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 .
Claim Rejections - 35 USC § 112
The following is a quotation of the first paragraph of 35 U.S.C. 112(a):
(a) IN GENERAL.—The specification shall contain a written description of the invention, and of the manner and process of making and using it, in such full, clear, concise, and exact terms as to enable any person skilled in the art to which it pertains, or with which it is most nearly connected, to make and use the same, and shall set forth the best mode contemplated by the inventor or joint inventor of carrying out the invention.
The following is a quotation of the first paragraph of pre-AIA 35 U.S.C. 112:
The specification shall contain a written description of the invention, and of the manner and process of making and using it, in such full, clear, concise, and exact terms as to enable any person skilled in the art to which it pertains, or with which it is most nearly connected, to make and use the same, and shall set forth the best mode contemplated by the inventor of carrying out his invention.
Claim 15 rejected under 35 U.S.C. 112(a) or 35 U.S.C. 112 (pre-AIA ), first paragraph, as failing to comply with the enablement requirement. The claim(s) contains subject matter which was not described in the specification in such a way as to enable one skilled in the art to which it pertains, or with which it is most nearly connected, to make and/or use the invention. Claim 15 states that the time-series data image is a dimensionally reduced scatter diagram, while the specification indicates that the dimensionally reduced scatter diagram is generated from the reduced features of the time-series data images (see [0051]). In other words, the dimensionally reduced scatter diagram is not the claimed time-series data image itself, but is instead generated from the time-series data image. For examination purposes, claim 15 is being interpreted as; “wherein the time-series data image is used to generate a dimensionally reduced scatter diagram.”
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.
Claim(s) 1-5, 8, 10-11 is/are rejected under 35 U.S.C. 103 as being unpatentable over Hirunori Murase et al. (hereinafter Murase) (JP2020144619A, 2020-09-10) in view of Kazuya Mori et al. (Hereinafter Mori) (JP2018051783A, 2018-04-05) further in view of Noriyuki Aoki et al. (hereinafter Aoki) (JP2018092453A, 2018-06-14)
Regarding claim 1, Murase teaches;
A non-transitory computer readable recording medium storing a computer program causing a computer to execute processing of detecting an abnormality of a machine having a movable part, wherein the computer executes the processing of:
PNG
media_image1.png
764
378
media_image1.png
Greyscale
NOTE: Murase teaches a device storing instructions to perform the methods presented in their disclosure.
acquiring physical quantity data on a time-series basis output from a sensor detecting a physical quantity related to a motion of the movable part;
([Abstract] sensor data obtaining means 11 for obtaining the signals from sensors 2 in time series as sensor data;)
NOTE: Teaches acquiring data on a time-series basis output from a sensor.
([pg.2] FIG. 1 shows the configuration of the abnormality detection device 1. The type of the sensor 2 is not limited, and examples thereof include a temperature sensor, a pressure sensor, a speed sensor, and a flow velocity sensor. Further, the number, combination, position, etc. of the sensors 2 are not particularly limited.)
NOTE: Teaches a sensor that detects a physical quantity related to a motion (speed) of a part of the machine. If you are measuring the speed of a part of a machine, then that part is a movable part.
converting the physical quantity data on a time-series basis acquired to a time-series data image representing the physical quantity data;
([Abstract] converting means 13 for converting the sensor data into a five-dimensional array combined so as to be a dimensionless number; image creating means 15 for creating a color image on the basis of the five-dimensional array;)
NOTE: Teaches converting the acquired physical quantity data on a time-series basis (the aforementioned sensor data) to a time-series data image representing the physical quantity data (the sensor data representing the time-series physical quantity data is converted into a five-dimensional array, which is then converted to a color image, which is therefore a time-series data image representing the physical quantity data).
inputting the time-series data image converted to a learning model to calculate a feature of the time-series data image,
([Abstract] feature-quantity extracting means 17 for inputting the color image into a convolution neural network, and which extracts the outputs by all coupled layers as a feature quantity;)
NOTE: Teaches inputting the converted time-series data image to a learning model (inputting the color image into a convolutional neural network) to calculate a feature of the time-series data image (extracts the outputs as a feature quantity) the learned model being trained with a feature of the time-series data related to the movable part in a normal condition.
the learning model being trained with a feature of the time-series data image related to the movable part in a normal condition;
[pg.2] In general, the parameters of each CNN layer trained using a large-scale supervised image data set are highly versatile, and the target image is obtained by applying fine tuning that tunes the parameters using the training data of the target task. The accuracy of recognition can be improved.
NOTE: Teaches the learning model being trained with a feature of the time-series data image related to the movable part in the normal condition (The CNN is trained using supervised image data, where the training data is related to the target task. The target task of the disclosure, as taught above, is to monitor sensor values of machines, which can include motion data. The time-series data images used to train the learning model are therefore related to the movable part in a normal condition)
and determining a presence or an absence of an abnormality of the industrial machine based on the feature calculated.
([Abstract] and exclusive identifying means 19 for performing an exclusive identification with the extracted feature quantity being input thereto to detect an abnormality.)
