Notice of Pre-AIA or AIA Status
1. The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
Detailed Action
2. This Final Office Action is responsive to Applicants’ Reply received 3/17/26. Subject to the Reply and its amendments, claims 18 and 22-33 are now pending, of which claims 18, 30, and 32-33 are independent.
Claim Rejections - 35 USC § 103
3. 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.
4. 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.
5. Claim 33 is rejected under 35 U.S.C. 103 as being unpatentable over KR 101822613 B1 (“Kang”) in view of Non-Patent Literature “Spot size, laser quality and welding performance” (“Verhaeghe”) and further in view of WO 2019239644 A1 (“Kondou”).
Regarding claim 33, KANG teaches A non-transitory machine-readable memory medium on which is stored a computer program (page 2 of the reference as printed, see e.g., the discussion found in the 3rd-4th paragraphs under the printed heading “DESCRIPTION-OF-EMBODIMENTS”, which detail the design, training, and inference/use of the trained artificial neural network, which one of ordinary skill in the art would understand to be a computer-implemented model thereby requiring a memory element for storing the model once designed, trained, and maintained for inference) for training a data-based model (page 2: “An artificial neural network (ANN) is a kind of function that connects inputs and outputs by connecting a virtual neuron device, a perceptron, which is a neuron of the human neuron, and a network. Perceptron's complex network makes it possible to reproduce or predict complex nonlinear phenomena relatively simply. Prediction methods using artificial neural networks are divided into learning stage and production stage. In the learning phase, a series of input / output relationships are supplied and thus a functional relationship is identified in the artificial neural network.” (where the learning stage as taught is akin to the recitation for training a data-based model, where the data used for the learning stage is further clarified on the reference’s page 3 (“spot welding data DB may be an existing spot welding data acquired and stored in advance by the experiment”))) to ascertain a variable which characterizes an energy input of a ... welding machine into a workpiece (page 3: “The input value acquiring step (S2) is a step of acquiring, as an input value, a part of the input power value data according to the time derived from the starting point to the end point of the spot welding ... the input value obtained is a value obtained by sampling the input power value from the start to the end of the welding in a predetermined time unit”, where the acquired input is processed by the neural network model to provide an output, e.g. page 4’s discussion of “a welding characteristic result output step (S3) for inputting the input value obtained in the input value obtaining step (S2) to the learned artificial neural network (ART) and outputting a welding characteristic result corresponding to the input value”), as a function of operating parameters of the ... welding machine (per page 4 of the reference, where the aforementioned output result is understood by the Examiner to be a characterization of the inputs as processed, where the welding characteristic result is understood by the Examiner to indicate a prediction corresponding to the welding processing conditions (e.g., current, voltage, time, pressuring force, etc.)), the computer program, when executed by a computer, causing the computer to perform the following:
training the data-based model ... and wherein the data-based model is trained to output as a function of the operating parameters the ascertained variable characterizing the energy input as a model output variable (as discussed above, page 2’s teaching where “Prediction methods using artificial neural networks are divided into learning stage and production stage” (where the learning stage as taught is akin to the recitation for training a data-based model, where the data used for the learning stage is further clarified on the reference’s page 3 (“spot welding data DB may be an existing spot welding data acquired and stored in advance by the experiment”)), and further where the acquired input is processed by the neural network model to provide an output, e.g. page 4’s discussion of “a welding characteristic result output step (S3) for inputting the input value obtained in the input value obtaining step (S2) to the learned artificial neural network (ART) and outputting a welding characteristic result corresponding to the input value”, where the aforementioned output result is understood by the Examiner to be a characterization of the inputs as processed, where the welding characteristic result is understood by the Examiner to indicate a prediction corresponding to the welding processing conditions (e.g., current, voltage, time, pressuring force, etc.) ).
