WO2021150099A1

WO2021150099A1 - Equipment spare part inventory optimization methods and systems

Info

Publication number: WO2021150099A1
Application number: PCT/MY2021/050001
Authority: WO
Inventors: M Anif B ADENAN; M Jamil Khan B NURUL AMIN; Ravishankar RAJAGOPALAN; Ahmad Aiman MOHAMAD; Jayakumar VISWANATHAN; Nurul Aizad M SAFIAN
Original assignee: Petroliam Nasional Berhad (Petronas)
Priority date: 2020-01-21
Filing date: 2021-01-17
Publication date: 2021-07-29

Abstract

Methods and systems for optimizing spare part inventory for equipment are described. In one embodiments a method of processing equipment maintenance report data to optimize spare part inventory for equipment comprises: identifying equipment maintenance reports relating to failed parts in the equipment report data; calculating a time-to-failure for each failed part in the equipment maintenance reports relating to failed parts and thereby generating time-to-failure data for a plurality of equipment-part combinations; for each equipment-part combination of the plurality of equipment-part combinations, fitting the time-to-failure data to a statistical distribution; estimating failure rates for each equipment-part combination of the plurality of equipment-part combinations using the respective statistical distributions; and optimizing spare part inventory for equipment using the estimated failure rates.

Description

EQUIPMENT SPARE PART INVENTORY OPTIMIZATION METHODS AND

SYSTEMS

TECHNICAL FIELD

The present disclosure relates to data processing methods and systems for optimizing equipment spare part inventories based on analysis of input data such as technical reports. BACKGROUND

In many industries and in particular in asset-intensive organizations such as those in the energy and utilities sectors it is necessary to keep an inventory of spare parts for use in the event of failure of a part or for routine maintenance. It is often the case that these inventories are kept at a sub-optimal level be it over-stocking or understocking. If an inventory is understocked this impact the time taken to repair the equipment since the time taken to obtain the necessary spare part will impact the mean time to repair of the equipment. This can have a negative impact on the efficiency of the plant of which the equipment forms part and may also have a financial impact. Conversely, overstocking inventory may take up unnecessary resources for storage and transportation and is also a liability from finance perspective.

Supply chain executives and maintenance/operations engineers often have conflicting goals which will cause a conundrum in an inventory decision making. Supply chain executives want a lower inventory to be carried to free up working capital whereas maintenance/operations engineers want it on a higher side since unavailability will lead to loss of opportunities and/or production which could mount to millions of dollars. The question on where to strike a balance in keeping an optimum amount of inventory can’t be answered by one side but need to be curated in such a way that would satisfy both parties. Stock-out and surplus are supply chain’s biggest dilemma especially in catering to the customer’s requirement. Just-in-case stocks which are usually hiding behind the surplus percentages exist due to the lack of assurance that supply will be arriving just in time when needed. Currently the method for optimizing the inventory is still sub-optimal. It is usually left to the decision of supply chain executives and maintenance engineers based on their tacit knowledge and less on analytics. In some cases, they use the historical consumption patterns to estimate the inventory required. However, the future inventory requirement is linked to the reliability of an equipment and how often it would fail. Any method that does not take into account the expected future failures would not provide optimal inventory recommendations.

SUMMARY

According to a first aspect of the present disclosure, a method of processing equipment maintenance report data to optimize spare part inventory for equipment is provided. The method comprises: identifying equipment maintenance reports relating to failed parts in the equipment report data; calculating a time-to-failure for each failed part in the equipment maintenance reports relating to failed parts and thereby generating time-to-failure data for a plurality of equipment-part combinations; for each equipment-part combination of the plurality of equipment-part combinations, fitting the time-to-failure data to a statistical distribution; estimating failure rates for each equipment-part combination of the plurality of equipment-part combinations using the respective statistical distributions; and optimizing spare part inventory for equipment using the estimated failure rates.