NOTE: Teaches determining a presence or absence of an abnormality of the machine (detect an abnormality) based on the feature calculated (with the extracted feature quantity)
Murase fails to teach but Mori teaches;
wherein the movable part comprises two or more of a rotational shaft of a molding machine, a first screw of a twin-screw kneading extruder, and a second screw of a twin-screw kneading extruder,
([pg. 7] the multi-screw kneading extruder provided with the first screw 14 and the second screw 16 and the vibration measuring means 60)
and wherein the physical quantity data includes time-series data ([pg. 11] For each of the directions, the vibration measurement sampling period was set to 1 millisecond, and the measurement for 1 second (1000 samplings) was performed once, and the vibration measurement was repeated 6000 times.) indicating two or more physical quantities selected from: a displacement of the rotational shaft, a torque of the rotational shaft, a rotational speed of the rotational shaft, a rotational acceleration of the rotational shaft, a displacement of that intersects a rotation center axis of the first screw (displacement/vibration in the y/z direction for the first screw 14 [see below]), a torque of the first screw, a rotational speed of the first screw, a rotational acceleration of the first screw, a displacement of that intersects a rotation center axis of the second screw (displacement/vibration in the y/z direction for the second screw 16 [see below]), a torque of the second screw, a rotational speed of the second screw, and a rotational acceleration of the second screw;
a first physical quantity selected from the two or more physical quantities in a first-axis direction (vibration/displacement of the first screw in the y-axis direction [see below]) and a second physical quantity selected from the two or more physical quantities in a second-axis direction (vibration/displacement of the first screw in the z-axis direction [see below]);
([pg.7] Therefore, the multi-screw kneading extruder provided with the first screw 14 and the second screw 16 and the vibration measuring means 60)
PNG
media_image2.png
393
866
media_image2.png
Greyscale
([pg.3] The vibration measuring means 60 measures at least vibrations in the longitudinal direction and the radial direction of the rotating shaft of the screw 12 [note: screw 12 represents both the first screw 14, and the second screw 16, as pictured above] … In the present specification, the “longitudinal direction” means a direction along the rotation axis of the screw 12 unless otherwise specified. The x-axis direction shown in FIGS. 1 to 4 is the longitudinal direction, and in this specification, the “x-axis direction” means the longitudinal direction. In the present specification, the “radial direction” means a direction orthogonal to the rotation axis (that is, the longitudinal direction) of the screw 12 unless otherwise specified. The y-axis direction and the z-axis direction shown in FIGS. 1 and 2 are radial directions.)
OBVIOUSNESS TO COMBINE MORI WITH MURASE:
Mori is analogous art to the present invention as it pertains to collecting measurements of the screws in a kneading extruder to determine if an abnormality exists.
Murase teaches an abnormality detection pipeline utilizing time-series sensor data, while Mori teaches acquiring time-series sensor data including displacement data (vibration data) of screws of a twin-screw kneading extruder in directions that intersect rotation center axes of the screws.
Mori further indicates that abnormalities in the vibrations can lead to damage of the extruder;
([pg. 5-6] excessive vibrations may occur and the device may break down.)
Therefore, it would have been obvious to one of ordinary skill in the art, before the effective filing date of the claimed invention, to perform the data processing limitations of Murase using the time-series displacement data of the screws of the twin-screw extruder of Mori to detect abnormal vibrations in the twin-kneading extruder before damage is done to the extruder, allowing preventative action to be taken.
Murase and Mori fail to teach but Aoki teaches
… wherein the time-series data image includes, in a single image, a first image rendering [a first and second physical quantity from sensors collected on a time-series basis]
([pg. 5] the sensor data may be expressed as two-dimensional image data. (sensor data table 102) in which a plurality of sensor data used by the learning apparatus 1 is collected. The sensor data table 102 of FIG. 8 is a collection of sensor data acquired by a plurality of sensors numbered 1 to 12 in time series… image data 202 obtained by converting and integrating the sensor data entered in the sensor data table 102 of FIG. 8 into luminance information.)
[fig. 8, 9]
PNG
media_image3.png
441
885
media_image3.png
Greyscale
PNG
media_image4.png
676
602
media_image4.png
Greyscale
OBVIOUSNESS TO COMBINE AOKI WITH MORI AND MURASE:
Aoki is analogous art to the present disclosure as it pertains to generating data images for time-series sensor data.
Mori already teaches acquiring sensor data of at least a first physical quantity selected from the two or more physical quantities in a first-axis direction (vibration/displacement of the first screw in the y-axis direction) and a second physical quantity selected from the two or more physical quantities in a second-axis direction (vibration/displacement of the first screw in the z-axis direction). Additionally, Mori indicates that there can be an individual sensor measuring each axis direction of displacement/vibration; ([Mori, pg. 4] The vibration measuring means 60 may be, a combination of single-axis vibrometers that measure vibrations in one direction).
Aoki teaches a single time-series data image 202 where each column is a rendering of a single sensor’s data over time intervals, and each row is a plurality of sensors data at a single time interval.
[fig. 9]
PNG
media_image4.png
676
602
media_image4.png
Greyscale
In combination with Mori, the first column of Aoki’s data image can represent the data from the sensor that measures displacement of the first screw in a first direction (y-axis direction), and the second column can represent the data from the sensor that measures displacement of the first screw in a second direction (z-axis direction).
In this interpretation, the first two columns of the time-series data image 202 would be a first image rendering a first physical quantity selected from the two or more physical quantities (displacement of the first screw) in a first-axis direction (y-axis direction) and a second physical quantity selected from the two or more physical quantities (displacement of the first screw) in a second-axis direction (z-axis direction).
Aoki further indicates that the benefit of generating their time-series data images is that it allows a plurality of sensor data to be learned by machine learning models at the same time;
([pg. 6] when numerical data output from a plurality of sensors is learned by machine learning, the numerical data is integrated and imaged, so that a plurality of sensors can be obtained at a time. … a plurality of numerical data can be converted into a format that can be easily analyzed together, and the plurality of numerical data can be learned at once by machine learning.)
Therefore, it would have been obvious to one of ordinary skill in the art, before the effective filing date of the claimed invention, to use the data imaging method of Aoki to generate a time-series data image of Mori’s physical quantity data to allow for the simultaneous analysis of a plurality of sensors by the machine learning model of Murase.
Regarding claim 2, Murase teaches;
The non-transitory computer readable recording medium according to claim 1, wherein the learning model is a model generated by machine learning.
([Abstract] feature-quantity extracting means 17 for inputting the color image into a convolution neural network, and which extracts the outputs by all coupled layers as a feature quantity;)
NOTE: The learning model is a CNN, which is an ML model and is therefore generated by ML.
Regarding claim 3, Murase teaches;
The non-transitory computer readable recording medium according to claim 1, wherein the learning model is a one-class classification model generated by machine learning.