Kang is silent as to its welding being explicitly laser welding, e.g. as specifically further recited, but the Examiner believes one of ordinary skill in the art would understand Kang’s operations and modeling and considerations to be reasonably extensible to a laser welding operation/paradigm. However, to the extent that such a conjecture is not soundly supported, the Examiner refers to VERHAEGHE to establish the use of laser welding equipment to perform the same type of welding, e.g. spot welding, that Kang contemplates. Hence, in view of Verhaeghe, the Examiner believes it reasonable to infer the applicability of Kang’s framework to welding that encompasses laser welding as Applicants have recited.
Both Kang and Verhaeghe contemplate improvements to the efficiency and quality of similar welding processes as measurable by like/similar considerations. Hence, they are similarly directed and therefore analogous. It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to apply a laser welding approach as Verhaeghe specifically contemplates to perform the same type of welding operations that Kang seeks to optimize using its taught framework.
Applicants’ claim further clarifies the training to be training the data-based model as a function of an ascertained number of spatters, wherein the ascertained number of spatters depends on the operating parameters, which Kang etc. do not teach. At best, Kang establishes that spattering “is a problem caused by excessive heat” (e.g., per Kang’s page 3) and that such heat is “generated when the current flows” as part of the welding process (e.g., per Kang’s page 2). From this, the Examiner the reasons that Kang contemplates that spattering from an operational condition such as excessive heat which may be generated via the welding operation’s applied current might be problematic. To provide what Kang etc. otherwise lacks, the Examiner relies upon KONDOU, see e.g., Kondou’s framework for generating a learned model that can detect spatter as discussed in relation to its FIG. 1 and more specifically steps S112-S113 (which permit a characterization of the welding conditions such as voltage applied resulting in a determinable number of spatters). Kondou clarifies that the model is learned in accordance with supervised learning using images featuring a number of spatters, and hence from this is follows that spatters and their amount is a consideration of the learning/training process for the model as taught.
Both Kang and Verhaeghe contemplate improvements to the efficiency and quality of similar welding processes as measurable by like/similar considerations. Hence, they are similarly directed and therefore analogous. It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to incorporate a spatter-based model per Kondou to a framework such as Kang’s, with a reasonable expectation of success, such that Kang’s welding process can be subject to monitoring for spatter in a manner that usefully determines an underlying operational condition causing the issue/defect, thereby improving quality/performance in the welding operation.
6. Claims 30-32 are rejected under 35 U.S.C. 103 as being unpatentable over Kang in view of Verhaeghe and Kondou and further in view of Non-Patent Literature “A Tutorial on Bayesian Optimization” (“Frazier”).
Regarding claim 30, Kang teaches A method for setting operating parameters of a ... welding machine ... of a data-based model (per page 4 of the reference, where the output result of modeling inputs and outputs for welding (see pages 3-4, referencing steps S2-S4) allows for a predictive benefit for performing welding operations) is understood by the Examiner to be a characterization of the inputs as processed, where the welding characteristic result is understood by the Examiner to indicate a prediction corresponding to the welding processing conditions (e.g., current, voltage, time, pressuring force, etc.) such that inputs that correlate to favorable outputs can be used to set operation parameters for a favorable operation), the method comprising the following steps:
training the data-based model ... and setting the operating parameters of the laser welding machine using the trained data-based model (as discussed above, page 2’s teaching where “Prediction methods using artificial neural networks are divided into learning stage and production stage” (where the learning stage as taught is akin to the recitation for training a data-based model, where the data used for the learning stage is further clarified on the reference’s page 3 (“spot welding data DB may be an existing spot welding data acquired and stored in advance by the experiment”)), and further where the acquired input is processed by the neural network model to provide an output, e.g. page 4’s discussion of “a welding characteristic result output step (S3) for inputting the input value obtained in the input value obtaining step (S2) to the learned artificial neural network (ART) and outputting a welding characteristic result corresponding to the input value”, where the aforementioned output result is understood by the Examiner to be a characterization of the inputs as processed, where the welding characteristic result is understood by the Examiner to indicate a prediction corresponding to the welding processing conditions (e.g., current, voltage, time, pressuring force, etc.) such that optimal inputs may be used, and where the “learning stage” as taught is akin to Applicants’ recitation of model training).