In an embodiment identifying equipment maintenance reports relating to failed parts in the equipment report data comprises classifying equipment maintenance reports as relating to failed parts corresponding to a scenario when a part is replaced due to failure or as relating to suspension corresponding to a scenario when a part is replaced prior to failure.

In an embodiment classifying equipment maintenance reports as relating to failed parts corresponding to a scenario when a part is replaced due to failure or as relating to suspension corresponding to a scenario when a part is replaced prior to failure comprises analyzing text of the maintenance reports. In an embodiment analyzing text of the maintenance reports comprises determining if a word indicating failure is present in a maintenance report and classifying the maintenance report as relating to failed parts if a word indicating failure is present. In an embodiment fitting the time-to-failure data to a statistical distribution comprises selecting a statistical distribution type from a plurality of statistical distribution types and estimating parameters for the selected statistical distribution type.

In an embodiment selecting a statistical distribution type comprises fitting the time-to- failure data to each of a plurality of statistical distribution types, determining an Akaike information criterion for each statistical distribution type and selecting the statistical distribution type with the lowest Akaike information criterion.

In an embodiment estimating failure rates for each equipment-part combination of the plurality of equipment-part combinations using the respective statistical distributions comprises performing a Monte Carlo simulation on the respective statistical distributions.

In an embodiment estimating failure rates for each equipment-part combination of the plurality of equipment-part combinations using the respective statistical distributions comprises grouping equipment-part combinations and estimating failure rates for the grouped equipment-part combinations.

According to a second aspect of the present disclosure, a method of estimating an initial equipment spare part inventory requirement for a plant is provided. The method comprises: identifying equipment maintenance reports relating to failed parts in the equipment report data; calculating a time-to-failure for each failed part in the equipment maintenance reports relating to failed parts and thereby generating time- to-failure data for a plurality of equipment-part combinations; for each equipment-part combination of the plurality of equipment-part combinations, fitting the time-to-failure data to a statistical distribution; estimating failure rates for each equipment-part combination of the plurality of equipment-part combinations using the respective statistical distributions; generating a failure library using the estimated failure rates; looking up estimated failure rates for equipment of the plant in the failure library; and estimating the initial spare part inventory requirement for the plant using the estimated failure rates.

According to a third aspect of the present disclosure a data processing system for processing equipment maintenance report data to optimize spare part inventory for equipment is provided. The system comprises a processor and a data storage device storing computer program instructions operable to cause the processor to: identify equipment maintenance reports relating to failed parts in the equipment report data; calculate a time-to-failure for each failed part in the equipment maintenance reports relating to failed parts and thereby generating time-to-failure data for a plurality of equipment-part combinations; for each equipment-part combination of the plurality of equipment-part combinations, fit the time-to-failure data to a statistical distribution; estimate failure rates for each equipment-part combination of the plurality of equipment-part combinations using the respective statistical distributions; and optimize spare part inventory for equipment using the estimated failure rates.

According to a fourth aspect of the present invention a data processing system for estimating an initial equipment spare part inventory requirement for a plant is provided. The system comprises a processor and a data storage device storing computer program instructions operable to cause the processor to: identify equipment maintenance reports relating to failed parts in the equipment report data; calculate a time-to-failure for each failed part in the equipment maintenance reports relating to failed parts and thereby generating time-to-failure data for a plurality of equipment-part combinations; for each equipment-part combination of the plurality of equipment-part combinations, fit the time-to-failure data to a statistical distribution; estimate failure rates for each equipment-part combination of the plurality of equipment-part combinations using the respective statistical distributions; generate a failure library using the estimated failure rates; look up estimated failure rates for equipment of the plant in the failure library; and estimate the initial spare part inventory requirement for the plant using the estimated failure rates.

