WATER DISTRIBUTION INFRASTRUCTURE RENEWAL METHOD

Abstract

The present invention relates to a method for planning the renewal of components of a physical fluid network infrastructure. The infrastructure, preferably a drinking water distribution infrastructure, comprises a set of components (pipes 50, connections 60 or valve accessories 40) including at least one component.The process includes steps of modelling a risk of failure (130) from a data exploration on a set of physical data (102) corresponding to physical parameters (101) of said set of components to constitute groupings (122) of similar components or to enrich the set of physical data from a shared database of physical data sets of several fluid network infrastructures, of prioritizing said component (140) of the set of components taking into account the risk of failure and the priority to renewal and of identifying the components (150) according to their classification.

Description

Scope of the invention The scope of the present invention relates to methods for planning the renewal of components in a physical fluid network infrastructure. It relates particularly, but not exclusively, to the planning of the renewal of components in a drinking water distribution network infrastructure. Anterior artery The provision of drinking water to populations began in France at the end of the 19th century, and the deployment of infrastructure saw a peak in activity between 1960 and 1980. Today, the length of pipes is estimated at nearly 900,000 kilometers, with an estimated value of 100 billion euros. Each year, local authorities withdraw nearly 6 billion cubic meters of water for distribution to water service users to meet their needs. These needs amount to 4.5 billion cubic meters. While some of the difference between the volumes withdrawn and the volumes used is necessary for the operation of the facilities, nearly one billion cubic meters are still lost during distribution. These losses are linked to leaks which mainly come from sealing defects and breaks in pipes, connections and their tap fittings: an often forgotten heritage which is at the origin of the vast majority of the number of leaks. According to some sources, the annual renewal rate of the pipes is only 0.6%, which would correspond to an average lifespan of just over 165 years. This lack of resources devoted to the renewal of heritage whose lifespan is not infinite is causing concern. Identifying drinking water distribution infrastructure that needs renewal is therefore a growing problem. Therefore, it is necessary to allocate resources, whether material, human, financial, or informational, to the most profitable actions in terms of reducing the risk incurred by the infrastructure of not ensuring its level of service. This approach assumes a mastery of the components of risk, and with regard to physical fluid network infrastructures, the prediction of the hazard of failure, such as the probability of breakage for example, and the potential consequences in case of occurrence, as well as a relevant formalization of their intersection, for example, with a criticality grid. Actions resulting from such an approach increase infrastructure performance, seen as the latter's ability to avoid risk. Numerous tools exist to implement such an approach. However, most of these tools, which incorporate a risk model, also include a failure model. These failure models are generally based on a parametric or semi-parametric approach to failures. These parametric or semi-parametric models require imposing constraints on the underlying failure distributions. These models require calibration, which consists of finding the best set of values ​​for the model parameters by defining a "score," which measures how well a certain set of parameter values ​​matches the observed data compared to another, commonly called the maximum likelihood. This calibration requires failure observation data which is collected manually by the operators of the drinking water distribution infrastructure and entered into Geographic Information Systems, i.e. GIS, and / or Computerized Maintenance Management Tools, i.e. CMMS, for failures. Among the shortcomings and limitations identified in existing tools, the absence of a covariate selection process in the failure modeling stage can be cited. The parametric and semi-parametric models used in existing tools are highly sensitive to increases in the dimensionality of the covariates. Indeed, increasing this dimensionality tends to make the data more sparse, and classical statistical methods requiring the principle of significance become less effective. Statistics are heavily impacted by the lack of data density in the space of these covariates. This is why covariate selection for dimensionality reduction is necessary. The lack of a failure analysis process for all network components can also be mentioned. Existing tools focus on pipelines, while other components such as service connections have failure rates equal to or higher than those of pipelines. The inability to implement existing tools when failure observation databases have insufficient depth of observation is also among these limitations. Existing methods do not allow for the calibration of failure models on drinking water distribution infrastructures for which failure data has not been recorded over a time window sufficient for the calibration of parametric or semi-parametric models, as these models are impacted by the lack of data, thus violating the principle of statistical significance. Other shortcomings and limitations identified in existing tools were also mentioned: The absence of a data preparation process prior to the failure modeling stage; and / or, The absence of a process for handling anomalies in the step failure modeling. Preparing the data prior to implementing existing tools can represent up to 80% of the total time. This data preparation involves discovering, structuring, cleaning, enriching, validating, and publishing raw data in a format suitable for analysis with existing tools. It is therefore a tedious task that the operator of the existing tools must perform before any analysis can take place. An anomaly is an observation that deviates significantly from the other observations in the data sample used to calibrate the failure model. These anomalies can: Resulting from incorrect data coding, for example, error assignment of the failure to the component; These failures can be due to the operator's management, for example, the failures are only reported when there are operations of active leak detection, therefore, determines the dating of failures. distorted if there are no continuous active leak detections; To result from a particular event, for example, extreme cold which causes damage to networks sensitive to ground freezing; and / or, Express random variations in failures. Failure to address these anomalies impacts the calibration process and the relevance of the resulting model. Non-convergence of model calibration