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 .
Response to Arguments
Applicant's arguments filed 07/02/2026 have been fully considered but they are not fully persuasive.
Regarding the 101 rejections, applicant’s arguments and amendments to the independent claims are persuasive and overcome the previous 101 rejections. Specifically, applicant’s amended limitations to independent claims of specifying a specific two-stage machine learning architecture for leak detection and leakage zone identification using specific algorithms and clustering steps provides a technical improvement of finding leaks in a water distribution network. See pg. 15-16 of “Remarks”: “The systems of claims 1 and 7 do not merely employ machine-learning models as generic analytical tools. Rather, the claims define a specific computational architecture comprising multiple interoperating machine-learning models arranged in interoperative stages. In the system of claim 1, the claimed partitioning stage first partitions the physical WDN into leakage zones based on leakage characteristics and physical connectivity. The resulting partitioned model is then processed by an unsupervised leakage detection model trained on non-leaking operating conditions to determine whether leakage is present. A separate localization model is applied to the leakage detection data generated by the leakage detection model identify the leakage zone. Thus, the systems of claims 1 and 7 provide a particular multi-stage computational architecture in which topology-aware partitioning, unsupervised leak detection, and localized machine- learning analysis cooperate to improve the computational efficiency and operation of the leak detection system itself while improving localization accuracy within a physical water distribution network. This further represents is a significant improvement over approaches like that described in Chen where a single model is used in an iterative and more computationally expensive framework.” Applicant’s amendments and corresponding arguments that the claimed invention provides a technical improvement to the field of leak detection in water distribution networks are persuasive. Therefore, the 101 rejections are withdrawn.
Regarding the 103 rejections, applicant's arguments filed with respect to the prior art rejections have been fully considered but they are moot. Applicant has amended the claims to recite new combinations of limitations. Applicant's arguments are directed at the amendment. Please see below for new grounds of rejection, necessitated by Amendment.
Applicant’s amendments and arguments regarding the Quinones-Grueiro “An Unsupervised Approach to Leak Detection and Location in Water Distribution Networks” reference have been fully considered and are persuasive. Applicant’s arguments that Quinones-Grueiro does not teach a PCA model that generates a leakage characteristic matrix in claims 4 and 8 are persuasive see “Remarks” pg. 22: “Even though Quinones-Grueiro may describe using PCA, Quinones-Grueiro does not describe using PCA to generate a leakage characterization matrix that is used within the partitioning stage of the claimed system consistent with claim 4. Instead, Quinones-Grueiro expressly assumes that partitioning has already been performed to define monitoring zones. Additionally, Quinones-Grueiro in combination with Chen provides no teaching or suggestion of utilizing PCA in the particular multi-stage framework recited in claim 4, which includes partitioning a WDN into leakage zones using a modified k-means clustering that uses physical connectivity data, a leakage detection ML model, and localization ML model.”.
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.
Claims 1-3, 5-6, and 14 are rejected under 35 U.S.C. 103 as being unpatentable over Chen, et al., Non-Patent Literature “An iterative method for leakage zone identification in water distribution networks based on machine learning” (“Chen”) in view of Quinones-Grueiro, et al., Non-Patent Literature “Novel Leak Location Approach in Water Distribution Networks with Zone Clustering and Classification” (“Quinones-Grueiro”) and further in view of Fan, et al., Non-Patent Literature “Machine learning model and strategy for fast and accurate detection of leaks in water supply network” (“Fan”).
Regarding claim 1, Chen discloses:
A system for leak detection and localization, comprising: one or more processors; and one or more non-transitory machine-readable media storing instructions that, when executed, cause the one or more processors to: (Chen, pg. 1948 col. 2, “All the computations summarized in this article were performed using an Intel(R) Core(TM) i7-6700CPU@ 3.40GHz, with 32 GB of RAM memory. A Windows 10 Home 64-bit operating system [A system for leak detection and localization, comprising: one or more processors; and one or more non-transitory machine-readable media storing instructions that, when executed, cause the one or more processors to:]”).
store a water distribution network (WDN) model representing structural, physical, topological, and/or hydraulic characteristics of the WDN; (Chen, pg. 1940 col. 1, “As shown in Figure 1, the proposed method can be performed in the following steps: (1) EPANET software is used to establish a hydraulic model of the WDN [store a water distribution network (WDN) model representing structural, physical, topological, and/or hydraulic characteristics of the WDN;]”).
