# US 12,119,070 B2 · Memory failure prediction > Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt - Number: US 12,119,070 B2 - Document ID: US12119070B2 - Status: Granted - Kind: B2 - Application number: 17/727,587 - Filed: 2022-04-22 - Published: 2022-11-10 - Granted: 2024-10-15 - Earliest claimed: 2021-05-06 - Assignees: Micron Technology, Inc. - Applicants: Micron Technology, Inc. - Inventors: Abhishek Chaurasia - Web page: https://patentgrep.com/patent/US12119070B2 ## CPC - G11C 29/08: Functional testing, e.g. testing during refresh, power-on self testing [POST] or distributed testing - G06N 3/0464: Convolutional networks [CNN, ConvNet] - G06N 3/08: Learning methods - G06N 3/09: Supervised learning - G11C 29/028: {with adaption or trimming of parameters} - G11C 29/10: Test algorithms, e.g. memory scan [MScan] algorithms; Test patterns, e.g. checkerboard patterns - G11C 29/16: using microprogrammed units, e.g. state machines - G11C 29/44: Indication or identification of errors, e.g. for repair - G11C 29/52: Protection of memory contents; Detection of errors in memory contents - G11C: STATIC STORES (semiconductor memory devices H10B) - G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS ## Abstract A system, method and apparatus of memory failure prediction through image analyses using an artificial neural network. A sequence of images indicative of progress of memory failures in a region of an integrated circuit die can be generated according to a physical layout of memory cells in the region. The artificial neural network can be trained to recognize graphical features in early images in the sequence and to predict, based on the recognized graphical features, memory failures shown in subsequent images in the sequence. A computing apparatus can use the artificial neural network to analyze an input image shown current memory failures in the region and to identify one or more memory cells in the region that are likely to have subsequent memory failures. ## Claims ### Claim 1 A method, comprising: generating, in a computing apparatus and according to a physical layout of a plurality of memory cells in a region of an integrated circuit die, a sequence of images indicative of progress of memory failures in the region of the integrated circuit die, wherein a first image in the sequence images is indicative of a first memory cell of the plurality of memory cells not having failed and wherein a second image in the sequence of images is indicative of the first memory cell having failed; training, in the computing apparatus using a machine learning technique, an artificial neural network to recognize graphical features in the images and to predict, based on a recognized graphical feature in the first image in the sequence, one or more memory failures shown in the second image appearing after the first image in the sequence; receiving, in the computing apparatus, a third image indicative of one or more memory failures in the region of the integrated circuit die; analyzing, by the computing apparatus using the artificial neural network, the third image; and identifying, based on the analyzing, one or more memory cells in the region, the one or more memory cells identified to have memory failures after a time of the third image. ### Claim 2 (depends on 1) The method of claim 1, wherein each respective memory cell having a memory failure in the region is represented by at least one pixel in the first image at a location corresponding to a location of the respective memory cell in the region on the integrated circuit die. ### Claim 3 (depends on 2) The method of claim 2, wherein the respective memory cell is represented, in the first image, by a plurality of pixels having a predefined shape. ### Claim 4 (depends on 3) The method of claim 3, wherein the plurality of pixels representative of the respective memory cell in the first image have a predefined color different from colors of pixels representative of memory cells having no memory failures. ### Claim 5 (depends on 4) The method of claim 4, wherein the first image includes pixels representative of structural features associated with the plurality of memory cells in the region. ### Claim 6 (depends on 4) The method of claim 4, wherein the first image includes pixels having attributes representative of data indicative of usages of the plurality of memory cells in the region. ### Claim 7 (depends on 4) The method of claim 4, wherein the artificial neural network is configured to perform image segmentation to identify a portion of the first image to identify a graphical feature in the portion. ### Claim 8 (depends on 4) The method of claim 4, further comprising: performing a data refreshing operation in the region in response to the one or more memory cells being identified to have memory failures. ### Claim 9 (depends on 4) The method of claim 4, further comprising: storing a redundant copy of data in the region in response to the one or more memory cells being identified to have memory failures. ### Claim 10 (depends on 4) The method of claim 4, further comprising: moving data from the region