procedures failures; and / or, Uncontrolled biases introduced into the model when the convergence is possible which leads to an erroneous model. This is why there is a need for new intelligent processes for classifying components of a physical fluid network infrastructure that are candidates for renewal, overcoming these defects and limitations. Description of the invention The present invention aims to resolve all or part of the aforementioned drawbacks by providing a method for planning the renewal of components in a physical fluid network infrastructure; this infrastructure, preferably a water distribution infrastructure, comprising a set of components including at least one component from among a distribution pipe, a branch, and a valve fitting; the method comprising the following steps: modeling a failure risk based on a data exploration on a dataset physical parameters corresponding to said physical parameters set of components; prioritization of the components of the set of components, taking into account the risk of failure and priority is given to renewal; identification of components candidates for renewal in a coherent whole based on the hierarchy. According to one provision, the process of planning the renewal of components of a physical fluid network infrastructure also makes it possible to detect data anomalies resulting from bad data coding which would be due to an error in assigning failures to components or duplication of failure recording, such as the same failure being recorded several times. Another advantage of this provision is that it detects anomalies that reflect the operator's management practices rather than the failure itself. This type of anomaly is particularly prevalent when the operator does not actively search for leaks continuously but only performs ad hoc operations; the dating of failures is therefore skewed because the recorded failure date only represents the date of the ad hoc leak detection operation, which may be carried out many days after the leaks appeared. According to one embodiment, the step of modeling a risk of failure is carried out by the Random Survival Forest method. According to one embodiment, the process for planning the renewal of components of a physical fluid network infrastructure includes a preliminary step of preparing data to process the physical parameters and / or physical data of said set of components for integration into a database. According to one embodiment, the planning process includes a step of correcting physical parameters and / or physical data. According to one embodiment, the process for planning the renewal of components of a physical fluid network infrastructure in which the preliminary data preparation stage includes structuring, cleaning, and enrichment steps. According to one embodiment, the failure risk modeling step includes a step of reducing physical parameters and / or physical data so as to improve the assessment of a failure risk and / or the failure model. According to one embodiment, the step of modelling a risk of failure includes a step of training and validating the model. According to one embodiment, the failure risk modeling step includes a failure model archiving step. According to one embodiment, the step of developing the modeling of a failure risk includes an estimation of the occurrence of a failure risk for each component. According to one embodiment, the component hierarchy step includes a step of establishing homogeneous physical parameters and / or physical data. According to one embodiment, the prioritization step of said section includes a multi-criteria analysis step with weighting of physical parameters and / or physical data. According to one embodiment, the planning process includes a step of grouping data sets so as to form physical data sets of similar components prior to modeling the risk of failure and / or detecting anomalies. According to one embodiment, the planning process includes a grouping step and a step to verify the similarity between the sets of components. According to one embodiment, the grouping step includes a step of enriching the physical data set of a fluid network infrastructure with another data set pooling the physical data of several fluid network infrastructures. Other features and advantages of the invention will become more apparent from the following description of an embodiment of the invention given by way of non-limiting example. List of figures The invention will be better understood with the aid of the detailed description set forth below opposite the drawing, in which: Figure 1 represents an example of infrastructure of drinking water distribution; Figure 2 shows an example of the constitution of data ; Figure 3 illustrates an example of the planning process the renewal of infrastructure components of physical fluid network according to a mode of realization : Figure 4 shows an example of the area under the curve for which is one of the indicators used to validate a failure model according to an embodiment: Figure 5 represents an example of a comparison of multi-criteria analyses that the planning process of renewal of infrastructure components physical fluid network implemented; Figure 6 presents an example of the expected effectiveness of variants. In the detailed description that will follow of the figures defined above, the same elements or elements fulfilling identical functions may retain the same references in order to simplify the understanding of the invention. Figure 1 represents a drinking water distribution infrastructure 1 which may, for example, include a submersible pump 10 in a water table underground. This submersible pump pumps water to fill a tower-mounted reservoir 30 via a supply pipe 20. The drinking water distribution infrastructure 1, shown in Figure 1, is configured to deliver water to domestic users 70 or industrial users 80 through a network of distribution pipes 50 and service connections 60. Additionally, the drinking water distribution infrastructure can be equipped with valve accessories such as valves 40 and fire hydrants 90. The present invention relates to a method for planning the renewal of components of a physical fluid network infrastructure to obtain a set of renewal actions to be undertaken based on all data relating to the physical network infrastructure, its environment, its failures or failures on infrastructures of the same type, data specific to the operation, maintenance and renewal work of the infrastructure. The process for planning the renewal of the present invention comprises several steps as illustrated in Figure 3: data preparation 110; data correction and storage 120; data grouping 122; Modeling the risk of failure of 130 components of the network ; ranking of said section 140 of the whole of sections 90 by risk of failure; and, identification of section 150 with risk of failure higher. During the preparation step 110 of the physical dataset 102 corresponding to the physical parameters 101 of said segment set 