([pg.8] The feature amount of each image is extracted by CNN, and the feature amount is input to the 1-class SVM.)
NOTE: The learning model is a one class classification model generated by machine learning (includes a one-class SVM).
Regarding claim 4, Murase fails to teach but Mori teaches;
a third physical quantity selected from the two or more physical quantities in a first-axis direction (vibration/displacement of the second screw in the y-axis direction [see below]) and a fourth physical quantity selected from the two or more physical quantities in a second-axis direction (vibration/displacement of the first screw in the z-axis direction [see below]).
([pg.7] Therefore, the multi-screw kneading extruder provided with the first screw 14 and the second screw 16 and the vibration measuring means 60)
PNG
media_image2.png
393
866
media_image2.png
Greyscale
([pg.3] The vibration measuring means 60 measures at least vibrations in the longitudinal direction and the radial direction of the rotating shaft of the screw 12 [note: screw 12 represents both the first screw 14, and the second screw 16, as pictured above] … In the present specification, the “longitudinal direction” means a direction along the rotation axis of the screw 12 unless otherwise specified. The x-axis direction shown in FIGS. 1 to 4 is the longitudinal direction, and in this specification, the “x-axis direction” means the longitudinal direction. In the present specification, the “radial direction” means a direction orthogonal to the rotation axis (that is, the longitudinal direction) of the screw 12 unless otherwise specified. The y-axis direction and the z-axis direction shown in FIGS. 1 and 2 are radial directions.)
OBVIOUSNESS:
Using the same reasoning from claim 1, it would have been obvious to one of ordinary skill in the art, before the effective filing date of the claimed invention, to perform the data processing limitations of Murase using the time-series displacement data of the screws of the twin-screw extruder of Mori to detect abnormal vibrations in the twin-kneading extruder before damage is done to the extruder, allowing preventative action to be taken.
Murase and Mori fail to teach buy Aoki teaches;
The non-transitory computer readable recording medium according to claim 1, wherein the time-series data image includes a second image rendering [a third and fourth physical quantity from sensors collected on a time-series basis]
([pg. 5] the sensor data may be expressed as two-dimensional image data. (sensor data table 102) in which a plurality of sensor data used by the learning apparatus 1 is collected. The sensor data table 102 of FIG. 8 is a collection of sensor data acquired by a plurality of sensors numbered 1 to 12 in time series… image data 202 obtained by converting and integrating the sensor data entered in the sensor data table 102 of FIG. 8 into luminance information.)
[fig. 8, 9]
PNG
media_image3.png
441
885
media_image3.png
Greyscale
PNG
media_image4.png
676
602
media_image4.png
Greyscale
OBVIOUSNESS:
Using the same reasoning from claim 1, in the combination of Aoki and Mori, the second two columns of the time-series data image 202 of Aoki would be a second image rendering a third physical quantity selected from the two or more physical quantities (displacement of the second screw) in a first-axis direction (y-axis direction) and a fourth physical quantity selected from the two or more physical quantities (displacement of the second screw) in a second-axis direction (z-axis direction).
And also using the reasoning from claim 1, it would have been obvious to one of ordinary skill in the art, before the effective filing date of the claimed invention, to use the data imaging method of Aoki to generate a time-series data image of Mori’s physical quantity data to allow for the simultaneous analysis of a plurality of sensors by the machine learning model of Murase.
Regarding claim 5, Murase fails to teach but Mori teaches;
The non-transitory computer readable recording medium according to claim 4, wherein each physical quantity is indicated by a numerical value.
([pg. 12] As shown in FIG. 8 (below), the vibration widths of the vibration acceleration data Ax, Ay, and Az in the x direction, the y direction, and the z direction)
[AltContent: textbox (y)][AltContent: textbox (z)][AltContent: textbox (x)]
PNG
media_image5.png
409
1012
media_image5.png
Greyscale
OBVIOUSNESS:
Using the same reasoning from claim 1, it would have been obvious to one of ordinary skill in the art, before the effective filing date of the claimed invention, to perform the data processing limitations of Murase using the time-series displacement data of the screws of the twin-screw extruder of Mori to detect abnormal vibrations in the twin-kneading extruder before damage is done to the extruder.
Regarding claim 8, Murase fails to teach but Mori teaches;
the physical quantity data includes time-series data ([pg. 11] For each of the directions, the vibration measurement sampling period was set to 1 millisecond, and the measurement for 1 second (1000 samplings) was performed once, and the vibration measurement was repeated 6000 times.) indicating displacements (vibrations) in a first-axis direction (y) and a second-axis direction (z) that intersect a rotation center axis (x) of the first screw (14) and time- series data indicating displacements in a third-axis direction (y) and a fourth axis direction (z) that intersect a rotation center axis (x) of the second screw (16) … displacements (vibrations) of the first screw (14) regarding the first-axis direction (y) and the second-axis direction (z) that intersect each other as coordinate axes, … displacements of the second screw (16) regarding the third-axis direction (y) and the fourth-axis direction (z) that intersect each other as coordinate axes.
([pg.7] Therefore, the multi-screw kneading extruder provided with the first screw 14 and the second screw 16 and the vibration measuring means 60)
[AltContent: textbox ([Fig. 2])]
PNG
media_image2.png
393
866
media_image2.png
Greyscale
([pg.3] The vibration measuring means 60 measures at least vibrations in the longitudinal direction and the radial direction of the rotating shaft of the screw 12 [note: screw 12 represents both the first screw 14, and the second screw 16, as pictured above] … In the present specification, the “longitudinal direction” means a direction along the rotation axis of the screw 12 unless otherwise specified. The x-axis direction shown in FIGS. 1 to 4 is the longitudinal direction, and in this specification, the “x-axis direction” means the longitudinal direction. In the present specification, the “radial direction” means a direction orthogonal to the rotation axis (that is, the longitudinal direction) of the screw 12 unless otherwise specified. The y-axis direction and the z-axis direction shown in FIGS. 1 and 2 are radial directions.)
