state, a series of Q-values are calculated using the previously determined or preset neural network model … This allows the selection of the action that maximizes the Q-value“; Par. [0078], “The third convolutional neural network model contains three convolutional layers and two fully connected output layers.” - the third neural network determines the Q-values used to determine the selected action corresponding to the adjusted change.).
Regarding claim 13, the combination of Haesendonckx, Zhao, and Yagi teaches all the limitations of the base claims as outlined above.
Haesendonckx further teaches wherein a predictive model is provided for determining the deviation from the target value of the physical parameter of a container (Par. [0017] “with the use of a simulation model, based on the measured parameters characterizing the blow molding process, at least one property of the finished blow molded container is computed and is compared to a desired value and that, based upon an eventual deviation between the desired value and the actual value”).
Haesendonckx, Zhao, and Yagi do not explicitly teach wherein:
the iteration process is based on a reinforcement learning model;
the reinforcement learning model comprises a first and a second component;
the first component is the predictive model and comprises a first neural network; and
the second component comprises a second and a third neural network.
However, Sun teaches wherein:
the iteration process is based on a reinforcement learning model (Par. [0002] “This invention relates to the field of artificial intelligence technology, and in particular to a deep reinforcement learning method for multi-value networks in time-sensitive environments”);
the reinforcement learning model comprises a first and a second component (Par. [0048], “The learner includes: a memory playback unit, a target value network, and a DQN error function update unit.”; Par. [0040], “The current value network group contains multiple value networks” – the target value network corresponds to the first component and the current value network group corresponds to the second component.);
the first component is the predictive model and comprises a first neural network (Par. [0052], “The target value network has the same structure as the current value network in the decision machine. The content of the target value network is copied from the neural network model in the decision machine every N steps”; Par. [0053], “The target value network calculates the target value Ym” – the target value network performs the predictive value determination and has the structure of a neural network.); and
the second component comprises a second and a third neural network (Par. [0040], “The current value network group contains multiple value networks. Each value network is a neural network model.”; Par. [0078] “The second convolutional neural network model contains two convolutional layers and two fully connected output layers. The third convolutional neural network model contains three convolutional layers and two fully connected output layers”).
Haesendonckx, Zhao, Yagi, and Sun are analogous art because they contain functional similarities. They use calculated values derived from process or state inputs to select or adjust parameters toward an optimization objective.
Therefore, at the time of effective filing date, it would have been obvious to a person of ordinary skill in the art to modify the above blow molding process parameter optimization system, as taught by Haesendonckx, Zhao, and Yagi, and incorporate predictive and iterative optimization using reinforcement learning neural network architecture, as taught by Sun.
One of ordinary skill in the art would have been motivated to improve learning efficiency as suggested by Sun (Par. [0017]).
Regarding claim 16, the combination of Haesendonckx, Zhao, and Yagi teaches all the limitations of the base claims as outlined above.
Haesendonckx, Zhao, and Yagi do not explicitly teach wherein the predictive model is a first neural network.
However, Sun teaches wherein the predictive model is a first neural network (Par. [0040] “Each value network is a neural network model. Each neural network model takes the current environment state as input and outputs a series of Q-values”; Par. [0052], “The target value network has the same structure as the current value network in the decision machine; Par. [0053], “The target value network calculates the target value Ym”).
Haesendonckx, Zhao, Yagi, and Sun are analogous art because they contain functional similarities. They use computational models to determine output values from input process or state information and use the determined values in selecting or adjusting parameters.
Therefore, at the time of effective filing date, it would have been obvious to a person of ordinary skill in the art to modify the above blow molding process parameter optimization method, as taught by Haesendonckx, Zhao, and Yagi, and incorporate a neural network as the predictive model, as taught by Sun.
One of ordinary skill in the art would have been motivated to improve learning efficiency as suggested by Sun (Par. [0017]).
Regarding claim 17, the combination of Haesendonckx, Zhao, Yagi, and Sun teaches all the limitations of the base claims as outlined above.