90, these physical parameters 101 relate to the components, network infrastructure, their environment, and their failures, and are processed for integration into a database. An example of these physical parameters is shown in Table 1. As shown in Figure 3, the data set preparation step 110 can include 2 steps: the collection of dataset 111: and, cleaning dataset 112. Physical parameters 101 are collected, during the data set collection step 111, as input to the process and include physical data set 102 of the components of the physical network infrastructure, for example for the distribution pipes 50 of the drinking water distribution network infrastructure 1, as shown in Table 1. This physical dataset (102) can be derived from a Geographic Information System (GIS) describing the structure and characteristics of the physical fluid network infrastructure, and from a Computerized Maintenance Management System (CMMS) describing component failures within the physical fluid network infrastructure. All other geographical and engineering characteristics of the physical fluid network infrastructure can also be used, as well as any other data relevant to a person skilled in the art. The set of physical data 102 collected is then structured, cleaned, enriched and validated before being published in a format suitable for continuing the process: this is the cleaning step 112. The cleaning step 112 of the physical dataset 102 allows for the logging of structuring, cleaning, and enrichment operations to revert to a previous state or to create dataset preparation scenarios 110 to be applied later to other physical datasets 102 of the same type. This cleaning step 112 may, however, be optional depending on the embodiment. In order to create and update the database specific to a physical fluid network infrastructure, or to build, update, and enrich a database sharing data from several physical fluid network infrastructures, this data is corrected before being stored. This data storage step is performed via a database. managing data specific to a single physical network infrastructure and shared data from multiple physical network infrastructures. A grouping step 122 of the data sets 112 aims to create physical data sets 102 of similar components prior to failure risk modeling 130 and / or anomaly detection. An analysis of the grouping of the physical data sets 102 ensures the grouping of components into groups, also called clusters. Each group ultimately has high intra-group similarity and low inter-group similarity. According to another embodiment, during the grouping step 122, the physical data sets 102 of components are injected into the datasets from the grouping analysis of a shared physical database of several fluid network infrastructures. If, at the end of this grouping, no similarity is detected, no dataset from the shared database enriches the data of the physical network infrastructure, and, if necessary, an enriched dataset is created. This constitution of a homogeneous data set facilitates the modeling of the risk of failure 130 of a component of the set of components including pipes 50, branches 60 or valve accessories 40. Indeed, the modeling of the risk of failure 130 of the network components makes it possible to build predictive models of the risk of failure of components from a physical data set 102 of one or more physical fluid network infrastructures. With regard to the modelling of the risk of failure of components 130, the corrected physical parameters 101 are grouped together to constitute homogeneous physical data sets 102. The modeling of the risk of failure 130 comprises 4 steps as illustrated in figure 3: reduction in the number of covariates 131; creation of training data samples and Validation of model 132: training and validation of model 133; and, archiving of model 134. Covariate reduction is essential for failure modeling. In practice, many covariates are often available as potential risk factors for failure. In the initial stage of modeling, it is common to introduce a large number of predictors in the hope of improving model accuracy. However, to improve model predictability and interpretation, a parsimonious model is desirable. Thus, reducing the number of covariates plays a significant role in the construction of failure models. Reduction methods are either univariate or multivariate. Examples of multivariate selection methods include cluster analysis, decision trees, and principal component analysis. In one embodiment of the invention, the reduction of the number of covariates on a selected dataset is carried out by first implementing the univariate selection method which eliminates the least significant covariates by training the model covariate by covariate and eliminating the covariates for which the Area Under the Curve indicator (proposed in Time-dependent ROC curves for censored survival data and a dinostic marker, PJ HEAGERTY et al, June 2002) is less than a threshold set at 0.7, then from the set of retained covariates, the selection of the relevant covariates is carried out by implementing the following method: training the model with all the retained covariates, then eliminating from the list of retained covariates the least significant covariate for a type I error threshold set at 0.35. During the creation of the training data samples132 and model validation, whether for parametric, semi-parametric, or non-parametric models, it is necessary to create two subsets of the data: the first to calibrate the models parametric or semi-parametric models or training non-parametric models, the second to validate the performance of the model obtained. Regarding the training and validation phase of model 133, the objective is to obtain, at the end of this phase, a model capable of predicting the risk of failure of physical network infrastructure components. To obtain this final model, it is necessary to perform calibration for parametric or semi-parametric models and training for non-parametric models on the data sample. The calibration of parametric or semi-parametric models can be performed using the maximum likelihood method. The training of non-parametric models is performed using a supervised method. In one embodiment of the invention, the non-parametric model implemented is the Random Survival Forest model described in "Random Survival Forest" - Hemant Ishwaran, Udaya B. Kogalur, Eugene H. Blackstone and Michael S. Lauer - 2008. For model validation, the method proposed by Y. Le Gat, "Evaluation of the performance of a model for predicting breakages in a drinking water network - DEA thesis - 2002," is implemented. This method consists of performing validation by measuring the model's effectiveness in identifying the components most at risk, based on the proportion of failures it would prevent, using the failure rate predicted by the model as the selection indicator. The constructed models are archived, during an archiving step 134, preserving the associative identifier of the dataset