OBVIOUSNESS:
Using the same reasoning from claim 1, it would have been obvious to one of ordinary skill in the art, before the effective filing date of the claimed invention, to perform the data processing limitations of Murase using the time-series displacement data of the screws of the twin-screw extruder of Mori to detect abnormal vibrations in the twin-kneading extruder before damage is done to the extruder.
Murase and Mori fail to teach but Aoki teaches;
and the time-series data image (image data 202 below) includes in the first image rendering [a first and second physical quantity from sensors collected on a time-series basis] (the first two columns of 202 can be considered a first image rendering two physical quantities) … and a second image rendering [a third and fourth physical quantity from sensors collected on a time-series basis] (the second two columns of 202 can be considered a second image rendering two physical quantities)
([pg. 5] the sensor data may be expressed as two-dimensional image data. (sensor data table 102) in which a plurality of sensor data used by the learning apparatus 1 is collected. The sensor data table 102 of FIG. 8 is a collection of sensor data acquired by a plurality of sensors numbered 1 to 12 in time series… image data 202 obtained by converting and integrating the sensor data entered in the sensor data table 102 of FIG. 8 into luminance information.)
[fig. 8, 9]
PNG
media_image3.png
441
885
media_image3.png
Greyscale
PNG
media_image4.png
676
602
media_image4.png
Greyscale
OBVIOUSNESS:
Using the same reasoning from claim 1, in the combination of Aoki and Mori, the first two columns of the time-series data image 202 of Aoki can be a first image, where the first column can render displacements of the sensor data of the first screw regarding the first-axis direction (y) and the second column can render displacements of the sensor data of the first screw regarding the second-axis direction (z) that intersect each other as coordinate axes. Similarly, in this interpretation, the second two columns of the time-series data image 202 of Aoki can be a second image, where the first column of the second image can render displacements of the sensor data of the second screw regarding the third-axis direction (y) and the second column of the second image can render displacements of the sensor data of the second screw regarding the fourth-axis direction (z) that intersect each other as coordinate axes.
And also using the reasoning from claim 1, it would have been obvious to one of ordinary skill in the art, before the effective filing date of the claimed invention, to use the data imaging method of Aoki to generate a time-series data image of Mori’s physical quantity data to allow for the simultaneous analysis of a plurality of sensors by the machine learning model of Murase.
Regarding claim 10,
Claim 10 is a method claim that is substantially similar to claim 1 and is rejected using the same reasoning.
Regarding claim 11,
Claim 11 is an apparatus claim that is substantially similar to claim 1 and is rejected using the same reasoning.
Claim(s) 6-7, 12 is/are rejected under 35 U.S.C. 103 as being unpatentable over Murase (JP2020144619A, 2020-09-10) in view of Mori (JP2018051783A, 2018-04-05) further in view of Aoki (JP2018092453A, 2018-06-14) as applied to claim 1 above, further in view of Miyazawa Masanori et al. (hereinafter Miyazawa) (JP 2000225641 A, 2000-08-15)
Regarding claim 6, Murase, Mori, and Aoki fail to teach but Miyazawa teaches:
the physical quantity data includes time-series data ([pg. 6] the operation of detecting the difference in rotation speed is continued (i.e., speed detected over time) … [pg. 4] determines a difference between the respective rotation speeds detected by the screw shaft rotation speed sensor) indicating a displacement of the rotational shaft ([pg. 8] screw shaft rotation speed) of the molding machine ([pg. 10] The extruder 101 ′ … extrudes a molded product) a torque of the rotational shaft, a rotational speed of the rotational shaft ([pg. 8] screw shaft rotation speed sensor 112), or a rotational acceleration of the rotational shaft.
OBVIOUSNESS TO COMBINE MIYAZAWA WITH MURASE, MORI, AOKI:
Miyazawa is analogous art to the present disclosure as it pertains to a twin screw extruder (as pictured in fig. 5, extruder 101 has 2 screws, 30 and 30’) capable of kneading ([pg. 3] apparatus capable of … kneading … with two screws) which is being utilized in a molding machine ([pg. 10] The extruder 101 … extrudes a molded product).
Murase provides the base abnormality detection pipeline using time-series data images, Mori teaches recording displacements in multiple axis directions for screws in a twin-kneading extruder using sensor data over time, Aoki teaches rendering data images for sensor data over time, and Miyazawa teaches measuring at least the rotational speed of a rotational shaft a twin-screw kneading extruder of a molding machine.
Additionally, Miyazawa indicates that monitoring rotational speeds of components in the molding machine and detecting abnormalities in associated data protects the molding machine from getting damaged.
([pg. 8] the torsional displacement of the screw shaft 32 may be detected by a combination of the above-described screw shaft rotation speed sensor 112 and screw rotation speed sensor 113… [pg. 4] detecting the torsional displacement of the screw shaft, an overload is detected by detecting a difference between the rotation speed of the screw itself and the rotation speed of the transmission gear driving the screw. Since the overload is detected at the portion where the occurrence occurs, there is no delay in detection, prompt response is possible, and the device can be effectively protected)
Therefore, it would have been obvious to one of ordinary skill in the art, before the effective filing date of the claimed invention, for the twin kneading extruder of Mori to be included as the twin kneading extruder of the molding machine of Miyazawa, and to also measure rotational speeds of components of the molding machine to additionally detect abnormalities (using the abnormality detection pipeline of Murase and data image generation of Aoki) in rotation data of the system, allowing action to be taken before the molding machine is damaged.
Regarding claim 7, Murase, Mori, and Aoki fail to teach but Miyazawa teaches:
the physical quantity data includes time-series ([pg. 6] the operation of detecting the difference in rotation speed is continued (i.e., speed detected over time) … [pg. 4] detecting a difference between the rotation speed of the screw itself) data indicating a screw displacement ([pg. 8] screw rotation speed sensor 113) of the molding machine ([pg. 10] The extruder 101 … extrudes a molded product), a torque of the first screw or the second screw, a rotational speed of the first screw or the second screw ([pg. 8] screw rotation speed sensor 113), or a rotational acceleration of the first screw or the second screw.