Kang is silent as to its welding being explicitly laser welding, e.g. as specifically further recited, but the Examiner believes one of ordinary skill in the art would understand Kang’s operations and modeling and considerations to be reasonably extensible to a laser welding operation/paradigm. However, to the extent that such a conjecture is not soundly supported, the Examiner refers to VERHAEGHE to establish the use of laser welding equipment to perform the same type of welding, e.g. spot welding, that Kang contemplates. Hence, in view of Verhaeghe, the Examiner believes it reasonable to infer the applicability of Kang’s framework to welding that encompasses laser welding as Applicants have recited.
Both Kang and Verhaeghe contemplate improvements to the efficiency and quality of similar welding processes as measurable by like/similar considerations. Hence, they are similarly directed and therefore analogous. It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to apply a laser welding approach as Verhaeghe specifically contemplates to perform the same type of welding operations that Kang seeks to optimize using its taught framework.
Applicants’ claim further clarifies the training to be training the data-based model as a function of an ascertained number of spatters, wherein the ascertained number of spatters depends on the operating parameters, which Kang etc. do not teach. At best, Kang establishes that spattering “is a problem caused by excessive heat” (e.g., per Kang’s page 3) and that such heat is “generated when the current flows” as part of the welding process (e.g., per Kang’s page 2). From this, the Examiner the reasons that Kang contemplates that spattering from an operational condition such as excessive heat which may be generated via the welding operation’s applied current might be problematic. To provide what Kang etc. otherwise lacks, the Examiner relies upon KONDOU, see e.g., Kondou’s framework for generating a learned model that can detect spatter as discussed in relation to its FIG. 1 and more specifically steps S112-S113 (which permit a characterization of the welding conditions such as voltage applied resulting in a determinable number of spatters). Kondou clarifies that the model is learned in accordance with supervised learning using images featuring a number of spatters, and hence from this is follows that spatters and their amount is a consideration of the learning/training process for the model as taught.
Both Kang and Verhaeghe contemplate improvements to the efficiency and quality of similar welding processes as measurable by like/similar considerations. Hence, they are similarly directed and therefore analogous. It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to incorporate a spatter-based model per Kondou to a framework such as Kang’s, with a reasonable expectation of success, such that Kang’s welding process can be subject to monitoring for spatter in a manner that usefully determines an underlying operational condition causing the issue/defect, thereby improving quality/performance in the welding operation.
Here, the Examiner notes the additional limitation of using Bayesian optimization in relation to the recited data-based model. The Examiner notes that the aforementioned references are silent as to this additional limitation, while also noting that it is common and prevalent to optimize models, neural networks, and the like such as those contemplated by Kang. To teach what Kang etc. lack, the Examiner then relies upon FRAZIER, see e.g. Frazier’s Abstract mentioning the use of Bayesian optimization of model objective functions to reduce evaluation time.
Kang relates to modeling to better welding operations, particularly in view of spatter. Frazier is far more generally directed to modelling apart from welding specifically, but contemplates improvements in modelling that can be leveraged to many types of substantive domains that can be modelled. Hence, the references to varying degrees are similarly directed and therefore analogous. It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to incorporate Frazier’s Bayesian optimization for use in a framework such as Kang’s modified framework, as discussed per claim 18 for example, with a reasonable expectation of success, thereby lessening time / cost for model development, validation, execution, and so forth.