on which they were built. They will be used later for prediction in the classification of components eligible for renewal 140. This classification of components eligible for renewal 140 comprises 3 steps as illustrated in Figure 3: Choice of failure model 410; Calculation of the risk of failure 420; and, Multi-criteria analysis 430. The failure model to be applied to the physical network infrastructure under study is chosen, during the selection of failure model 410, from among the models archived in the archiving step of model 134. Thus, it is possible to apply an archived model built from physical datasets 102 specific to a physical fluid network infrastructure or from datasets physical 102 derived from shared data from several physical fluid network infrastructures thus making it possible to overcome a possible insufficiency of failure data. The calculation of the risk of failure 420 of each of the components of the physical fluid network infrastructure which is subject to a hierarchy of components candidates for renewal is carried out by calculating a failure indicator from the model chosen in the previous step, i.e. when choosing the failure model 410. In one embodiment of the invention, this indicator of the risk of failure is the number of breakages predicted for a given time horizon. As for the Multi-criteria Analysis step 430, it is broken down into 4 steps; Construction of the multi-criteria analysis framework 431; Calculation of criteria for each component 432; Calculation of the score for each component 433; and, Comparison of different multi-criteria analysis frameworks 434. The 430 multi-criteria analysis is carried out to classify components eligible for renewal by taking into account several criteria. These criteria and their associated weights reflect the preferences expressed by the manager of the physical fluid network infrastructure with regard to, for example, the risk of failure, the criticality of the component in the network, the need for reinforcement, the cost of repairs avoided, the reduction of water losses, and the need for coordination with other roadworks, for example.In one embodiment of the invention, it is possible to establish the framework for the multi-criteria analysis by constructing the criteria to be taken into account in the analysis, by defining the weights of each of the criteria and by entering additional data allowing the criteria to be calculated if necessary, such as the unit cost of repairing leaks to calculate an indicator of the cost of repairs avoided by replacing the faulty component. Regarding the step of calculating the criteria for each component 432, during the calculation of the criteria, according to the previously defined framework, a normalization is performed using a transformation function which allows us to bypass the problem of incompatibility of units of measurement of the different criteria. This is followed by a step of calculating the final score433 by a weighted sum aggregation method after which the results are saved. During a step comparing different multi-criteria analysis frameworks 434, the results of several multi-criteria analyses are compared based on a selection of the first n components sorted according to their previously calculated score value and representing a certain percentage of the total number of components to be renewed. In one embodiment of the invention, this comparison relies on the selection of a specific percentage of the total number of components to be renewed, on the construction of indicators from the criteria selected for the multi-criteria analysis, and the performance of the multi-criteria analysis is evaluated by calculating, for each selected indicator, an expected benefit. This benefit is calculated as the sum of the indicator value for each of the first n selected components, representing a certain percentage of the total number of components, divided by the sum of the indicator value for all the components of the network infrastructure. Identifying the 150 components to be renewed is necessary to build coherent sets of components to be renewed in terms of site organization and budget envelope. This step of identifying the 150 components to be renewed is broken down into 4 steps: Creation of a cost database renewal 151 Creation of subsets of candidate components renewal 152: Creation of renewal projects 153: Editing of construction site sheets 154: According to one embodiment, it is possible to create a unit price base for renewal 151 per component type. Three groups of components classified as candidates for renewal (152) are created based on three budget allocations. The first subset is formed by grouping the top n components for renewal, classified according to their multi-criteria score, whose total renewal cost is roughly equal to the first budget allocation. The same process is followed for the other two subsets. During the step of creating renewal projects 153 via a dedicated user interface, the user selects from these 3 groups the components to be renewed to constitute coherent projects through a tabular view of these components or a map view in which the group belonging is identified by a color coding of the components. A summary sheet including a map extract of the site and a table of the characteristics and costs of the site is published during a step of editing site sheets 154. In the following example, the process of planning the renewal of components of a physical fluid network infrastructure will be illustrated using a case study focusing mainly on sections. Furthermore, this example can be transposed to other components, such as a branch, with the following parameters: Material: steel / cast iron / HDPE / lead / other; Diameter ; Length ; Date of installation; Material thickness; Type of cover: under roadway / under sidewalk / under natural terrain / other; Disinfectant type: Chlorine gas / Chlorine dioxide / Ozone / other; Subscriber type: detached house / apartment building / commerce / office / craftsman / industrial / Hospital / other; Operating pressure; and / or, Presence of a branch shut-off valve: or a valve, with the following parameters: Type: butterfly / operculum / conical bushel / bushel spherical / guillotine / valve / other; Material ; Nominal diameter; Date of installation; Operating pressure; Connection type: flanged / welded to a push-fit / butt welded / other; Valve actuation: manual / motorized; Status: open / closed: and / or, Location: room / in the ground. As an example, an extract of 18 pipe type 50 components is shown from a set of 7,971 pipes 50. This set represents a total length of 486 km for which 225 failures were recorded over a period of 32 years, from January 1, 1969, to December 31, 2001. The table below illustrates an example of a physical data set 102 corresponding to physical parameters 101 whose abbreviations represent "ID" for identifier, "LONG." for length, "AN" for year of installation, "©" for diameter, "MAT." for material, "PE." for polyethylene, "Fgrise" for