OBVIOUSNESS:
Using the same reasoning from claim 6, it would have been obvious to one of ordinary skill in the art, before the effective filing date of the claimed invention, for the twin kneading extruder of Mori to be included as the twin kneading extruder of the molding machine of Miyazawa, and to also measure rotational speeds of components of the molding machine to additionally detect abnormalities (using the abnormality detection pipeline of Murase and data image generation of Aoki) in rotation data of the system, allowing action to be taken before the molding machine is damaged.
Regarding claim 12, Murase, Mori, and Aoki teach;
the abnormality detection apparatus according to claim 11;
using the same reasoning from claim 1, which is substantially similar to claim 11
Murase, Mori, and Aoki fail to teach but Miyazawa teaches;
A molding machine system ([pg. 10] The extruder 101 … extrudes a molded product), comprising: … and a molding machine, wherein the abnormality detection apparatus is adapted to detect an abnormality ([pg. 4] an overload is detected) of the molding machine ([pg. 10] The extruder 101 … extrudes a molded product).
OBVIOUSNESS:
Using the same reasoning from claim 6, it would have been obvious to one of ordinary skill in the art, before the effective filing date of the claimed invention, for the twin kneading extruder of Mori to be included as the twin kneading extruder of the molding machine of Miyazawa, and to also measure rotational speeds of components of the molding machine to additionally detect abnormalities (using the abnormality detection pipeline of Murase and data image generation of Aoki) in rotation data of the system, allowing action to be taken before the molding machine is damaged.
Claim(s) 9 is/are rejected under 35 U.S.C. 103 as being unpatentable over Murase (JP2020144619A, 2020-09-10) in view of Mori (JP2018051783A, 2018-04-05) further in view of Aoki (JP2018092453A, 2018-06-14) as applied to claim 1 above, further in view of Anthony Thyssen (hereinafter Thyssen) (“ImageMagick Examples -- Montage, Arrays of Images”, 2009-12-27).
Regarding claim 9, Murase teaches;
time series data image
using the same reasoning from claim 1
Murase, Mori, and Aoki fail to teach but Thyssen teaches;
wherein the
[pg. 4]
PNG
media_image6.png
129
536
media_image6.png
Greyscale
PNG
media_image6.png
129
536
media_image6.png
Greyscale
NOTE: Thyssen discloses that if you specify ‘-tile 2x2’ but only use 2 images, you will get a substantially square image including the 2 images, where the part of the image other than the 2 input images is a blank image. This therefore teaches a substantially square image including a first image and a second image as well as a blank image that fills a part other than the first image and the second image.
OBVIOUSNESS TO COMBINE THYSSEN:
Thyssen is analogous art to the present disclosure as it solves the same problem of combining two images in a single image and making the single image square by filling the remaining space with a blank image.
Murase teaches using time-series data images as input to a machine learning model, Mori teaches time-series sensor data, Aoki teaches generating time-series data images comprising multiple sub-images of sensor data, and Thyssen provides an image editing software capable of combining and padding images.
Machine learning models can be configured to operate on inputs having specific dimensions, and input deviating from the specified dimensions can cause a model to operate less effectively.
Therefore, it would have been obvious to one of ordinary skill in the art, before the effective filing date, to use the methods of Thyssen to pad the time-series data images provided by the combination of Mori and Aoki to provide consistently formatted inputs to the model of Murase, predictably improving the performance of the model.
Claim(s) 14-15 is/are rejected under 35 U.S.C. 103 as being unpatentable over Murase (JP2020144619A, 2020-09-10) in view of Mori (JP2018051783A, 2018-04-05) further in view of Aoki (JP2018092453A, 2018-06-14) as applied to claim 1 above, further in view of Sungho Suh et al (hereinafter Suh) (“Supervised Health Stage Prediction Using Convolutional Neural Networks for Bearing Wear,” 2020-10-16)
Regarding claim 14, Murase, Mori, and Aoki fail to teach but Suh teaches;
wherein the time-series data image is a scatter diagram.
([pg. 3] transforming the time series data of raw vibration data into a nested-scatter plot image (NSP) for feature extraction)
OBVIOUSNESS TO COMBINE SUH WITH MURASE, MORI, AND AOKI:
Suh is analogous art to the present disclosure as it pertains to time-series data images represented as scatter diagrams.
Murase teaches the time-series image abnormality detection pipeline, Mori teaches deriving time-series sensor data on different axes from screws of a twin-screw kneading extruder, Aoki teaches generating data images rendering a plurality of sensors / variables, and Suh teaches time series data images being represented as scatter diagrams.
Suh further states;
([pg. 4] NSP…is an efficient imaging method for multi-variable correlation analysis)
Therefore, it would have been obvious to one of ordinary skill in the art, before the effective filing date, to configure the time-series data images of Murase as modified by Mori and Aoki to be the scatter diagrams generated using the process of Suh to improve efficiency for multi-variable correlation analysis.
Regarding claim 15, Murase teaches;
time-series data image [features] ([Abstract] creates a color image from pieces of time-series data obtained by a large number of sensors, … extracts the feature of the image to detect the abnormality of a facility)
Murase, Mori, and Aoki fail to teach but Suh teaches;
dimensionally reduced scatter diagram [generated using features from time-series data images]
([pg. 3] Phase one consists of transforming the time series data of raw vibration data into a nested-scatter plot image (NSP) for feature extraction … [pg. 12] The 500-dimensional features were then reduced to 100 dimensions by using principal component analysis (PCA) [30] and further reduced to two dimensions by using t-SNE … Figure 11 and Figure 12 show the reduced features of tested bearings in C1 and C2 mapped into a two-dimensional plane.)