partition the WDN model into partition zones by applying a modified k-means clustering algorithm to leakage characteristic data and physical connectivity data,… (Chen, pg. 1940 col. 1, “As shown in Figure 1, the proposed method can be performed in the following steps: (1) EPANET software is used to establish a hydraulic model [and physical connectivity data,…] of the WDN; 23 (2) assuming that l (l>=1) simultaneous leakages occur and the leakage characteristic (leakage sample, the residual vector between normal and abnormal sensor values)…(3) the leakage matrix of the identified leakage zone is generated; [to leakage characteristic data] (4) the k-means clustering is used to divide the identified leakage zone into two parts according to the leakage matrix; [partition the WDN model into partition zones by applying a modified k-means clustering algorithm]”).
receive sensor data from the WDN; (Chen, pg. 1942 col. 2, “Because there are flow sensors and pressure sensors in WDNs [receive sensor data from the WDN;]”).
…and apply a localization machine-learning model to the leak detection data to generate localization data identifying a leakage zone in the WDN, wherein the localization machine-learning model is trained based on labeled leakage data. (Chen, pg. 1940 col. 1 and Figure 1, “(7) each leakage combination of the candidate leakage zones is used as a category label [wherein the localization machine-learning model is trained based on labeled leakage data.] of the classifier model, and the selected features are used to train the RF classifier; […and apply a localization machine-learning model] and (8) if the final accuracy (Acc) of the model is greater than 95%, DSl is input into the trained RF classifier, and the leakage zones and the number of leakage nodes in each leakage zone are output [to the leak detection data to generate localization data identifying a leakage zone in the WDN,]”).
While Chen teaches a system for determining leak locations using a leak characteristic matrix, Chen does not explicitly teach:
wherein the physical connectivity data represents a physical distance between respective junctions in the WDN;
apply a trained unsupervised leakage detection machine-learning model to the partition zones and the sensor data to detect occurrence of leakage in the WDN and generate leak detection data indicative of the occurrence of leakage in the WDN, wherein the leakage detection machine-learning model is trained based on non-leaking data;
Quinones-Grueiro teaches wherein the physical connectivity data represents a physical distance between respective junctions in the WDN; (Quinones-Grueiro, pg. 39, “Clustering is usually applied to define the shape and dimension of network zones. Given a data set of topological parameters D = {c}ni=1 of n nodes with c E Rm, the clustering task can be formulated as finding the z clusters of nodes Gzj=1 = c1,…,cnj that maximize/minimize an optimization function. The three main variables considered for the vector are the geographical coordinates (X,Y) and the topological height of each node. Different methods have been applied for zone division in WDNs [5]. Since uniformity is the main concern within the scope of this paper, the k-medoids clustering algorithm will be used. The optimization problem is the following: equation (2); equation 2 shows that the clustering is performed using the distance between junctions/nodes (i.e. wherein the physical connectivity data represents a physical distance between respective junctions in the WDN;)”).
Chen and Quinones-Grueiro are both in the same field of endeavor (i.e. water leakage). It would have been obvious for a person having ordinary skill in the art before the effective filing date of the claimed invention to combine Chen and Quinones-Grueiro to teach the above limitation(s). The motivation for doing so is that clustering using distances between junctions improves leak location assessments (cf. Quinones-Grueiro, pg. 45, “The simulated experiments with a real network demonstrate that it is possible to obtain a reliable leak location with fewer sensors and within a shorter time horizon than other recently proposed leak location methods. The latter represents superior reliability and lower costs in the leak location task.”).
While Chen in view of Quinones-Grueiro teaches a system for determining leak locations using a leak characteristic matrix and distances, the combination does not explicitly teach:
apply a trained unsupervised leakage detection machine-learning model to the partition zones and the sensor data to detect occurrence of leakage in the WDN and generate leak detection data indicative of the occurrence of leakage in the WDN, wherein the leakage detection machine-learning model is trained based on non-leaking data;
Fan teaches apply a trained unsupervised leakage detection machine-learning model to the partition zones and the sensor data to detect occurrence of leakage in the WDN and generate leak detection data indicative of the occurrence of leakage in the WDN, wherein the leakage detection machine-learning model is trained based on non-leaking data; (Fan, pg. 10 col. 1, “The ANN model achieved excellent performance by utilizing the water pressure data at multiple nodes…A variation of ANN model, the autoencoder neural (AE) network, is developed for leak detection to resolve the challenge of unbalanced data. As an unsupervised ML model, the AE model features unique advantages to work with unbalanced data [apply a trained unsupervised leakage detection machine-learning model to the partition zones and the sensor data to detect occurrence of leakage in the WDN and generate leak detection data indicative of the occurrence of leakage in the WDN,].”, and Fan pg. 10 col. 2, “Since the AE model features the ability to detect abnormal samples from dataset of normal samples, only the non-leaking samples are used for the training purpose [wherein the leakage detection machine-learning model is trained based on non-leaking data;].”).