gray cast iron, "Fductile" for ductile iron, "AMC" for asbestos cement, "PVC" for polyvinyl chloride, "EMP." for location, "Schemin" for underpass, "Schaussée" for sub-roadway, "PRES." for pressure and "ILP" for linear loss index. LONG ID. AN Oo MAT. EMP. CLOSE. ILP 28963 179.91 1996 63 PE Schemin 45.46 0.64 29968 17.51 ​​1969 150 Fgrise Schemin 67.52 4.08 LONG ID. AN G MAT. EMP. CLOSE. ILP 30050 2.75 1993 150 Fuctile Schaussée 6341 0 30861 63.86 1969 150 Fductile Schaussée 44.42 4.08 28981 45.14 1976 125 AMC Schaussée 61.46 2.46 29152 542.53 2001 63 PVC Schemin 79.27 2.46 30862 3.64 1989 63 PE Schaussée 47.62 4.08 30863 60.66 1989 63 PE Schaussée 47.62 4.08 30864 62.55 1969 150 Fductile Schaussée 48.97 4.08 30865 1.78 1989 63 PE Schaussée 50.39 4.08 30866 41.64 1989 63 PE Schaussée 48.94 4.08 60867 73.09 1989 63 PE Schaussée 55.37 4.08 30868 87.60 1989 63 PE Schaussée 51.38 4.08 30051 852.97 2001 63 PE Schaussée 59.44 0 30869 63.01 1989 90 PE Schaussée 53.95 4.08 30870 1.16 1989 50 PE Schaussée 53.38 4.08 30871 2.12 1989 63 PE Schemin 51.38 4.08 29153 2.78 2001 63 Ppve Schemin 70.20 2.46 Table 1: Data Collection The following table represents records of pipeline failures with leak type identification "leak ID" based on identifiers: ID DATEDEFUITE = —IDDEFUITE 32957 26 / 05 / 2004 2488 31211 29 / 11 / 2007 2594 13126 30 / 05 / 2013 2646 31904 01 / 07 / 2000 2506 33654 21 / 08 / 2013 2677 30536 07 / 08 / 2013 2678 34870 13 / 08 / 2013 2679 31156 22 / 08 / 2013 2680 29679 28 / 09 / 2013 2681 31649 14 / 10 / 2013 2682 30926 22 / 11 / 2013 2683 33860 23 / 11 / 2013 2684 Table 2: Failure Record This set of physical data 102 corresponding to physical parameters 101 of said set of pipes 50 is prepared so as to integrate them into a database during the preparation step 110. This step in preparing datasets 110 collects the dataset. In order to create and update the database specific to a physical fluid network infrastructure or to create, update and enrich a database pooling data from several physical fluid network infrastructures, this data is corrected before being stored 120. The dataset is then grouped in a grouping step 122 to form physical datasets 102 of similar components prior to failure risk modeling 130 and / or anomaly detection. This Nomogenes dataset buildup facilitates the assessment of a risk of failure 130 of a pipeline in the set of pipelines 50. Addressing these anomalies is necessary so that the failure model construction step 130 leads to a model that expresses random variations in failures and is not biased by coding errors or failure recording practices of the operator, for example. Groups and / or anomalies are detected through a statistical study of the number of objects contained in a group relative to the total number: a percentage of elements in a group is less than 0.55% of the total number, the elements of this group are considered as anomalies; certain groups stand out with higher percentages at 40% of the total number. Each group constitutes a game independent data; and, a group, above 75% of the total number, generates the creation of a single dataset. The clustering analysis is performed on the sample, keeping only the following covariates: length ; date of installation; age; material ; number of failures. The identifier is kept for the purpose of finding the group / raw data correspondence. This analysis is performed using a BIRCH-type clustering algorithm (Balanced Iterative Reducing and Clustering using Hierarchies, Zhang, T.; Ramakrishnan, R.; Livny, M. - 1996). The resulting component groups, whether for a physical fluid network infrastructure or for the dataset from a database pooling data from several physical fluid network infrastructures, constitute the data on which failure risk modeling is subsequently performed. The BRICH algorithm used for classified clustering and anomaly detection is unsupervised. The number of output clusters is equal to the number of covariates. Each segment is returned as an integer value, that is, a value belonging to the group of natural numbers N. This integer value represents the cluster to which the segment belongs and corresponds to the "BRICH" column in the following table: AN.DE NB DE ID LONG. AGE MAT. BRICH POSING FAULT. 28963 179.91 38 0 6 3 29968 17.51 ​​11 2 0 30050 2.75 35 0 3 0 30861 63.86 11 0 3 0 28981 45.14 18 0 0 0 29152 542.53 43 0 2 AN.DE NB DE ID LONG. AGE MAT. BRICH POSING FAULT. 30862 3.64 31 0 6 0 30863 60.66 31 0 6 0 30864 62.55 11 0 3 0 30865 1.78 31 0 6 0 30866 41.64 31 0 6 0 60867 73.09 31 0 6 0 30868 87.60 31 0 6 0 30051 852.97 43 0 6 4 30869 63.01 31 0 6 0 30870 1.16 31 0 6 0 30871 2.12 31 0 6 0 Table 3: Data analysis using BRICH A statistical analysis is carried out on the number of pipes per group and the corresponding percentage. GROUP NUMBER OF SECTIONS PERCENTAGE OF TOTAL NUMBER 0 7168 88.93 1 11 0.14 2184 2.24 3,652 8.09 Table 4: Statistical analysis of the number of segments The analysis allows us to define two actions: Group 1 represents 0.14% of the total sample, or a value less than 0.5%. These 50 pipes will be considered anomalies and will be eliminated from the game of the dataset; Group O represents 89% of the total sample and clearly surpass the other subgroups. Once the With anomalies eliminated, all remaining data constitutes the dataset. The Failure Risk Modeling step 130 is carried out on the datasets from the data grouping step 122. Regarding the failures suffered by the components of the physical fluid network infrastructure, it should be noted that the main characteristic of this data is that the failure event that we seek to model may not appear for some components because its non-failure time or lifetime is greater than the observation window of the data series, we thus speak of right-censored data, or because the components enter the observation window at different ages and others may have been taken out of service before the start of the observation window, we then speak of left-truncated data. For the analysis of this type of data, classical statistical methods are ineffective, and a so-called survival analysis approach is used on recurring events such as component failures, as shown in Figure 2. Several parametric or semi-parametric models, such as COXPH or Linear Extended Yule Process LEYP, can be implemented for this type of analysis. An alternative that avoids imposing parametric constraints on the underlying failure distributions is to use non-parametric, machine-learning-type models. These models provide a way to automatically handle high-level interactions and higher-order terms in the covariates, enabling accurate prediction. In one embodiment of the invention, the failure model is a random forest model adapted for survival analysis. Its non-parametric nature requires no assumptions about the distribution between covariates and explained covariates. This model is described in "Random Survival Forest" - Hemant Ishwaran, Udaya B. Kogalur, Eugene H. Blackstone and Michael S. Lauer - 2008. It allows the model to benefit from the robustness of random forest methods, their accuracy, and their ability to discover highly complex interrelationships between covariates, even for large datasets. The risk assessment of failure 230 comprises 4 steps as mentioned previously, including a covariate reduction step 131, of type Univariate or Multivariate, which plays a significant role in the construction of failure models. In this example, the reduction of the number of covariates on a selected dataset is carried out automatically by first implementing the univariate selection method which eliminates the least significant covariates by training the model covariate by covariate and eliminating the covariates for which the "Area under the Curve AUC" indicator (proposed in Time-dependent ROC curves for censored survival data and a dinostic marker, PJ HEAGERTY et al, June 2002) is less than a threshold set at 0.7, then from the set of retained covariates the selection of the relevant covariates is carried out by implementing the following method: training the model with all the retained covariates, then eliminating from the list of retained covariates the least significant covariates for a type I error threshold set at 0.35. During the step of creating the samples132 of training data and validating the model, whether for parametric, semi-parametric or non-parametric models, it is necessary to create 2 subsets of the data the first to calibrate the parametric or semi-parametric models or to train the non-parametric models, the second to validate the quality of the model obtained. Regarding the training and validation phase of model 133, the objective is to obtain, at the end of this phase, a model capable of predicting the risk of failure of physical network infrastructure components. To obtain this final model, it is necessary to perform calibration for parametric or semi-parametric models and training for non-parametric models on the data sample. The calibration of parametric or semi-parametric models can be performed using the maximum likelihood method. In the embodiment where the failure model is of the Random Survival Forest type, | For model validation, the method proposed by Y. Le Gat (Evaluation of the performance of a breakage prediction model in a drinking water network - DEA thesis - 2002) is implemented. It consists of performing validation by measuring the model's effectiveness in identifying the most... risk, based on the proportion of failures that it would help to avoid, using the failure rate predicted by the model as a selection indicator. The constructed models are archived, during an archiving step 134, and will be used later in prediction for the prioritization of components candidates for renewal 140. The model in our example is built on the entire recording period, from January 1, 1998, to January 25, 2016. The calibration period begins on January 1, 1998, and ends on January 17, 2012. Validation is performed from January 17, 2012, to January 25, 2016. The univariate selection method allowed us to retain only the covariates: Length ; Year of installation; ILP; Pressure. The other covariates, having AUC values ​​less than 0.7, are not included. The result is illustrated in the table below: COVARIABLES =— LONG. AN G EMP. ILP PRES. AUC 0.994 0.793 .0647 .0494 0.783 0.979 Table 5: Covariates with AUC The convergence of parametric, semi-parametric, and non-parametric models is based on the p-value test. This test measures the probability that a given model, under the null hypothesis, will obtain the same or an even more extreme value than the observed one. This value must be less than 0.3. The p-values ​​provide an initial indicator of the model's convergence and the importance of the covariates selected in the previous steps. The measurement of the predictive performance of the model is based on the cross-validation procedure described in Y.Le Gat, Modeling of the risk of rupture in drinking water distribution networks - 2012. This method allows us to plot the predictive performance curve of the model represented in figure 4. The curve representing predicted failures (breakdowns) avoided by the model is represented by a dashed line, and the curve representing observed failures avoided is represented by a solid line. The model's predictive performance is calculated from the Area Under the Curve (AUC) of the observed failure curve. This value ranges from 0 to 1. The closer it is to 1, the better the model. In the illustrated case, this value is 0.963. In the prioritization step 140, the first step is to choose a failure model from among the models associated with the data of the studied infrastructure archived in the model archiving step 134. This is followed by the calculation of the failure risk 420 of each of the components of the physical fluid network infrastructure which is the subject of a prioritization of candidate components. In our example, the AUC indicator allows us to choose the best model for forecasting the number of failures. A value of 0.963 is sufficient to select this model and apply the forecast to the period from January 17, 2012, to January 17, 2020. An example of the forecast result is illustrated in the table below with the abbreviations representing "ID" for identifier, "AN" for installation year, "LONG." for length, "©" for diameter, "MAT." for material, "PE." for polyethylene, "Fgrise" for gray cast iron, "Fductile" for ductile iron, "AMC" for asbestos cement, "PVC" for polyvinyl chloride, "EMP." for location, "Schemin" for under-path, "Schaussée" for under-road, "PRES." for pressure and "ILP" for linear loss index, "NB DE DEF. Number of failures, "TAUX DE DEF." for Failure Rate: NB. RATE ID YEAR LONG. @MAT. EMP. CLOSE. ILP DE DE 17.51 ​​150 — Fgrey Road 67.52 4.08 0.36 2.55 2.75 150 Fductiie Schaussée 6341 0 00 jm 012 6386 150 Fductie Shaussee 4442 408 0.15 =—0.28 45.14 125 AMC Road 6146 246 004 010 542.553 63 PVC Schemin 7927 = 2.46 003 001 DEF. DEF. 28963 | 1996 179.91 63 PE Schemin 4546 064 0.00 : 000 29968 | 1969 1751 150 Fgrise …Schaussée 6752 408 0.36 2.55 30050 | 1993 2.75 150 Fductile Schaussée 63.41 0 0.00 012 30861 | 1969 6386 150 Fductile Schaussée 4442 408 0.15 0.28 28981 | 1976 4514 125 AMC Schaussée 6146 246 0.04 0.10 29152 | 2001 54253 63 Pvc Schemin 7927 2.46 003 —001 NB. RATE ID YEAR LONG. @MAT. EMP. CLOSE. ILP DE DE 60,66 63 PE Road 4762 408 000 001 6255 —150 Road construction 4897 408 0.15 0.30 1.78 63 PE Road 5039 408 0.00 019 41.64 63 PE Road 4894 408 003 010 DEF. DEF. 30862 | 1989 3.64 63 PE Road 47.62 4.08 0.00 0.08 30863 | 1989 60.66 63 PE Road 47.62 4.08 0.00 0.01 30864 | 1969 6255 150 Fductile Road 48.97 4.08 0.15 0.30 30865 | 1989 1.78 63 PE Road 50.39 4.08 0.00 0.19 30866 | 1989 41.64 63 PE Road 4894 408 0.03 0.10 Table 6: Forecast Results Multi-criteria analysis is performed to classify components eligible for renewal, taking into account several criteria such as: The cost of repair; Hydraulic criticality; Disruption to road traffic; The