to a different region in response to the one or more memory cells being identified to have memory failures. ### Claim 11 An apparatus, comprising: memory storing instructions; and at least one processor configured via the instructions to: receive a sequence of images generated according to a physical layout of a plurality of memory cells in a region of an integrated circuit die to show progress of memory failures in the region of the integrated circuit die, wherein a first image in the sequence images is indicative of a first memory cell of the plurality of memory cells not having failed and wherein a second image in the sequence of images is indicative of the first memory cell having failed; train, using a machine learning technique, an artificial neural network to recognize graphical features in the images and to predict, based on a recognized graphical feature in the first image in the sequence, one or more memory failures shown in the second image appearing after the first image in the sequence; receive a third image indicative of one or more memory failures in the region of the integrated circuit die; and identify, based on an analysis of the third image using the artificial neural network, one or more memory cells in the region, the one or more memory cells identified to have memory failures after a time of the third image. ### Claim 12 (depends on 11) The apparatus of claim 11, wherein each respective memory cell having a memory failure in the region is represented by at least one pixel in a respective image in the sequence at a location corresponding to a location of the respective memory cell in the region on the integrated circuit die. ### Claim 13 (depends on 12) The apparatus of claim 12, wherein the respective memory cell is represented, in the respective image, by a plurality of pixels having a shape different from a shape representative of a memory cell having no memory failure. ### Claim 14 (depends on 13) The apparatus of claim 13, wherein the plurality of pixels representative of the memory cell in the respective image have a predefined color different from one or more colors of pixels representative of memory cells having no memory failures. ### Claim 15 (depends on 14) The apparatus of claim 14, wherein the first image includes pixels representative of structural features associated with the plurality of memory cells in the region. ### Claim 16 (depends on 14) The apparatus of claim 14, wherein the first image includes pixels having attributes representative of data indicative of usages of the plurality of memory cells in the region. ### Claim 17 (depends on 14) The apparatus of claim 14, wherein the artificial neural network is configured to perform image segmentation to identify a portion of the first image to identify a graphical feature in the portion. ### Claim 18 A non-transitory computer readable storage medium storing instructions which, when executed by a microprocessor in a computing device, causes the computing device to perform a method, comprising: receiving first data identifying memory failures in a region of an integrated circuit die in a period of time; receiving second data identifying a physical layout of a plurality of memory cells in the region; generating, based on the first data and the second data, a sequence of images indicative of progress of memory failures in the region of the integrated circuit die, wherein a first image in the sequence images is indicative of a first memory cell of the plurality of memory cells not having failed and wherein a second image in the sequence of images is indicative of the first memory cell having failed; training, using a machine learning technique, an artificial neural network to recognize graphical features in the images and to predict, based on a recognized graphical feature in the first image in the sequence, one or more memory failures shown in the second image appearing after the first image in the sequence; receiving third data identifying one or more memory failures in the region of the integrated circuit die; generating, according to the physical layout, a third image showing the one or more memory failures in the region of the integrated circuit die; analyzing, using the artificial neural network, the third image; and identifying, based on the analyzing, one or more memory cells in the region, the one or more memory cells identified to have memory failures after a time of the third image. ### Claim 19 (depends on 18) The non-transitory computer readable storage medium of claim 18, wherein pixels representative of the plurality of memory cells in the region are configured to be at locations corresponding to locations of the plurality of memory cells in the region of the integrated circuit die. ### Claim 20 (depends on 19) The non-transitory computer readable storage medium of claim 19, wherein the first data includes health and usage information of the plurality of memory cells in the region; and the pixels representative of the plurality of memory cells in the region are configured to have attributes representative of the health and usage information.