Chen, in view of Quinones-Grueiro, and Fan are both in the same field of endeavor (i.e. water leakage). It would have been obvious for a person having ordinary skill in the art before the effective filing date of the claimed invention to combine Chen, in view of Quinones-Grueiro, and Fan to teach the above limitation(s). The motivation for doing so is that using an unsupervised model allows for the use of unbalanced data (cf. Fan, pg. 10 col. 1, “As an unsupervised ML model, the AE model features unique advantages to work with unbalanced data.”).
Regarding claim 2, Chen in view of Quinones-Grueiro and Fan teaches the system of claim 1. Chen further teaches wherein the sensor data includes water pressure data. (Chen, pg. 1942 col. 2, “Because there are flow sensors and pressure sensors in WDNs [wherein the sensor data includes water pressure data.]”).
Regarding claim 3, Chen in view of Quinones-Grueiro and Fan teaches the system of claim 1. Chen further teaches wherein the leakage characteristic data includes a leakage characteristic matrix. (Chen, pg. 1940 col. 1, “(3) the leakage matrix of the identified leakage zone is generated [wherein the leakage characteristic data includes a leakage characteristic matrix.]”).
Regarding claim 5, Chen in view of Quinones-Grueiro and Fan teaches the system of claim 3. Chen teaches wherein the leakage characteristic matrix is calculated as seen in claim 1. Chen further teaches based on a leakage matrix of monitored pressure when leakage occurs at respective junctions in the WDN, (Chen, pg. 1940 col. 2 and see the leakage matrix on pg. 1941, “Leakage matrix For a network, assume that there are N nodes, NP pressure sensors, and NQ flow sensors [based on a leakage matrix of monitored pressure when leakage occurs at respective junctions in the WDN,].”).
Fan further teaches using a trained autoencoder (AE) neural network to provide an AE-based leakage characteristics matrix…wherein the trained AE neural network is trained based on a training dataset of non-leaking data. (Fan, pg. 10 col. 1, “A variation of ANN model, the autoencoder neural (AE) network, is developed for leak detection to resolve the challenge of unbalanced data. As an unsupervised ML model, the AE model features unique advantages to work with unbalanced data [using a trained autoencoder (AE) neural network].” and Fan pg. 10 col. 2, “Since the AE model features the ability to detect abnormal samples from dataset of normal samples, only the non-leaking samples are used for the training purpose [to provide an AE-based leakage characteristics matrix…wherein the trained AE neural network is trained based on a training dataset of non-leaking data.].”).
It would have been obvious to one of ordinary skill in the art before the effective filling date of the present application to combine the teachings of Fan with the teachings of Chen and Quinones-Grueiro for the same reasons disclosed in claim 1.
Regarding claim 6, Chen in view of Quinones-Grueiro and Fan teaches the system of claim 1. Fan further teaches wherein the leakage detection machine-learning model further comprises at least one of a principal component analysis (PCA) machine-learning model or an autoencoder machine learning model, in which the PCA and/or autoencoder machine learning models are trained based on non-leaking data. (Fan, pg. 10 col. 1, “A variation of ANN model, the autoencoder neural (AE) network, is developed for leak detection to resolve the challenge of unbalanced data. As an unsupervised ML model, the AE model features unique advantages to work with unbalanced data [wherein the leakage detection machine-learning model further comprises at least one of a principal component analysis (PCA) machine-learning model or an autoencoder machine learning model,].” and Fan pg. 10 col. 2, “Since the AE model features the ability to detect abnormal samples from dataset of normal samples, only the non-leaking samples are used for the training purpose [in which the PCA and/or autoencoder machine learning models are trained based on non-leaking data.].”).
It would have been obvious to one of ordinary skill in the art before the effective filling date of the present application to combine the teachings of Fan with the teachings of Chen and Quinones-Grueiro for the same reasons disclosed in claim 1.
Regarding claim 14, Chen in view of Quinones-Grueiro and Fan teaches the system of claim 1. Chen further teaches wherein the localization machine-learning model comprises a random forest model. (Chen, abstract, “An iterative method combining k-means clustering with the random forest classifier is proposed to identify the leakage zones [wherein the localization machine-learning model comprises a random forest model.].”).