According to a fifth aspect of the present disclosure a computer readable medium storing processor executable instructions which when executed on a processor cause the processor to carry out a method as set out above is provided. BRIEF DESCRIPTION OF THE DRAWINGS

In the following, embodiments of the present invention will be described as non-limiting examples with reference to the accompanying drawings in which:

FIG.1 is a flowchart showing an outline of a method of optimizing equipment spare part inventory according to an embodiment of the present invention; FIG.2 is a block diagram showing a data processing system for optimizing equipment spare part inventory according to an embodiment of the present invention;

FIG.3 is a flowchart showing a method of optimizing equipment spare part inventory according to an embodiment of the present invention;

FIG.4 is a flowchart showing a method of pre-processing text of maintenance report data according to an embodiment of the present invention;

FIG.5 is a flowchart showing a method of classifying a maintenance report line item as a failure or a suspension in an embodiment of the present invention;

FIG.6 is a table showing rules for classifying a maintenance report as either a suspension or a failure used in embodiments of the present invention; FIG.7 is a flowchart showing a method of calculating time to failure for parts according to an embodiment of the present invention;

FIG.8 shows statistical distributions used for fitting time to failure data in embodiments of the present invention;

FIG.9 is a table showing the parameters of the statistical distributions shown in FIG.8;

FIG.10 is a flowchart showing a method of selecting the best fitting statistical distribution for time to failure data in embodiments of the present invention; FIG.11 is a table showing example results of a Monte Carlo simulation of number of failures per year obtained using a method according to an embodiment of the present invention;

FIG.12 is a flowchart showing a method of estimating failure statistics according to an embodiment of the present invention;

FIG.13 is a table showing grouping of equipment, material and plants used to generate homogenous data for use in embodiments of the present invention;

FIG.14 is a table showing an example of a failure library used in embodiments of the present invention; FIG.15 is a flowchart showing a method of optimizing equipment spare part inventory for a new plant according to an embodiment of the present invention; and

FIG.16 shows material consumption pattern associated with planned and corrective maintenance activities in an example scenario.

DETAILED DESCRIPTION

The present disclosure relates to the optimization of equipment spare part inventory using the results of analysis of maintenance reports. Embodiments of the present invention help achieve total inventory optimization which can render balance between supply chain and engineering department. Embodiments provide a complete view from demand to supply perspective by taking into consideration spare reliability, its replenishment time and overall cost of ownership. An outline such a method is shown in FIG.1.

FIG.1 is a flowchart showing an outline of a method of optimizing equipment spare part inventory according to an embodiment of the present invention. As shown in FIG.1 , the method 10 comprises four main steps. In step 12, text analytics is performed on maintenance report data to identify planned consumption and unplanned consumption of spare part inventory. In the present disclosure, maintenance events are classified as either failure or suspension. Events are classified as a failure if part fails while in use (unplanned consumption) and are classified as a suspension when a part is replaced as a part of a planned maintenance routine (planned consumption).

In step 14, the failure events identified in step 12 are analyzed and are fitted to failure distributions. Thus, a time to failure (TTF) for each failure event can be determined. The TTF is then modelled and fitted to an automated model.

In step 16, the TTF is fitted to a statistical distribution and a model for mean time between failures (MTBF) is generated. This allows forecasting of the MTBF and therefore forecasting of the likely occurrences of failures.

Finally, in step 18, the optimized model is used to provide insights such as recommended stocking levels for spare parts. The simulation optimization of step 18 is used to arrive at the optimal inventory levels by taking into consideration the downtime cost, double jeopardy and simulation of optimum stocking norms for a specific inventory. FIG.2 is a block diagram showing a data processing system for optimizing equipment spare part inventory according to an embodiment of the present invention. The data processing system 100 is a computer system with memory that stores computer program modules which implement equipment spare part inventory optimization methods according to embodiments of the present invention.

The data processing system 100 comprises a processor 110, a working memory 112, an input module 114, an output module 116, a user interface 118, program storage 120 and data storage 130. The processor 110 may be implemented as one or more central processing unit (CPU) chips. The program storage 120 is a non-volatile storage device such as a hard disk drive which stores computer program modules. The computer program modules are loaded into the working memory 112 for execution by the processor 110. The input module 114 is an interface which allows data, for example maintenance reports to be received by the well log data processing system 100. The output module 116 is an output device which allows data and optimized inventory information calculated by the data processing system 100 to be output The output module 116 may be coupled to display device or a printer. The user interface 118 allows a user of the data processing system 100 to input selections and commands and may be implemented as a graphical user interface.