[fig. 11] Resulting scatter diagram:
PNG
media_image7.png
233
951
media_image7.png
Greyscale
OBVIOUSNESS:
Suh is analogous art to the present disclosure as it pertains to deriving features from time-series data images, then deriving scatter plots from said features.
Murase already teaches extracting features from time-series data images, while Suh teaches dimensionally reducing high-dimensional features extracted from time-series data images and generating a scatter diagram from the reduced features, allowing for visual evaluation of the features.
Therefore, it would have been obvious to one of ordinary skill in the art, before the effective filing date, to further process the extracted features from the system of Murase as modified by Mori and Aoki by generating a dimensionally reduced scatter diagram as taught by Suh, to allow for visual evaluation of the features extracted from the time-series data images.
Claim(s) 13 is/are rejected under 35 U.S.C. 103 as being unpatentable over Murase (JP2020144619A, 2020-09-10) in view of Aoki (JP2018092453A, 2018-06-14) further in view of Mori (JP2018051783A, 2018-04-05).
Regarding claim 13, Murase teaches;
and based on training data sets including the plurality of time-series data images generated and a plurality of reference images having any feature,
generating a learning model that outputs a feature according to a normal operation and an abnormal operation
([Abstract] feature-quantity extracting means 17 for inputting the color image into a convolution neural network, and which extracts the outputs by all coupled layers as a feature quantity; and exclusive identifying means 19 for performing an exclusive identification with the extracted feature quantity being input thereto to detect an abnormality.)
NOTE: Teaches generating a learning model which outputs a feature (feature quantity) according to a normal and abnormal operation (using the feature quantity to detect an abnormality)
([pg.2] the parameters of each CNN layer trained using a large-scale supervised image data set are highly versatile, and the target image is obtained by applying fine tuning that tunes the parameters using the training data of the target task.)
NOTE: The learning model (CNN) is generated based on training data including the plurality of time-series data images generated (data of the target task) and a plurality of reference images having any feature (using supervised image data).
Reasoning as to why it would be obvious for the output feature of the model to be generated according to a normal operation and an abnormal operation of the twin-screw kneading extruder in a case where a time-series data image including a first image and a second image rendering displacements of the rotation center axes of the first screw and the second screw is input will be explained further below.
Murase fails to teach but Aoki teaches;
generating a plurality of time-series data images each including, in a single image, a first image rendering, and a second image rendering
([pg.5] Here, an example is given and demonstrated about the image data which the data conversion means 13 produces | generates. FIG. 6 is an example (sensor data table 101) of a table in which a plurality of sensor data used by the learning apparatus 1 is collected. The sensor data table 101 in FIG. 6 is a collection of sensor data acquired by a plurality of sensors numbered 1 to 12. For example, the sensor data table 101 is a collection of sensor data acquired by each sensor at a predetermined timing.)
PNG
media_image8.png
190
1500
media_image8.png
Greyscale
PNG
media_image9.png
110
883
media_image9.png
Greyscale
NOTE: Aoki discloses a single time-series data image which is composed of multiple image renderings (each square in 201 above is an individual sensor value image rendering at a given time, and together the collection of sensor image renderings creates a single time series image). Therefore, teaches a first and second image rendering (at least two image rendering are shown in the image) in a single image.
PNG
media_image3.png
441
885
media_image3.png
Greyscale
PNG
media_image4.png
676
602
media_image4.png
Greyscale
NOTE: The above images show that a plurality of these time series data images are generated. Reasoning as to why is would have been obvious to one of ordinary skill in the art before the effective filing date to combine Aoki with Murase is provided further below.
Murase and Aoki fail to teach but Mori teaches;
A method of [taught by Murase] detecting an abnormality of a twin-screw kneading extruder having a first screw and a second screw, the method [taught by Murase]
([pg.4] As the vibration measuring means 60 used in the kneading extruder 10 according to the present embodiment, vibrations in three directions orthogonal to each other are measured from the viewpoint of reducing the number of parts provided in the kneading extruder 10 and specifying an abnormal part… first screw 14… second screw 16.)
NOTE: Mori pertains to a method of detecting abnormalities of a twin-screw kneading extruder having a first and second screw.
data indicating displacements in a first-axis direction (y) and a second-axis direction (z) that intersect a rotation center axis (x) of the first screw (14) of the twin-screw kneading extruder (10)
data indicating displacements in a third-axis direction (y) and a fourth axis direction (z) that intersect a rotation center axis (x) of the second screw (16) of the twin-screw kneading extruder, (10)
([pg.7] In the multi-axis kneading extruder, a vibration measuring means 60 for measuring vibration in the direction between the screw shafts (y direction in FIG. 2) and vibration in the direction included in the second plane (zx plane in FIGS. 1 and 2))
PNG
media_image2.png
393
866
media_image2.png
Greyscale
([pg.3] In the present specification, the “longitudinal direction” means a direction along the rotation axis of the screw 12 (NOTE: Screw 12 includes the first screw 14 and second screw 16, as pictured above) unless otherwise specified. The x-axis direction shown in FIGS. 1 to 4 is the longitudinal direction, and in this specification, the “x-axis direction” means the longitudinal direction. In the present specification, the “radial direction” means a direction orthogonal to the rotation axis (that is, the longitudinal direction) of the screw 12 unless otherwise specified. The y-axis direction and the z-axis direction shown in FIGS. 1 and 2 are radial directions.)
NOTE: Discloses data indicating displacements (vibrations) in a first and third axis direction (y) and a second and fourth axis direction (z) which intersect a rotation center axis of the first screw and second screw (x, which is the rotation axis of the first and second screws of the twin-kneading extruder).