Claims 7, 9, 12-13, and 15-16 are rejected under 35 U.S.C. 103 as being unpatentable over Chen, et al., Non-Patent Literature “An iterative method for leakage zone identification in water distribution networks based on machine learning” (“Chen”) in view of Quinones-Grueiro, et al., Non-Patent Literature “Novel Leak Location Approach in Water Distribution Networks with Zone Clustering and Classification” (“Quinones-Grueiro”) and further in view of Fan, et al., Non-Patent Literature “Machine learning model and strategy for fast and accurate detection of leaks in water supply network” (“Fan”) and Sarrate, et al., Non-Patent Literature “Sensor placement for leak detection and location in water distribution networks” (“Sarrate”).
Regarding claim 7, Chen discloses:
A computer-implemented method, comprising: accessing, from non-transitory computer-readable memory, (Chen, pg. 1948 col. 2, “All the computations summarized in this article were performed using an Intel(R) Core(TM) i7-6700CPU@ 3.40GHz, with 32 GB of RAM memory. A Windows 10 Home 64-bit operating system [A computer-implemented method, comprising: accessing, from non-transitory computer-readable memory,]”).
a water distribution network (WDN) model, the WDN model representing structural, physical, topological, and/or hydraulic characteristics of the WDN; (Chen, pg. 1940 col. 1, “As shown in Figure 1, the proposed method can be performed in the following steps: (1) EPANET software is used to establish a hydraulic model of the WDN [a water distribution network (WDN) model, the WDN model representing structural, physical, topological, and/or hydraulic characteristics of the WDN;]”).
performing, by one or more processors, clustering on a leakage characteristics matrix and physical connectivity data to partition the WDN into partitioned clusters that define leakage zones, in which the leakage characteristics matrix describes the leakage behaviors of each of a plurality of junctions in the WDN, (Chen, pg. 1940 col. 1, “(2) assuming that l (l>=1) simultaneous leakages occur and the leakage characteristic (leakage sample, the residual vector between normal and abnormal sensor values)…(3) the leakage matrix of the identified leakage zone is generated; [in which the leakage characteristics matrix describes the leakage behaviors of each of a plurality of junctions in the WDN,] (4) the k-means clustering is used to divide the identified leakage zone into two parts according to the leakage matrix; [performing, by one or more processors, clustering on a leakage characteristics matrix and physical connectivity data to partition the WDN into partitioned clusters that define leakage zones,]”).
…determining, by the one or more processors, centroids of the partitioned clusters, (Chen, pg. 1940 col. 1, “(4) the k-means clustering is used to divide the identified leakage zone into two parts according to the leakage matrix; […determining, by the one or more processors, centroids of the partitioned clusters,]”).
…and applying, by the one or more processors, a leakage localization machine-learning model to the leakage detection data to locate a leakage zone in the WDN, in which the leakage localization machine-learning model is trained based on labeled leakage data. (Chen, pg. 1940 col. 1 and Figure 1, “(7) each leakage combination of the candidate leakage zones is used as a category label [in which the leakage localization machine-learning model is trained based on labeled leakage data.] of the classifier model, and the selected features are used to train the RF classifier; […and applying, by the one or more processors, a leakage localization machine-learning model to the leakage detection data] and (8) if the final accuracy (Acc) of the model is greater than 95%, DSl is input into the trained RF classifier, and the leakage zones and the number of leakage nodes in each leakage zone are output [to locate a leakage zone in the WDN,]”).
While Chen teaches a system for determining leak locations using a leak characteristic matrix and a classifier model, Chen does not explicitly teach:
and the physical connectivity data represents a physical distance between respective junctions in the WDN;
wherein sensors are positioned in the WDN at locations corresponding to the centroids;
applying, by the one or more processors, a leakage detection machine-learning model to detect leakage that occurs in the WDN and provide leakage detection data based on sensor data associated with the leakage zones, in which the leakage detection machine-learning model is trained based on non-leaking data;
Quinones-Grueiro teaches and the physical connectivity data represents a physical distance between respective junctions in the WDN; (Quinones-Grueiro, pg. 39, “Clustering is usually applied to define the shape and dimension of network zones. Given a data set of topological parameters D = {c}ni=1 of n nodes with c E Rm, the clustering task can be formulated as finding the z clusters of nodes Gzj=1 = c1,…,cnj that maximize/minimize an optimization function. The three main variables considered for the vector are the geographical coordinates (X,Y) and the topological height of each node. Different methods have been applied for zone division in WDNs [5]. Since uniformity is the main concern within the scope of this paper, the k-medoids clustering algorithm will be used. The optimization problem is the following: equation (2); equation 2 shows that the clustering is performed using the distance between junctions/nodes (i.e. and the physical connectivity data represents a physical distance between respective junctions in the WDN;)”).