The program storage 120 stores a maintenance report analysis module 122, a time to failure calculation module 124, a statistical modeling module 126, and an inventory optimization module 128. The computer program modules cause the processor 110 to execute various well log data processing which is described in more detail below. The program storage 120 may be referred to in some contexts as computer readable storage media and/or non-transitory computer readable media. As depicted in FIG.2, the computer program modules are distinct modules which perform respective functions implemented by the data processing system 100. It will be appreciated that the boundaries between these modules are exemplary only, and that alternative embodiments may merge modules or impose an alternative decomposition of functionality of modules. For example, the modules discussed herein may be decomposed into sub-modules to be executed as multiple computer processes, and, optionally, on multiple computers. Moreover, alternative embodiments may combine multiple instances of a particular module or sub-module. It will also be appreciated that, while a software implementation of the computer program modules is described herein, these may alternatively be implemented as one or more hardware modules (such as field-programmable gate array(s) or application-specific integrated circuit(s)) comprising circuitry which implements equivalent functionality to that implemented in software.

The data storage 130 stores various maintenance report and failure rate data. As shown in FIG.2, the data storage 130 stores maintenance report data 132 which comprises maintenance reports 134 and a failure library 136 which stores failure rate data 138 for various equipment spare parts.

The maintenance report data 132 comprises historical failure and maintenance data which forms the basis for Life Data Analysis (LDA) which is central to the reliability analysis used in embodiments of the present invention. The maintenance report data 132 may be obtained from an Enterprise Resource Planning (ERR) system and is comprised of the following tables: (i) Notification Report; (ii) Work Order Long Text; (iii) Work Order Services and Materials; (iv) Work Order Operations Data and (v) Catalogue Profile. For each maintenance report 134, the field engineers enter information which includes a short/long description of the work performed, type of maintenance work (planned versus unplanned), the material and the quantity replaced.

Previously, Reliability Engineers review these records manually for each maintenance order to identify the materials replaced after failure (henceforth referred to as failures) and materials replaced before failure (henceforth referred to as suspensions). The identification of the failures and suspension is laborious and could take one to two months to complete for one plant which comprises of thousands of such records. The identification is done through rules derived from interviews with the maintenance personnel as well as tacit domain knowledge of the engineer. However, as is described in detail below, in embodiments of the present invention this identification process is automated by text analytics.

The failure library 136 stores failure rate data 138 which is estimated using the methods described below. An example of the failure rate data 138 stored in the failure library 136 is shown in FIG.14 and is described in more detail below.

FIG.3 is a flowchart showing a method of optimizing equipment spare part inventory according to an embodiment of the present invention. The method 300 shown in FIG.3 is carried out by the data processing system 100 shown in FIG.2.

In step 302, the maintenance report analysis module 122 identifies equipment maintenance reports 134 which relate to filed parts from the maintenance reports stored as maintenance report data 132. Step 302 comprises a text analysis of the maintenance reports 134 to classify each of the maintenance reports as corresponding to a scenario in which a part is replaced due to failure or as relating to a suspension in which a part is replaced prior to failure. Step 302 may comprise a pre-processing step to standardize the text before searching for materials and failure related words. FIG.4 provides the steps followed during text preprocessing. In the present disclosure, the term “material” is used to mean “part" or “spare part”.

FIG.4 is a flowchart showing a method of pre-processing text of maintenance report data according to an embodiment of the present invention. The steps of the method 400 shown in FIG.4 may be carried out by the maintenance report analysis module 122 of the data processing system 100 during step 302 of the method 300 shown in FIG.3.