OBVIOUSNESS TO COMBINE:
Below is the reasoning to combine Murase, Aoki, and Mori for limitation: outputs a feature according to a normal operation and an abnormal operation of the twin-screw kneading extruder in a case where a time-series data image including a first image and a second image rendering displacements of the rotation center axes of the first screw and the second screw is input.
generating a learning model that outputs a feature according to a normal operation and an abnormal operation [Murase]
([Abstract] feature-quantity extracting means 17 for inputting the color image into a convolution neural network, and which extracts the outputs by all coupled layers as a feature quantity; and exclusive identifying means 19 for performing an exclusive identification with the extracted feature quantity being input thereto to detect an abnormality.)
NOTE: Teaches generating a learning model which outputs a feature (feature quantity) according to a normal and abnormal operation (using the feature quantity to detect an abnormality)
of the twin-screw kneading extruder [Mori]
([pg.4] As the vibration measuring means 60 used in the kneading extruder 10 according to the present embodiment, vibrations in three directions orthogonal to each other are measured from the viewpoint of reducing the number of parts provided in the kneading extruder 10 and specifying an abnormal part. It is preferable to use a three-axis vibrometer. At least one of the three directions in which the triaxial vibrometer used as the vibration measuring means 60 measures vibration is the longitudinal direction of the rotational axis of the screw 12 or the radial direction of the rotational axis of the first screw 14. The direction included in the first plane including the rotation axis of the first screw 14 and the rotation axis of the second screw 16.)
NOTE: Mori pertains to a method of detecting abnormalities of a twin-screw kneading extruder having a first and second screw.
in a case where a time-series data image including a first image and a second image rendering [Aoki]
([pg.5] Here, an example is given and demonstrated about the image data which the data conversion means 13 produces | generates. FIG. 6 is an example (sensor data table 101) of a table in which a plurality of sensor data used by the learning apparatus 1 is collected. The sensor data table 101 in FIG. 6 is a collection of sensor data acquired by a plurality of sensors numbered 1 to 12. For example, the sensor data table 101 is a collection of sensor data acquired by each sensor at a predetermined timing.)
PNG
media_image8.png
190
1500
media_image8.png
Greyscale
PNG
media_image9.png
110
883
media_image9.png
Greyscale
NOTE: Aoki discloses a single time-series data image which is composed of multiple image renderings (each square in 201 above is an individual sensor value image rendering at a given time, and together the collection of sensor image renderings creates a single time series image), which includes a first and second image rendering
displacements of the rotation center axes of the first screw and the second screw is input [Mori].
([pg.7] In the multi-axis kneading extruder, a vibration measuring means 60 for measuring vibration in the direction between the screw shafts (y direction in FIG. 2) and vibration in the direction included in the second plane (zx plane in FIGS. 1 and 2). When used, excessive vibration due to mounting angle failure is measured as vibration in the direction between the screw shafts, but not as vibration in the direction included in the second plane. Therefore, the multi-screw kneading extruder provided with the first screw 14 and the second screw 16 and the vibration measuring means 60 for measuring at least vibration in the direction between the screw shafts and in the direction included in the second plane is vibration measurement. It is possible to easily determine whether or not the cause of excessive vibration measured by the means 60 is a problem in the mounting angle.)
PNG
media_image2.png
393
866
media_image2.png
Greyscale
([pg.3] In the present specification, the “longitudinal direction” means a direction along the rotation axis of the screw 12 unless otherwise specified. The x-axis direction shown in FIGS. 1 to 4 is the longitudinal direction, and in this specification, the “x-axis direction” means the longitudinal direction. In the present specification, the “radial direction” means a direction orthogonal to the rotation axis (that is, the longitudinal direction) of the screw 12 unless otherwise specified. The y-axis direction and the z-axis direction shown in FIGS. 1 and 2 are radial directions.)
NOTE: Mori teaches data indicating displacements (vibrations) of the rotation center axes of the first screw and the second screw.
As taught by Mori above, there already exists a method of sensing displacements of the axes of the first screw and the second screw.
This data can be collected at different time intervals to be used to generate a plurality of time series data images using the methods taught by Aoki above.
The plurality of time series data images produced from the process of Aoki could then be used as input for the generated learning model taught by Murase above (the model taught by Murase is configured to take time-series data images as input: [Abstract] feature-quantity extracting means 17 for inputting the color image {which is a time-series data image} into a convolution neural network, and which extracts the outputs by all coupled layers as a feature quantity), which would allow the generated model to output a feature according to a normal operation and an abnormal operation (the model taught by Murase is configured to output a feature according to a normal and abnormal operation in a system: [Abstract] and exclusive identifying means 19 for performing an exclusive identification with the extracted feature quantity being input thereto to detect an abnormality) of the twin-screw kneading extruder in a case where a time-series data image including a first image and a second image rendering displacements of the rotation center axes of the first screw and the second screw is input.
Each of these combinations is simply substituting the data which is being used as input for each process with another appropriate data source (Screw displacement data from Mori being used as input to the time-series data image rendering process taught above by Aoki, which takes sensor data as input, to generate time series data images -> the generated time-series data images being used as input to the learning model taught by Murase which is configured to take time-series data images as input).
Additionally, Murase, Aoki, and Mori are analogous art (using the same reasoning as in claim 8).
Therefore, it would have been obvious to one of ordinary skill in the art, before the effective filing date of the claimed invention, to generate the learning model configured to detect abnormalities [taught by Murase] using a generated plurality of time series data images including a first and second rendering [taught by Aoki] of displacements of the screws in a twin-screw kneading extruder [taught by Mori] as input in order to tailor the learning model be used in the context of detecting abnormalities of a twin-screw kneading extruder system, predictably allowing preventative action to be taken before damage occurs to the system.