Chen and Quinones-Grueiro are both in the same field of endeavor (i.e. water leakage). It would have been obvious for a person having ordinary skill in the art before the effective filing date of the claimed invention to combine Chen and Quinones-Grueiro to teach the above limitation(s). The motivation for doing so is that clustering using distances between junctions improves leak location assessments (cf. Quinones-Grueiro, pg. 45, “The simulated experiments with a real network demonstrate that it is possible to obtain a reliable leak location with fewer sensors and within a shorter time horizon than other recently proposed leak location methods. The latter represents superior reliability and lower costs in the leak location task.”).
While Chen in view of Quinones-Grueiro teaches a system for determining leak locations using a leak characteristic matrix and distances, the combination does not explicitly teach:
wherein sensors are positioned in the WDN at locations corresponding to the centroids;
applying, by the one or more processors, a leakage detection machine-learning model to detect leakage that occurs in the WDN and provide leakage detection data based on sensor data associated with the leakage zones, in which the leakage detection machine-learning model is trained based on non-leaking data;
Fan teaches applying, by the one or more processors, a leakage detection machine-learning model to detect leakage that occurs in the WDN and provide leakage detection data based on sensor data associated with the leakage zones, in which the leakage detection machine-learning model is trained based on non-leaking data; (Fan, pg. 10 col. 1, “The ANN model achieved excellent performance by utilizing the water pressure data at multiple nodes…A variation of ANN model, the autoencoder neural (AE) network, is developed for leak detection to resolve the challenge of unbalanced data. As an unsupervised ML model, the AE model features unique advantages to work with unbalanced data [applying, by the one or more processors, a leakage detection machine-learning model to detect leakage that occurs in the WDN and provide leakage detection data based on sensor data associated with the leakage zones,].”, and Fan pg. 10 col. 2, “Since the AE model features the ability to detect abnormal samples from dataset of normal samples, only the non-leaking samples are used for the training purpose [in which the leakage detection machine-learning model is trained based on non-leaking data;].”).
Chen, in view of Quinones-Grueiro, and Fan are both in the same field of endeavor (i.e. water leakage). It would have been obvious for a person having ordinary skill in the art before the effective filing date of the claimed invention to combine Chen, in view of Quinones-Grueiro, and Fan to teach the above limitation(s). The motivation for doing so is that using an unsupervised model allows for the use of unbalanced data (cf. Fan, pg. 10 col. 1, “As an unsupervised ML model, the AE model features unique advantages to work with unbalanced data.”).
While Chen in view of Quinones-Grueiro and Fan teaches a system for determining leak locations using a leak characteristic matrix, distances, and unsupervised models, the combination does not explicitly teach:
wherein sensors are positioned in the WDN at locations corresponding to the centroids;
Sarrate teaches wherein sensors are positioned in the WDN at locations corresponding to the centroids; (Sarrate, pg. 7, “In this work, a reduction in the number of candidate sensors is proposed by applying the k means algorithm to partition the n initial sensors into l groups (l ≤ n). Then, a representative sensor will be selected for each cluster, setting up the new candidate sensor set.”, and Sarrate, pg. 8, “Once the elements xj (sensors) have been grouped in l clusters, the most representative sensors ci, i=1, …, l can be chosen as the nearest ones to the cluster centroids among the elements of each cluster [wherein sensors are positioned in the WDN at locations corresponding to the centroids;].”).
Chen, in view of Quinones-Grueiro and Fan, and Sarrate are both in the same field of endeavor (i.e. water leakage). It would have been obvious for a person having ordinary skill in the art before the effective filing date of the claimed invention to combine Chen, in view of Quinones-Grueiro and Fan, and Sarrate to teach the above limitation(s). The motivation for doing so is that k-means centroids can find the optimal placement of sensors in a WDN (cf. Sarrate, pg. 11, “This work proposes a methodology to decide the optimal placement of pressure sensors to maximize the water network leak diagnosability. The goal is to characterize and determine a sensor set that guarantees a maximum degree of diagnosability while a budgetary constraint is satisfied. To overcome the complexity of the problem this work proposes the combination of a branch and bound search based on a structural model of the distribution network and clustering techniques.”).