The input to the method 400 is the raw text 402 of the maintenance reports 134. First, numbers are removed in step 404, special characters are removed in step 406 and punctuation is removed in step 408 from the raw text columns of the notifications and work order data. Then, the text is standardized by converting it to lower case in step 410. A dictionary-based spell check is performed in step 412 to correct spelling errors. A customized spell check dictionary is developed based on data from several plants. In step 414, word expansions are performed to expand shortened forms of the words. In step 416, stop words are then removed from the text Stop words are those that are either common words in English such as “the”, “a”, “an” or domain specific words such as plant names which typically are not helpful in identifying failures from suspensions. In step 418 stemming is performed. Stemming is the process identifying the root/stem word so that different variations of the same word are treated the same. Stemming completes initial pre-processing of the text and the output of the method 400 is normalized text 420. The normalized text 420 is used for failure identification.

The normalized text is check for two key pieces of information: (i) material consumed is present in the material description, notification long text, work order long text or order operation text; and (ii) words describing failure is present in damage code, notification text, work order long text or work order operation text columns.

A line item in a maintenance report is classified as a failure if both the conditions (i) and (ii) mentioned above are met. Otherwise the line item is considered as suspension. An example of the classification is shown in FIG.5. FIG.5 is a flowchart showing a method of classifying a maintenance report line item as a failure or a suspension in an embodiment of the present invention. The steps of the method 500 shown in FIG.5 may be carried out by the maintenance report analysis module 122 of the data processing system 100 during step 302 of the method 300 shown in FIG.3.

The input to the method 500 is the normalized text 420 output by from the method 400 described above with reference to FIG.4. In step 502, it is determined whether material was consumed in the maintenance report line item. If no material was consumed, the event is classified as a suspension in step 504. If material was consumed, then the method moves to step 506. In step 506, it is determined whether failure words are present in the damage code, notification text, work order long text or work order operation text columns of the maintenance report. If no failure words are present, then the event is classified as a suspension in step 508. If failure words are present, then the event is classified as a failure in step 510.

In addition to the classification described above in relation to FIG.5, in some embodiments, rules may be applied to the maintenance reports to classify events. FIG.6 shows examples of such rules. The rules shown in FIG.6 may take precedence over the classification shown in FIG.5.

FIG.6 is a table showing rules for classifying a maintenance report as either a suspension or a failure used in embodiments of the present invention. As shown in the table 600 of FIG.6, replacement of a control valve or mechanical seal assembly or suction strainer materials 602 is classified as a failure; material consumed as part of a plant statutory inspection 604 is classified as a suspension; plant preventive maintenance for inspection, calibration, rectification, service or test 606 is classified as a failure; first time consumption 608 is classified as a suspension; and material consumed for overhaul, hot gas path inspection or combustion inspection 610 is classified as a suspension. The output from step 302 of the method 300 is a classification of each data point as a failure or suspension. This classification is important for the failure calculations explained in the next step. Returning now to FIG.3, in step 304, the time to failure calculation module 124 calculates a time to failure for each failure event identified in step 302. The objective of later steps is to determine the best fitting distribution for the time to failure data and use the distribution to estimate the failure characteristics, e.g. failure rates, mean life and standard deviation. This is done for each material and equipment combination to ensure homogeneity of the operating conditions and usage patterns. In order to do this, it is necessary to first calculate time to failure from the historical failure data. Several common distributions are fitted to Time to Failure Data and the distribution that best fits the data is selected based on a statistical criterion. The best fitting distribution is an approximation of the data and this distribution is further used as one of the key inputs for Monte Carlo simulations in the Simulation Optimization stage for determining the optimal inventory levels.

In addition, best fitting distribution forms the basis for estimating the failure characteristics, which are used by the end users to validate the inventory recommendations. These failure statistics characterize the failure distribution of a material used in a specific equipment. These are calculated using a Monte Carlo simulation as well.