losses avoided; Other parameters such as the failure rate or the number of failures. The definition of the criteria is carried out in the Construction stage of the multi-criteria analysis framework 430. Regarding the calculation step of the criteria for each component 432, during the calculation of the criteria, a normalization is carried out according to a transformation function to make them comparable and thus circumvent the problem of incompatibility of units of measurement of the different criteria. This is followed by a step of calculating the final score 433 by a weighted sum aggregation method after which the results are saved. During a step comparing different multi-criteria analysis frameworks 434, the results of several multi-criteria analyses are compared based on a selection of the first n components sorted by their score value calculated previously and representing a certain percentage of the total number of components to be renewed. The criteria created in our example are: The relative weight of the annual cost of repairing the section by report on the amortization of its renewal CREP; The cost of CRP repairs; and, PT road traffic disruption. Multi-criteria analysis involves weighting these criteria. An example of two multi-criteria analyses and the weights assigned to the criteria is illustrated in the following table: SUM ANALYSIS CRP CREP PT MULTI-CRITERIA WEIGHTING FIRST CRITERIA 0.55 0.225 0.225 ANALYSIS SECOND 0.56 0.11 0.33 ANALYSIS Table 7: Multi-criteria analysis The comparison indicators are: repair cost; hydraulic criticality; disruption of road traffic; losses avoided; Other parameters include failure rates. The comparison result is based on 1% of the total linear footage. It is shown in Figure 6. The sections are then classified. This classification, also called prioritization, of the components eligible for renewal 130 makes it possible to identify the components of a physical fluid network infrastructure in taking into account the risk of failure and the operator's preferences in terms of priority for renewal and therefore identification. The score calculated by this analysis is used subsequently. It allows us to create 4 classes of sections with values ​​from 1 to 4, where 1 represents the class of critical and priority 90 sections for renewal and 4 represents the class of the lowest priority 90 sections. An example of the results is shown in the table below: LONG ID. AN Go MAT. EMP. CLOSE. ILP PRIORITY LAID 28963 | 179.91 1996 63 PE Schemin 45.46 0.646 4.0 29968 17.51 ​​1969 150 FGrise Schaussée 67.52 4.087 40 30050 2.75 1993 150 Fductile Schaussée 63.41 0 4.0 30861 63.86 1969 150 Fductile Schaussée 44.42 4.087 40 28981 | 45.14 1976 125 AMC Schaussée 61.46 2.468 4.0 29152 | 542.53 2001 63 PVC Schemin 79.27 2.468 40 30862 3.64 1989 63 PE Schaussée 47.62 4.087 4.0 30863 60.66 1989 63 PE Schaussée 47.62 4.087 4.0 30864 62.55 1969 150 Fductile Schaussée 48.94 4.087 10 30865 1.78 1989 63 PE Schaussée 50.39 4.087 40 30866 | 41.67 1989 63 PE Schaussée 48.94 4.087 40 30867 | 73.09 1989 63 PE Schaussée 55.37 4.087 40 30868 | 87.60 1989 63 PE Schaussée 51.38 4.087 3.0 30051 852.97 2001 63 PE Schaussée 59.44 0 4.0 30869 63.01 1989 90 PE Schaussée 53.95 4.087 4.0 30870 1.16 1989 50 PE Schaussée 53.38 4.087 40 30871 | 2.12 1989 63 PE Schemin 51.38 4.087 4.0 29153 2.78 2001 63 PVC Schemin 70.20 2.468 40 Table 8: Classification Results The results of the ranking derived from a multi-criteria analysis framework are visualized through a dedicated user interface in tabular form or as a map if the component descriptive data has a spatial reference. The analysis result is saved in a database along with its set of criteria and associated weights. Next, identifying the components to be renewed 150 is necessary to build coherent sets of components to be renewed in terms of site organization and budget envelope. A unit price base for renewal 151 per component type is created and a creation of 3 groups of components classified as candidates for renewal 152 is carried out on the basis of 3 budget envelopes for example. During the renewal project creation step 153, the components to be renewed are selected from these three groups to form coherent projects. This selection is made using a tabular view of these components or a map view in which the component group is identified by color coding. A summary sheet, including a map excerpt of the project and a table of the project's characteristics and costs, will then be generated. In another embodiment, the datasets of a physical fluid network infrastructure are injected into the datasets of the shared database resulting from the clustering analysis. Then, a similarity check between the components is performed by searching, through clustering analysis, for the new resulting class containing the most data. If, after this clustering, the datasets of the physical network infrastructure are not mixed with those of the shared database dataset, this process is repeated with another dataset from the shared database. If no similarity is detected, no dataset from the shared database enriches the physical network infrastructure data, and, if necessary, an enriched dataset is created. Thus, thanks to this provision, the process of planning the renewal of components of a physical fluid network infrastructure is applicable to a physical fluid network infrastructure for which failure data would not have been observed over a sufficiently long period to provide enough information to calibrate a failure model by benefiting from failure data from a shared database of several fluid network infrastructures. Furthermore, one of the advantages of the component renewal planning process is the presence of a component renewal project planning stage. This planning stage for component renewal projects includes a step to create a database of renewal costs, via a dedicated user interface, allowing the creation of a unit price database for renewal by component type. This stage of creating a cost database is followed by a stage of creating subsets of components that are candidates for renewal. For example, an initial selection of the ranked components eligible for renewal is made into three groups based on three budget allocations. The first subset is formed by grouping the top n components eligible for renewal, ranked according to their multi-criteria score, whose total renewal cost is roughly equal to the first budget allocation. The same process applies to the other two subsets. This is followed by a step of creating renewal projects, also via a dedicated user interface, through which the user selects one of the three groups, the components to be renewed in order to create coherent projects through a tabular view of these components or a map view in which the group belonging is identified by a color coding of the components, for example. Finally, a site sheet editing step produces a summary sheet including a map extract of the site and a table of the site's characteristics and costs.