Regarding claim 9, Chen in view of Quinones-Grueiro, Fan, and Sarrate teaches the method of claim 7. Chen teaches further comprising calculating the leakage characteristics matrix as seen in claim 7. Fan further teaches using an autoencoder neural network, wherein the autoencoder neural network is trained based on non-leaking data. (Fan, pg. 10 col. 1, “A variation of ANN model, the autoencoder neural (AE) network, is developed for leak detection to resolve the challenge of unbalanced data. As an unsupervised ML model, the AE model features unique advantages to work with unbalanced data [using an autoencoder neural network,].” and Fan pg. 10 col. 2, “Since the AE model features the ability to detect abnormal samples from dataset of normal samples, only the non-leaking samples are used for the training purpose [wherein the autoencoder neural network is trained based on non-leaking data.].”).
It would have been obvious to one of ordinary skill in the art before the effective filling date of the present application to combine the teachings of Fan with the teachings of Chen, Quinones-Grueiro, and Sarrate for the same reasons disclosed in claim 7.
Regarding claim 12, Chen in view of Quinones-Grueiro and Fan teaches the system of claim 1. Chen further teaches wherein the sensor data is pressure sensor data from respective sensors in the WDN, (Chen, pg. 1942 col. 2, “Because there are flow sensors and pressure sensors in WDNs [wherein the sensor data is pressure sensor data from respective sensors in the WDN,]”).
While the combination teaches a system for determining leak locations using a leak characteristic matrix, distances, and unsupervised models, the combination does not explicitly teach and the instructions to partition further cause the processor to: determine sensor placement for the respective sensors within each of the partition zones.
Sarrate teaches and the instructions to partition further cause the processor to: determine sensor placement for the respective sensors within each of the partition zones. (Sarrate, pg. 7, “In this work, a reduction in the number of candidate sensors is proposed by applying the k means algorithm to partition the n initial sensors into l groups (l ≤ n). Then, a representative sensor will be selected for each cluster, setting up the new candidate sensor set.”, and Sarrate, pg. 8, “Once the elements xj (sensors) have been grouped in l clusters, the most representative sensors ci, i=1, …, l can be chosen as the nearest ones to the cluster centroids among the elements of each cluster [determine sensor placement for the respective sensors within each of the partition zones.].”).
Chen, in view of Quinones-Grueiro and Fan, and Sarrate are both in the same field of endeavor (i.e. water leakage). It would have been obvious for a person having ordinary skill in the art before the effective filing date of the claimed invention to combine Chen, in view of Quinones-Grueiro and Fan, and Sarrate to teach the above limitation(s). The motivation for doing so is that k-means centroids can find the optimal placement of sensors in a WDN (cf. Sarrate, pg. 11, “This work proposes a methodology to decide the optimal placement of pressure sensors to maximize the water network leak diagnosability. The goal is to characterize and determine a sensor set that guarantees a maximum degree of diagnosability while a budgetary constraint is satisfied. To overcome the complexity of the problem this work proposes the combination of a branch and bound search based on a structural model of the distribution network and clustering techniques.”).
Regarding claim 13, Chen in view of Quinones-Grueiro, Fan, and Sarrate teaches the system of claim 12. Sarrate further teaches wherein the placement for each sensor is determined as a centroid of a respective partition zone. (Sarrate, pg. 7, “In this work, a reduction in the number of candidate sensors is proposed by applying the k means algorithm to partition the n initial sensors into l groups (l ≤ n). Then, a representative sensor will be selected for each cluster, setting up the new candidate sensor set.”, and Sarrate, pg. 8, “Once the elements xj (sensors) have been grouped in l clusters, the most representative sensors ci, i=1, …, l can be chosen as the nearest ones to the cluster centroids among the elements of each cluster [wherein the placement for each sensor is determined as a centroid of a respective partition zone.].”).
It would have been obvious to one of ordinary skill in the art before the effective filling date of the present application to combine the teachings of Sarrate with the teachings of Chen, Quinones-Grueiro, and Fan for the same reasons disclosed in claim 12.