The first step in the determination of failure characteristics of a material is to calculate the time to failure. The work order data forms the basis of this calculation. The historical failures from work orders are extracted and are ordered by a combination of material, equipment and failure date. The time for the first failure is calculated from the installation date of the equipment and the time for the subsequent failures is calculated as the time difference between them. In case of missing installation date, 1 Jan 2005 is assumed to be the installation date. In case the installation is not available, the first failure considered as a Suspension (which indicates replacement before failure) and not a failure.

An overview of the calculation of the time to failure in step 304 is shown in FIG.7. FIG.7 is a flowchart showing a method of calculating time to failure for parts according to an embodiment of the present invention. The method 700 shown in FIG.7 may form part of the step 304 of the method 300.

The input to the method 700 is historical data 702 which indicates the past failures of equipment parts. In step 704, it is determined whether the installation date of the equipment is known. If the installation date of the equipment is known, the method moves to step 706. In step 706, the time to failure (TTF) for the first failure is calculated from the installation date. If the installation date of the equipment is not known, then the method moves to step 708. In step 708, the first failure is assumed to be a suspension and the time to first suspension is calculated from an assumed start date. For subsequent failures, the TTF is calculated from the previous (last) failure date and the current date in step 710.

Returning again to FIG.3, in step 306, the statistical modelling module 126 fits the time to failure data calculated in step 304 to statistical models. Once the time to failure is calculated for a material-equipment combination, the next step is to determine which distribution fits this data the best. For time to failure data, typically one of the following distributions fits the best (i) Weibull Distribution; (ii) Exponential Distribution; (iii) Log Normal Distribution; (iv) Gamma Distribution and (v) Normal Distribution. A summary of these distributions is shown in FIG. 8.

FIG.8 shows statistical distributions used for fitting time to failure data in embodiments of the present invention. As shown in FIG.8, the statistical distributions are the Weibull Distribution 802, the exponential distribution 804, the log normal distribution 806, the gamma distribution 808 and the normal (or Gaussian) distribution 810.

Each distribution is characterized by a set of parameters. These are shown in FIG.9. FIG.9 is a table showing the parameters of the statistical distributions shown in FIG.8. As shown in FIG.9, Weibull distribution is characterized by shape (β) and scale (η) parameters whereas Exponential distribution has only one parameter rate (λ). The log normal distribution is characterized by log (μ) and log (σ), the gamma distribution is characterized by shape (k) and scale (ϑ). The normal distribution is characterized by expected value (μ) and standard deviation (σ).

The parameters of each of the distribution corresponding the Time to Failure data is estimated using Maximum Likelihood Estimate (MLE) method. MLE is a widely used approach in statistics that identifies the optimum parameters by maximizing a likelihood function specific to a distribution.

The likelihood corresponding to the optimal parameter for each distribution can be used to calculate a metric Akaike Information Criteria (AIC) as follows:

AIC = 2k -2 In (L)

Where k is the number of parameters being estimated and L is the likelihood for a given distribution and data. The distribution with the lowest AIC is considered the best fitting distribution.

FIG.10 is a flowchart showing a method of selecting the best fitting statistical distribution for time to failure data in embodiments of the present invention. The method 1000 shown in FIG.10 may form part of the step 306 of the method 300. The input to the method 1000 is time to failure data 1002, which may be the output of the method 700 shown in FIG.7.

In step 1004, the time to failure data 1002 is fitted to each of the different statistical distributions using maximum likelihood estimate (MLE). In step 1006, the Akaike Information Criterion (AIC) is calculated for each distribution. In step 1008, one distribution is selected based on the lowest AIC.

Returning again to FIG.3, in step 308, the statistical modelling module 126 estimates failure rates for equipment-part combinations. Once the best distribution for time to failure is identified it can be used to perform Monte Carlo simulations by generating random numbers from the distribution. To calculate the failure rate and mean time between failures, random numbers are generated from best fit distribution. Since the distribution was modeled on time to failure, the random numbers generated are representative of random failure times. Once a sizable random sample is generated for time to failure, the cumulative time to failure is calculated by sequentially adding the random time to failure.