Claims

CLAIMS 1. Computer-implemented planning method (100) for the renewal of components (90) of a physical fluid network infrastructure; this infrastructure, preferably a water distribution infrastructure, comprising a set of components (90) comprising at least one component from a distribution pipe (50), a connection (60) and a tap accessory (40) such as a valve; the planning method (100) comprising the following steps: - modeling a risk of failure (130) from a data exploration on a set of physical data (102) corresponding to physical parameters (101) of said set of components (90); - prioritization of the components (140) of the set of components (90) taking into account the risk of failure and the priority for renewal; - identification of components (150) candidates for renewal in a coherent set based on the hierarchy.

2. Planning method (100) according to claim 1, in which the step of modeling a risk of failure (130) is carried out by the Random Survival Forest method.

3. Planning method (100) according to claim 1 or 2, which comprises a preliminary data preparation step (110) for processing the physical parameters (101) and / or physical data (102) of said set of components (90) with a view to their integration into a database.

4. Planning method (100) according to claim 1 or 2, which comprises a step of correcting the physical parameters (101) and / or physical data (102).

5. Planning method (100) according to claim 3, wherein the preliminary data preparation step (110) comprises structuring, cleaning, and enrichment steps.

6. Planning method (100) according to any one of the preceding claims, wherein the step of modeling a failure risk (130) comprises a step of reducing (131) physical parameters (101) and / or physical data (102) so as to improve the assessment of a failure risk (130) and / or the failure model.

7. Planning method (100) according to any one of the preceding claims, in which the step of modeling a risk of failure (130) comprises a step of training and validating the model (133).

8. Planning method (100) according to any one of the preceding claims, wherein the step of modeling a failure risk (130) comprises a step of archiving (134) the failure model.

9. Planning method (100) according to any one of the preceding claims, in which the step of hierarchizing the components (140) comprises a step of constituting homogeneous physical parameters (101) and / or physical data (102).

10. Planning method (100) according to any one of the preceding claims, wherein the step of prioritizing said section comprises a step of multi-criteria analysis (430) with weighting of the physical parameters (101) and / or physical data (102).

11. Planning method (100) according to any one of the preceding claims, wherein the step of developing modeling of a failure risk (130) comprises an estimation of an occurrence of a failure risk (130) for each component for a given duration.

12. Planning method (100) according to any one of the preceding claims, which comprises a step of grouping (122) the data sets (112) so as to form physical data sets (102) of similar components prior to modeling the risk of failure 130 and / or detecting anomalies.

13. Planning method (100) according to any one of the preceding claims, which comprises a grouping step (122) comprising a step of checking the similarity between the sets of components (90). 5 14. Planning method (100) according to the preceding claim, in which the grouping step comprises a step of enriching the physical data set of a fluid network infrastructure with another data set pooling the physical data of several fluid network infrastructures. 10

Images