Regarding claim 15, Chen in view of Quinones-Grueiro, Fan, and Sarrate teaches the method of claim 7. Sarrate further teaches wherein clustering is further performed to partition the WDN based on an influence of sensor placement within the leakage zones into which the WDN is partitioned. (Sarrate, pg. 7, “In this work, a reduction in the number of candidate sensors is proposed by applying the k means algorithm to partition the n initial sensors into l groups (l ≤ n) [wherein clustering is further performed to partition the WDN]. Then, a representative sensor will be selected for each cluster, setting up the new candidate sensor set.”, and Sarrate, pg. 8, “Once the elements xj (sensors) have been grouped in l clusters, the most representative sensors ci, i=1, …, l can be chosen as the nearest ones to the cluster centroids among the elements of each cluster [based on an influence of sensor placement within the leakage zones into which the WDN is partitioned.].”).
It would have been obvious to one of ordinary skill in the art before the effective filling date of the present application to combine the teachings of Sarrate with the teachings of Chen, Quinones-Grueiro, and Fan for the same reasons disclosed in claim 7.
Regarding claim 16, Chen in view of Quinones-Grueiro, Fan, and Sarrate teaches the method of claim 7. Chen further teaches wherein the localization machine-learning model comprises a random forest model. (Chen, abstract, “An iterative method combining k-means clustering with the random forest classifier is proposed to identify the leakage zones [wherein the localization machine-learning model comprises a random forest model.].”).
Allowable Subject Matter
Claims 4 and 8, are objected to as being dependent upon a rejected base claim, but would be allowable if rewritten in independent form including all of the limitations of the base claim and any intervening claims. The following is a statement of reasons for indication of allowable subject matter:
Regarding claim 4, Below are the closest cited references, each of which disclose various aspects of the claimed invention:
Quinones-Grueiro, et al., “An Unsupervised Approach to Leak Detection and Location in Water Distribution Networks” discloses an unsupervised approach to leak detection and leak locations in water distribution networks. As mentioned above, this reference was previously interpreted to teach generating a leakage characteristic matrix using a PCA model. However, applicant’s arguments are convincing and even though Quinones-Grueiro teaches an unsupervised approach to leak detection using a PCA model, Quinones-Grueiro does not explicitly teach a leakage characteristic matrix calculated using a PCA model to provide a PCA-based leakage characteristics matrix based on a leakage matrix of monitored pressure when leakage occurs at respective junctions in a WDN, and wherein the PCA model is trained based on a training dataset of non-leaking data.
Jang, et al., “Estimation of Leakage Ratio Using Principal Component Analysis and Artificial Neural Network in Water Distribution Systems” discloses a system that combines a neural network with a PCA model to estimate volume of leakage in a water distribution system. The PCA used in the system is used as a data preprocessing step to reduce the number of features available. While Jang teaches system that uses PCA with a neural network, Jang does not explicitly teach a leakage characteristic matrix calculated using a PCA model to provide a PCA-based leakage characteristics matrix based on a leakage matrix of monitored pressure when leakage occurs at respective junctions in a WDN, and wherein the PCA model is trained based on a training dataset of non-leaking data.
Gertler, et al., “Leak detection and isolation in water distribution networks using principal component analysis and structured residuals” discloses a system for leak detection and isolation in water distribution networks using PCA residuals. The PCA is applied for fault diagnosis in the water distribution network. Even though Gertler teaches the use of a PCA model for leak detection, Gertler does not explicitly teach a leakage characteristic matrix calculated using a PCA model to provide a PCA-based leakage characteristics matrix based on a leakage matrix of monitored pressure when leakage occurs at respective junctions in a WDN, and wherein the PCA model is trained based on a training dataset of non-leaking data.
While the above prior arts disclose the aforementioned concepts, however, none of the prior arts, individually or in reasonable combination, discloses all the limitations in the manner recited in claim 4. Specifically, the claim requires wherein the leakage characteristic matrix is calculated using a principal component analysis (PCA) model to provide a PCA-based leakage characteristics matrix based on a leakage matrix of monitored pressure when leakage occurs at respective junctions in the WDN, and wherein the PCA model is trained based on a training dataset of non-leaking data. While the references cited above mention aspects of using a PCA model in a leakage environment, they do not recite wherein the leakage characteristic matrix is calculated using a principal component analysis (PCA) model to provide a PCA-based leakage characteristics matrix based on a leakage matrix of monitored pressure when leakage occurs at respective junctions in the WDN, and wherein the PCA model is trained based on a training dataset of non-leaking data. Therefore, claim 4 is allowable over the prior art.
Regarding claim 8, the claim is similar to claim 4 and is allowable over the art for the similar reasons in claim 4 because claim 8 has the identified allowable subject matter of a calculating a leakage characteristic matrix using a PCA model and the PCA model being trained on non-leaking data.