Regarding claim 31, Kang in view of Verhaeghe and Kondou and further in view of Frazier teach the method as recited in claim 30, as discussed above. The aforementioned references further teach the additional limitation wherein following the setting of the operating parameters, the laser welding machine is operated using the operating parameters thus set (Kondou’s statement of a technical field makes clear that its taught machine learned model and related framework is effective to support the setting of welding conditions, see e.g., its discussion of step S113 where the result of the processing is used to effectively modify the voltage as applied). The motivation for combining the references is as discussed above in relation to claim 30.
Regarding claim 32, the claim includes the same or similar limitations as discussed above in relation to claim 30, and is therefore rejected under the same rationale. Here, with the present claim, there is a specific recitation of a test stand for a laser welding machine, which is further taught per Kondou’s discussion per FIGs. 1-2 teaching a practical setup for welding that the Examiner equates with the recitation of a test stand.
7. Claims 18 and 22-23 are rejected under 35 U.S.C. 103 as being unpatentable over Kang in view of Verhaeghe and Kondou and further in view of Non-Patent Literature “Studies on the spatter behavior when welding” (“Chang”) and Non-Patent Literature “Why is Gaussian the King of all distributions” (“Chugh”).
Regarding claim 18, KANG teaches A method for training a data-based model (page 2: “An artificial neural network (ANN) is a kind of function that connects inputs and outputs by connecting a virtual neuron device, a perceptron, which is a neuron of the human neuron, and a network. Perceptron's complex network makes it possible to reproduce or predict complex nonlinear phenomena relatively simply. Prediction methods using artificial neural networks are divided into learning stage and production stage. In the learning phase, a series of input / output relationships are supplied and thus a functional relationship is identified in the artificial neural network.” (where the learning stage as taught is akin to the recitation for training a data-based model, where the data used for the learning stage is further clarified on the reference’s page 3 (“spot welding data DB may be an existing spot welding data acquired and stored in advance by the experiment”))) to ascertain a variable which characterizes an energy input of a ... welding machine into a workpiece (page 3: “The input value acquiring step (S2) is a step of acquiring, as an input value, a part of the input power value data according to the time derived from the starting point to the end point of the spot welding ... the input value obtained is a value obtained by sampling the input power value from the start to the end of the welding in a predetermined time unit”, where the acquired input is processed by the neural network model to provide an output, e.g. page 4’s discussion of “a welding characteristic result output step (S3) for inputting the input value obtained in the input value obtaining step (S2) to the learned artificial neural network (ART) and outputting a welding characteristic result corresponding to the input value”), as a function of operating parameters of the ... welding machine (per page 4 of the reference, where the aforementioned output result is understood by the Examiner to be a characterization of the inputs as processed, where the welding characteristic result is understood by the Examiner to indicate a prediction corresponding to the welding processing conditions (e.g., current, voltage, time, pressuring force, etc.)), the method further comprising:
training the data-based model ... and wherein the data-based model is trained to output as a function of the operating parameters the ascertained variable characterizing the energy input as a model output variable (as discussed above, page 2’s teaching where “Prediction methods using artificial neural networks are divided into learning stage and production stage” (where the learning stage as taught is akin to the recitation for training a data-based model, where the data used for the learning stage is further clarified on the reference’s page 3 (“spot welding data DB may be an existing spot welding data acquired and stored in advance by the experiment”)), and further where the acquired input is processed by the neural network model to provide an output, e.g. page 4’s discussion of “a welding characteristic result output step (S3) for inputting the input value obtained in the input value obtaining step (S2) to the learned artificial neural network (ART) and outputting a welding characteristic result corresponding to the input value”, where the aforementioned output result is understood by the Examiner to be a characterization of the inputs as processed, where the welding characteristic result is understood by the Examiner to indicate a prediction corresponding to the welding processing conditions (e.g., current, voltage, time, pressuring force, etc.) ).