Sun further teaches wherein the first component of the reinforcement learning model is the predictive model (Par. [0053], “The target value network calculates the target value Ym” – the target value network corresponding to the first component is the predictive model.).
Regarding claim 18, the combination of Haesendonckx, Zhao, Yagi, and Sun teaches all the limitations of the base claims as outlined above.
Sun further teaches wherein the first component of the reinforcement learning model is the first neural network or comprises the first neural network (Par. [0052], “The target value network has the same structure as the current value network in the decision machine.”; Par. [0040], “Each value network is a neural network model.” – the target value network corresponding to the first component is a neural network.).
Regarding claim 19, the combination of Haesendonckx, Zhao, Yagi, and Sun teaches all the limitations of the base claims as outlined above.
Sun further teaches wherein the second component of the reinforcement learning model consists of a third neural network (Par. [0078] “third convolutional neural network model contains three convolutional layers and two fully connected output layers.” – the selected third convolutional neural network model corresponds to the second component.).
Regarding claim 20, the combination of Haesendonckx, Zhao, Yagi, and Sun teaches all the limitations of the base claims as outlined above.
Sun further teaches wherein the first iteration step is performed by the predictive model (Par. [0016] “The target value network is used to calculate the target value by combining the current environmental state, the reward value of the previous action, and the action selected by the neural network model.” – the target value network corresponds to the predictive model.), and the second iteration step is performed by the third neural network (Par. [0027], “Based on the current environmental state, a series of Q-values are calculated using the previously determined or preset neural network model … This allows the selection of the action that maximizes the Q-value“; Par. [0078], “The third convolutional neural network model contains three convolutional layers and two fully connected output layers” - the third neural network calculates the Q-values used to determine the selected action corresponding to the adjusted change.).
Citation of Pertinent Prior Art
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure.
Cowden et al [US 11772317 B2] teaches an inspection system capable of measuring the material distribution of individual blow-molded containers. The measurements are made using a series of emitters and sensors that are located either within or downstream of the blow molder.
Diraddo et al. [Modeling of Membrane Inflation in Blow Molding: Neural Network Prediction of Initial Dimensions from Final Part Specifications, 1993] teaches the use of neural networks in the modeling of the inflation stage of the blow molding process.
Duesterhoeft [DE 10000859 A1] teaches an automatic process for reshaping a thin side wall of a compartment involving measuring deviations between actual and set geometries.
Jebadurai et al. [WO 2020/226921 A1] teaches a computer-based predictive preventative maintenance system that uses a data model to make preventative maintenance predictions for a container-forming apparatus that produces containers.
Conclusion
Any inquiry concerning this communication or earlier communications from the examiner should be directed to PETER XU whose telephone number is (571)272-0792. The examiner can normally be reached Monday-Friday 9am-5pm.
Examiner interviews are available via telephone, in-person, and video conferencing using a USPTO supplied web-based collaboration tool. To schedule an interview, applicant is encouraged to use the USPTO Automated Interview Request (AIR) at http://www.uspto.gov/interviewpractice.
If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, Mohammad Ali can be reached at (571) 272-4105. The fax phone number for the organization where this application or proceeding is assigned is 571-273-8300.
Information regarding the status of published or unpublished applications may be obtained from Patent Center. Unpublished application information in Patent Center is available to registered users. To file and manage patent submissions in Patent Center, visit: https://patentcenter.uspto.gov. Visit https://www.uspto.gov/patents/apply/patent-center for more information about Patent Center and https://www.uspto.gov/patents/docx for information about filing in DOCX format. For additional questions, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free). If you would like assistance from a USPTO Customer Service Representative, call 800-786-9199 (IN USA OR CANADA) or 571-272-1000.
/PETER XU/ Examiner, Art Unit 2119
/MOHAMMAD ALI/ Supervisory Patent Examiner, Art Unit 2119