Claim 10 is allowed. The following is an examiner’s statement of reasons for allowance:
Claim 10 is considered allowable since when reading the claim in light of the specification, per MPEP 2111.01, the references of record alone or in combination do not disclose or suggest the limitations found within claim 10. The claim as a whole with regards to technical features recited by the claim limitations are directed to:
(exemplar claim 10 limitations):
“store, for a water distribution network (WDN) subjected to damage, a graph representation of the WDN including structural, physical, topological, and hydraulic characteristics of the WDN, and performance data of service nodes of the WDN; apply a graph convolution neural network (GCN) model to the graph representation of the WDN to encode information of the WDN and generate output values corresponding to respective repair actions for damaged pipes of the WDN; and apply a deep reinforcement learning method to the output values to select one or more repair actions from the respective repair actions and determine a repair sequence for the damaged pipes based on a reward associated with recovery performance of the WDN”
Below are the closest prior arts, each of which disclose various aspects of the claimed invention:
Yang, et al., CN114245337A discloses a system for arranging leak location sensors in water supply pipe network based on a graph convolution network, which combines graph analysis and topological structure information of water supply distribution network to solve the problem of optimal arrangement of sensors for leak detection. While Yang teaches using a GCN to encode information about a water distribution network, Yang is focused on sensor placement and not pipe repair. Yang also does not teach a specific combination of GCN and reinforcement learning models as claimed in claim 10. Thus, Yang does not explicitly teach a GCN that outputs values corresponding to respective repair actions for damaged pipes of a WDN or a deep reinforcement learning method to the output values to select one or more repair actions from the respective repair actions and determine a repair sequence for the damaged pipes based on a reward associated with recovery performance of the WDN.
Song, et al., “Resilience-based post-earthquake recovery optimization of water distribution networks” discloses a system for seismic resilience analysis framework of WDNs with a focus on the optimization of pipe recovery sequence. The system creates a pipe recovery model that that considers the recovery pipes based on user demands and tiered importance to the infrastructure of a WDN. While Song teaches a system that is concerned with the optimal repair sequence of pipes in a WDN, Song does not teach the use of a GCN or reinforcement learning and thus does not explicitly teach a GCN that outputs values corresponding to respective repair actions for damaged pipes of a WDN or a deep reinforcement learning method to the output values to select one or more repair actions from the respective repair actions and determine a repair sequence for the damaged pipes based on a reward associated with recovery performance of the WDN.
Hu, et al., “Multi-objective deep reinforcement learning for emergency scheduling in a water distribution network” discloses a system that focuses on preventing the spread of contaminated water in a water distribution network. The system prevents the spread of contamination by using a multi-objective reinforcement learning scheme that schedules the optimal response sequence to water valves and hydrants in a WDN to reduce the spread of contaminated water. While Hu teaches the use of reinforcement learning in a WDN, Hu is focused on water contamination instead of pipe leakage. Hu also does not use a specific combined GCN and reinforcement learning models. Therefore, Hu does not explicitly teach a GCN that outputs values corresponding to respective repair actions for damaged pipes of a WDN or a deep reinforcement learning method to the output values to select one or more repair actions from the respective repair actions and determine a repair sequence for the damaged pipes based on a reward associated with recovery performance of the WDN.
In summary, the references made of record, fail to disclose the required claimed technical features recited by the noted claim limitations as a whole and the related dependent claims are found allowable.
Furthermore, the references of record alone or in combination fail to disclose or suggest the combination of limitations found within the independent claim as a whole without hindsight reasoning.
The dependent claims, being further limiting to the independent claims, definite, and enable by the Specification are also allowed.
Any comments considered necessary by applicant must be submitted no later than the payment of the issue fee and, to avoid processing delays, should preferably accompany the issue fee. Such submissions should be clearly labeled “Comments on Statement of Reasons for Allowance”.
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure. Zanfei, et al., “Novel approach for burst detection in water distribution systems based on graph neural networks” discloses a system that uses graph convolutional neural networks for identifying pipe bursts in a water distribution network.
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 NICHOLAS S WU whose telephone number is (571)270-0939. The examiner can normally be reached Monday - Friday 8:00 am - 4:00 pm EST.
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, Michelle Bechtold can be reached at 571-431-0762. 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.
/N.S.W./Examiner, Art Unit 2148
/MICHELLE T BECHTOLD/Supervisory Patent Examiner, Art Unit 2148
Prosecution Insights