Kang is silent as to its welding being explicitly laser welding, e.g. as specifically further recited, but the Examiner believes one of ordinary skill in the art would understand Kang’s operations and modeling and considerations to be reasonably extensible to a laser welding operation/paradigm. However, to the extent that such a conjecture is not soundly supported, the Examiner refers to VERHAEGHE to establish the use of laser welding equipment to perform the same type of welding, e.g. spot welding, that Kang contemplates. Hence, in view of Verhaeghe, the Examiner believes it reasonable to infer the applicability of Kang’s framework to welding that encompasses laser welding as Applicants have recited.
Both Kang and Verhaeghe contemplate improvements to the efficiency and quality of similar welding processes as measurable by like/similar considerations. Hence, they are similarly directed and therefore analogous. It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to apply a laser welding approach as Verhaeghe specifically contemplates to perform the same type of welding operations that Kang seeks to optimize using its taught framework.
Applicants’ claim further clarifies the training to be training the data-based model as a function of an ascertained number of spatters, wherein the ascertained number of spatters depends on the operating parameters, which Kang etc. do not teach. At best, Kang establishes that spattering “is a problem caused by excessive heat” (e.g., per Kang’s page 3) and that such heat is “generated when the current flows” as part of the welding process (e.g., per Kang’s page 2). From this, the Examiner the reasons that Kang contemplates that spattering from an operational condition such as excessive heat which may be generated via the welding operation’s applied current might be problematic. To provide what Kang etc. otherwise lacks, the Examiner relies upon KONDOU, see e.g., Kondou’s framework for generating a learned model that can detect spatter as discussed in relation to its FIG. 1 and more specifically steps S112-S113 (which permit a characterization of the welding conditions such as voltage applied resulting in a determinable number of spatters). Kondou clarifies that the model is learned in accordance with supervised learning using images featuring a number of spatters, and hence from this is follows that spatters and their amount is a consideration of the learning/training process for the model as taught.
Both Kang and Verhaeghe contemplate improvements to the efficiency and quality of similar welding processes as measurable by like/similar considerations. Hence, they are similarly directed and therefore analogous. It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to incorporate a spatter-based model per Kondou to a framework such as Kang’s, with a reasonable expectation of success, such that Kang’s welding process can be subject to monitoring for spatter in a manner that usefully determines an underlying operational condition causing the issue/defect, thereby improving quality/performance in the welding operation.
Applicants claim has now been amended to include the limitations of cancelled claims 19-21. The Examiner will address these limitations here:
The aforementioned references Kang in view of Verhaeghe and Kondou do not teach the entirely of the further limitation for training of the data-based model is carried out as a function of the number of spatters as an experimentally ascertained measured variable, and the training also being carried out as a function of a simulatively ascertained variable characterizing the energy input as a simulatively ascertained simulation variable. Rather, the Examiner relies upon CHANG to teach what Kang etc. otherwise lack, see e.g. Chang’s comparable study to correlate operational parameters such as power with a welding result in terms of spatter (Abstract, Introduction: as found on page 1769). In particular, Chang’s sections 4.1 and 4.3 respectively detail the amassing of data via experimental and numerical modeling means, which are then subject to a comparison and correlation as discussed in section 5. Chang’s discussion in these aforementioned sections at least suggests there is a value of amassing data both ways and that the data as a whole has value in solving the sort of challenge/problem it contemplates, which is similar to what is contemplated by Kang etc.
Like Kang, Chang relates to modeling to better welding operations, particularly in view of spatter. Hence, the references are similarly directed and therefore analogous. It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to incorporate Chang’s more explicit and broader sense of data collection for purposes of providing more training data for use in a framework such as Kang’s modified framework, as discussed per claim 18 for example, with a reasonable expectation of success, thereby lessening the challenges that exist in the state of the art relating to the obtaining substantial training data for model/network training, validation, and so forth.
The aforementioned references Kang in view of Verhaeghe and Kondou and further in view of Chang do not teach the entirely of the further limitations for:
wherein during the training, the measured variable and/or the simulation variable are transformed using an affine transformation and wherein in the affine transformation, the measured variable and/or the simulation variable is multiplied by a factor, and the factor is selected as a function of a simulative model uncertainty and as a function of an experimental model uncertainty.