Response to Arguments
Applicant's arguments filed 06/03/2026 regarding the 35 U.S.C. 112(f) invocation and resulting rejection under 35 U.S.C. 112(b) have been fully considered but they are not persuasive. Following several citations from the MPEP, the applicant remarks that “the claim terms "acquisition unit," "conversion unit," and "determination unit" provide at least as much structure as the terms "circuit," "detent mechanism," "digital detector," and "reciprocating member." However, terms such as circuit, detent mechanism, digital detector, and reciprocating member all recite some level of structure, whereas vague ‘units’ capable of performing some action (acquisition, conversion, determining), provide no structure. The applicant further remarks; “Additionally the claims provide connections between different components that indicate structure. Applicant thus submits that the claims provide a description of the function that would "connote sufficient structure to one of ordinary skill in the art," per the MPEP guidance. As the claim terms recite sufficient structure to a person having ordinary skill in the art, Applicant submits that the claims do not invoke § 112(f). For the foregoing reasons, Applicant submits that claim 11 does not recite any features invoking § 112(f). As such, Applicant submits the § 112(b) rejection is rendered moot.” These remarks state that the 112(f) invocation and resulting 112(b) rejection are allegedly improper, but the remarks are presented in a conclusory manner. The applicant does not explain how the claims ‘provide connections between different components that indicate structure’, or why the claims would ‘connotate sufficient structure to one of skill in the art’. From this analysis, the rejection of claim 11 under 35 U.S.C 112(b) stands.
Applicant's arguments filed 06/03/2026 regarding the 35 U.S.C. 102 rejections have been fully considered but they are not persuasive. Starting on page 3, the applicant submits that “Murase does not provide any description of what sensor data is obtained, for example, two or more of "a displacement of that intersects a rotation center axis of the first screw, a torque of the first screw, a rotational speed of the first screw..." as recited in claim 1. Thus, Murase fails to show all features of claims 1, 10, and 11; therefore claims 1, 10, and 11 are novel over Murase.” However, as necessitated by the amendments to claims 1, 10, and 11, Murase, Mori, and Aoki are now being used to reject independent claims 1, 10, and 11, under 35 U.S.C. 103, where Mori teaches the limitation requiring two or more of the physical quantities, as reflected by the current office action.
Applicant's arguments filed 06/03/2026 regarding the 35 U.S.C. 103 rejections have been fully considered but they are not persuasive. Starting on page 4, the applicant remarks that;
“the Office Action alleges in part "Aoki teaches: a time-series data image including, in a single image, a first and second image rendering of sensor data." Applicant respectfully disagrees.
Aoki demonstrates a grid structure, where each cell (or pixel) represents a single sensor reading. If each cell were a separate image rendering, each rendering would only show a single sensor reading at a single time. This interpretation fails to show time-series data with physical quantities represented in each axis.
If one were to expand the rendering of Aoki, such that each row were a single image rendering, each row would present a collection of single-dimension data at a single time. Such an interpretation would only show a collection of single-dimension data at a single time. It fails to show time-series data.
Alternatively, if each column of Aoki were interpreted as a single image rendering, each column would show time-series data of a single dimension. Thus under this third interpretation, Aoki fails to show "a first physical quantity...in a first-axis direction and a second physical quantity...in a second-axis direction."
Finally, a fourth interpretation of Aoki is that the grid structure were broken into smaller sub- grids (e.g., 2 X 2, 3 X 3, etc.). Such an interpretation does not cure the deficiencies of the row- and/or column-based interpretations. In fact, in the sub-grid interpretation, each sub-grid would show a collection of single-dimension data over time. Such an interpretation would still fails to show "a first physical quantity...in a first-axis direction and a second physical quantity...in a second-axis direction."
Under any one of these interpretations, Aoki fails to teach multi-dimensional readings as a time series in a single image. Moreover, there is no teaching or suggestion on how to modify the grid-based imaging of Aoki to demonstrate a multi-dimensional time series.
Mori does not cure this defect. Specifically, the Office Action points to Mori to demonstrate vibration measurements in a kneading extruder. Even if vibrations are measured in different planes or axes, such information does not teach or suggest how to generate a time-series data image as recited in claims 1, 10, and 11.”
However, as reflected by the office action, the combination of Aoki and Mori can be interpreted differently than the interpretations presented by the applicant, in a way that does teach on these limitations. Mori teaches multiple single-axis sensors that each measure displacements of the first and second screw on a different axis. Aoki teaches a time-series data image where each row is a time-step and each column is a sensor reading. From this, a reasonable interpretation of the combination of Aoki and Mori that does read on the limitations in question can be that, the first two columns of the time-series data image of Aoki can be a first image, where the first column can render displacements from the sensor data of the first screw regarding the first-axis direction (y) and the second column can render displacements from the sensor data of the first screw regarding the second-axis direction (z) that intersect each other as coordinate axes. Similarly, in this interpretation, the second two columns of the time-series data image of Aoki can be a second image, where the first column of the second image can render displacements of the sensor data of the second screw regarding the third-axis direction (y) and the second column of the second image can render displacements of the sensor data of the second screw regarding the fourth-axis direction (z) that intersect each other as coordinate axes.
From this analysis, the rejections in the current office action under 35 U.S.C. 103 stand.
Applicant's arguments filed 06/03/2026 regarding the new claims have been fully considered but they are not persuasive. The applicant remarks that the “Applicant has added new claims 14 and 15 to describe additional features of the time-series image data. Applicant submits that these details are not disclosed or taught by the cited references. Applicant thus submits that these claims are novel and non-obvious.” However, new claims 14 and 15 are rejected under 35 U.S.C. 103 as being anticipated by Suh as cited in the current office action. Additionally, as reflected by the current office action, claim 15 has been rejected under 35 U.S.C. 112(a) as being unsupported by the specification.
CONCLUSION
Applicant's amendment necessitated the new ground(s) of rejection presented in this Office action. Accordingly, THIS ACTION IS MADE FINAL. See MPEP § 706.07(a). 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.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to Matthew Alan Cady whose telephone number is (571) 272-7229. The examiner can normally be reached Monday - Friday, 7:30 am - 5:00 pm ET.
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, Cesar Paula can be reached on (571)272-4128. 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.
/MATTHEW ALAN CADY/ Examiner, Art Unit 2145
/CESAR B PAULA/ Supervisory Patent Examiner, Art Unit 2145
Prosecution Insights