Rather, the Examiner relies upon CHUGH to teach what Kang etc. otherwise lack, see e.g., Chugh’s pages 3-4 discussing the benefit of using Gaussian distributions for values in ML data, which can entail an affine transformation as discussed on page 6, and further per page 6’s discussion of affine transformation, use of multiplication with a scalar is taught.
Like Kang, Chugh relates to modeling to better welding operations, particularly in view of spatter. Chugh is far more generally directed to modelling apart from welding specifically, but contemplates improvements in modelling that can be leveraged to many types of substantive domains that can be modelled. Hence, the references to varying degrees are similarly directed and therefore analogous. It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to incorporate Chugh’s transformation and Gaussian form in a framework such as Kang’s combined framework, with a reasonable expectation of success, thereby making a safer assumption about the distribution of the real world data that is involved in the welding operations as used for training.
Regarding claim 22, Kang in view of Verhaeghe and Kondou and then further in view of Chang and Chugh teach the method as recited in claim 18, as discussed above. The aforementioned references further teach the additional limitation wherein the factor is selected as a function of a quotient of the simulative model uncertainty and the experimental model uncertainty (Chugh’s page 6 discussing affine transformations). The motivation for combining the references is as discussed above in relation to claim 18.
Regarding claim 23, Kang in view of Verhaeghe and Kondou and then further in view of Chang and Chugh teach the method as recited in claim 18, as discussed above. The aforementioned references further teach the additional limitation wherein the data-based model includes a simulatively trained first partial model which is a Gaussian process model, and an experimentally trained second partial model which is a Gaussian process model, the simulative model uncertainty being ascertained using the first partial model, and the experimental model uncertainty being ascertained using the second partial model (Chang’s sections 4.1 and 4.3 respectively detail the amassing of data via experimental and numerical modeling means, which are then subject to a comparison and correlation as discussed in section 5, such that Chang’s discussion in these aforementioned sections at least suggests there is a value of amassing data both ways and that the data as a whole has value in solving the sort of challenge/problem it contemplates, which can be subject to modeling the two types of data separately to engage with the comparison and correlation discussed per Chang’s section 5). The motivation for combining the references is as discussed above in relation to claim 18.
8. Claims 24, 27, and 29 are rejected under 35 U.S.C. 103 as being unpatentable over Kang in view of Verhaeghe and Kondou and then Chang and Chugh and further in view of U.S. Patent Application Publication No. 2021/0241037 (“Lisowska”).
Regarding claim 24, Kang in view of Verhaeghe and Kondou and then Chang and Chugh teach the method as recited in claim 23, as discussed above. The aforementioned references do not teach the further limitation wherein the data-based model includes an experimentally trained third partial model which is a Gaussian process model, and which is trained to output a difference between the experimentally ascertained measured variable and an output variable of the first partial model. Rather, the Examiner relies upon LISOWSKA to teach what they otherwise lack, see e.g., [0051] discussing modelling that finds the difference between predictions from other models in a model chain, which helps to determine uncertainty values that otherwise help model performance.
Kang relates to modeling to better welding operations, particularly in view of spatter. Chang is similarly directed. Lisowska is more generally directed to modelling and its development apart from welding specifically, but contemplates improvements in modelling that can be leveraged to many types of substantive domains that can be modelled. Hence, the references to varying degrees are similarly directed and therefore analogous. It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to incorporate Lisowska’s difference prediction for use in a framework such as Kang’s modified framework as modified in view of Chang, with a reasonable expectation of success, thereby improving the predictive aspect of the model as it is being developed.
Regarding claim 27, Kang in view of Verhaeghe and Kondou and then Chang and Chugh and further in view of Lisowska teach the method as recited in claim 24, as discussed above. The aforementioned references teach the additional limitation wherein when ascertaining the transformed measured variable, the measured variable is transformed using the affine transformation, and the difference is multiplied by the factor (Chugh’s page 6 discussing affine transformations, including multiplication with a scalar, where the scalar could be the differential determined by Lisowska’s model as discussed above per claim 24 for purposes of transforming the measurement in a way that also incorporates an additional dimension of certainty/confidence). The motivation for combining the references is as discussed above in relation to claim 24.
Regarding claim 29, Kang in view of Verhaeghe and Kondou and then Chang and Chugh and further in view of Lisowska teach the method as recited in claim 24, as discussed above. The aforementioned references teach the additional limitation wherein to ascertain an uncertainty of the model output variable of the data-based model, the uncertainty is ascertained using the second partial model (Lisowska’s [0051] discussing modelling that finds the difference between predictions from other models in a model chain, which helps to determine uncertainty values that otherwise help model performance). The motivation for combining the references is as discussed above in relation to claim 24.
9. Claims 25-26 and 28 are rejected under 35 U.S.C. 103 as being unpatentable over Kang in view of Verhaeghe and Kondou and then Chang and Chugh and Lisowska and further in view of Non-Patent Literature “Honey I shrunk the target variable” (“Wilhelm”).
Regarding claim 25, Kang in view of Verhaeghe and Kondou and then Chang and Chugh and Lisowska teach the method as recited in claim 24, as discussed above. The aforementioned references do not teach the further limitation wherein the second partial model is not trained with the transformed measured variable, but is trained using the measured variable. Rather, the Examiner relies upon WILHELM to teach what they otherwise lack, see e.g. pages 1-2 for a general discussion of deciding whether or not to provide a transformation to data subject to modelling, such that there is a choice by the operative administrator of the data based on whether it improves the model.
Like Lisowska, and the other references more generally, Wilhelm relates to transformations of data that is subject to modeling. Hence, they are similarly directed and therefore analogous. It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to apply Wilhelm’s discretion to choose either to transform or not a particular data in a model, such as one contemplated by Kang’s modified framework, based on whether this would improve the model’s performance.
Regarding claim 26, Kang in view of Verhaeghe and Kondou and then Chang and Chugh and Lisowska and further in view of Wilhelm teach the method as recited in claim 25, as discussed above. The aforementioned references wherein the third partial model is trained using the transformed measured variable (Wilhelm’s pages 1-2 for a general discussion of deciding whether or not to provide a transformation to data subject to modelling, such that there is a choice by the operative administrator of the data based on whether it improves the model). The motivation for combining the references is as discussed above in relation to claim 25.
Regarding claim 28, Kang in view of Verhaeghe and Kondou and then Chang and Chugh and Lisowska teach the method as recited in claim 24, as discussed above. The aforementioned references
wherein to ascertain the model output variable of the data-based model, an output variable of the first partial model and an output variable of the third partial model are added (Chang’s sections 4.1 and 4.3 respectively detail the amassing of data via experimental and numerical modeling means, which are then subject to a comparison and correlation as discussed in section 5, such that Chang’s discussion in these aforementioned sections at least suggests there is a value of amassing data both ways and that the data as a whole has value in solving the sort of challenge/problem it contemplates, which can be subject to modeling the two types of data separately to engage with the comparison and correlation discussed per Chang’s section 5, and it reasons that the outputs can be subject to a combination/integration step if they are deemed to be correlated) and transformed using an inverse of the affine transformation (Wilhelm’s page 2 discussing, in the last paragraph under the heading Choosing the right error measure, that models and predictions are transformed back for evaluation).
Like Lisowska, and the other references more generally, Wilhelm relates to transformations of data that is subject to modeling. Hence, they are similarly directed and therefore analogous. It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to apply Wilhelm’s discretion to choose either to transform or not a particular data in a model, such as one contemplated by Kang’s modified framework, based on whether this would improve the model’s performance.
Response to Arguments
10. Applicants’ arguments filed 3/17/26 have been fully considered but are respectfully not persuasive.
On page 7 of those Applicants’ arguments, Applicants essentially argue that the Examiner has not sufficiently/successfully mapped to the recited term “energy input” and then cites to a portion of the Examiner’s mapping that does not even include the first recitation for that term.
If Applicants revisit the rejection as previously made and as presently maintained, they will see the mapping to that recited term in the portion of the claim where that term is first used- which is not the training limitation but rather the claim preamble. In the Examiner’s mapping addressing the claim’s preamble, the Examiner has provided a mapping to the recited term “energy input” that equates it to Kang’s teaching of an “input power value data” that is acquired per Kang’s acquiring step S2 (see page 3) and fed as an input into the taught learned artificial neural network (page 4) to obtain that model’s output of a welding characteristic result.
Applicants’ arguments as reviewed by the Examiner make no accounting for this mapping at all, and rather focus on other portions of the reference. Hence, the arguments are not properly responsive to the rejection and its mapping as made of record and as presently maintained, and hence cannot be considered persuasive. If Applicants would address the rejection as it was made, including the proper mapping, then the Examiner will appropriately reconsider the rejection. But as it stands, Applicants have not rebutted the rejection as it was made.
In making these same arguments, Applicants also take the position that Kang does not even teach a trained model that outputs a characterization of an energy input. See the paragraph beginning at the bottom of page 7 of the Reply. The Examiner disagrees. Kang’s model is obviously trained. A “learning phase” for the model is clearly taught and the Examiner has mapped to it in making the rejection previously and in maintaining it. Moreover, the output is explicitly “a welding characteristic result.” The Examiner finds these characterizations of the cited reference as made by Applicants’ Representative to be incredibly inaccurate.
Applicants further argue that the Kondou reference does not sufficiently teach features that read on the limitation for “training the data-based model as a function of an ascertained number of spatters.” See Applicants’ Reply beginning at the bottom of page 8 and continuing onto page 9. The Examiner disagrees for at least two reasons.
As a first matter, Applicants admit that Kondou teaches “a learned model generated by supervised learning using a plurality of images including the sputter and a plurality of images not including the spatter as teacher data.” Hence, on its face, Applicants appear to acknowledge a model that is trained based on images that include and don’t include spatter.
Applicants appear to emphasize that the training is not performed based on a number of spatters. However, as a second matter, the Examiner reasons that a model that is trained on ascertaining more than one spatter would be understood to be trained based on an ascertained number of spatters as recited, which the Examiner believes Kondou, even when subjected to Applicants’ characterization, teaches, mostly because “a plurality of images including the sputter” as ascertained by the model in its training is equivalent to training as a function of an ascertained number of spatters.
Conclusion
11. THIS ACTION IS MADE FINAL. Applicant is reminded of the extension of time policy as set forth in 37 CFR 1.136(a).
A shortened statutory period for reply to this final action is set to expire THREE MONTHS from the mailing date of this action. In the event a first reply is filed within TWO MONTHS of the mailing date of this final action and the advisory action is not mailed until after the end of the THREE-MONTH shortened statutory period, then the shortened statutory period will expire on the date the advisory action is mailed, and any nonprovisional extension fee (37 CFR 1.17(a)) pursuant to 37 CFR 1.136(a) will be calculated from the mailing date of the advisory action. In no event, however, will the statutory period for reply expire later than SIX MONTHS from the mailing date of this final action.
12. Any inquiry concerning this communication or earlier communications from the examiner should be directed to SHOURJO DASGUPTA whose telephone number is (571)272-7207. The examiner can normally be reached M-F 8am-5pm CST.
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, Tamara Kyle can be reached at 571 272 4241. 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.
/SHOURJO DASGUPTA/Primary Examiner, Art Unit 2144
Prosecution Insights