Number of failures expected in Year 1 to Year 20 is calculated by identifying the number of failures within a specific time period. For example, number of failures in Year 1 is estimated by counting the number of cumulative time to failures of less than 365 days.

The above simulation is typically repeated for thousands of iterations (typically between 1000 and 100000) to generate several random samples and total count of failures in each time period is calculated as the sum over all the iterations.

FIG.11 is a table showing example results of a Monte Carlo simulation of number of failures per year obtained using a method according to an embodiment of the present invention. FIG.11 shows an example of the total expected failures in Year 1, Year 2, ... Year n, for 10 materials based on 100000 simulations.

The failure rate is estimated by dividing this count by the number of simulations. Failure rate indicates probability of failure per unit time. Mean time between failures is calculated as the time for the first failure. FIG.12 provides a summary of the calculations involved.

FIG.12 is a flowchart showing a method of estimating failure statistics according to an embodiment of the present invention. The method 1200 shown in FIG.12 may form part of step 308 of the method 300. The input to the method 1200 is the best fitting distribution 1202 determined according to the method 1000 shown in FIG.10.

The method 1200 comprises repeating a Monte Carlo simulation 1210 multiple times on the best fitting distribution 1202. The Monte Carlo simulation 1210 comprises generating random times to failure 1212 from the best fitting distribution 1202. Cumulative times to failure are calculated 1214 from the random times to failure. From this result, an estimate of the number of failures in each year is calculated 1216. This allow failure rate and mean time between failures 1220 to be calculated as the time for the first failure.

In the failure rate analysis so far, it is assumed that there is sufficient historical data available to build the reliability models. Typically, at least 5 failure points need to be available for fitting failure distributions mentioned earlier. Ideal case, availability of sufficient data for a plant-equipment-material combination. This would help to estimate the failure rates for each equipment-material combination. In case of lack of sufficient data, grouping as shown in FIG.13 is used to create groups of homogenous data that is expected to exhibit same behavior as the ungrouped data. FIG.13 is a table showing grouping of equipment, material and plants used to generate homogenous data for use in embodiments of the present invention.

As shown in FIG.13, the first grouping is the material - plant group level 1310. For the same material, data is collected at plant level (Case 1), similar plants (at plant level, Case 2), similar plants combined (Case 3) and all plants combined (Case 4). These cases are in the order of preference. If there is sufficient data for a material for Case 1 , remaining cases will not be executed.

The second grouping is at the material family type level 1320. For example, if sufficient data is not available for a specific material number of a gasket, then data is group by gasket type which could spiral wound gaskets, flat ring gaskets etc. This grouped data is analyzed at different plant combination levels which leads to Case 5 - Case 8.

The third grouping is at the material family level 1330. For example, all the data of gaskets is analyzed together at different combination of plants. This leads to Cases 9 - 12. The failure rates obtained from the grouped data is used to represent the failure rates of all the materials within that group. FIG.13 shows a summary of the 12 cases resulting from different groups of materials and plants. At this stage, failure rates for materials and its cluster are assembled into one virtual failure dictionary or failure library. As shown in FIG.2, the failure library 136 is stored in the data storage 130 of the data processing system 100 and comprises failure rate data 138 for a plurality of parts or material clusters. The failure library 136 may store failure rate data for equipment-part combinations. These combinations may be specific to particular plant or OPU (operating unit) segments.

FIG.14 is a table showing an example of a failure library used in embodiments of the present invention. These references in a failure library will produce insights for non- moving Inventory which has less or no data for analytics. The library will be able to tell a spiral wound gasket installed in a lean charge pump has n years of mean life thus providing an indication whether keeping the gasket for n number of years is making sense. The failure library shown in FIG.14 may be used to determine an initial spare part inventory proposal for new plants which have no historical failure data would be able to find a match inside the virtual failure library.

Returning again to FIG.3, in step 310, the inventory optimization module 128 optimizes a spare part inventory for the equipment using the estimated failure rates. In some cases, this optimization is carried out for an existing plant, in other cases, the optimization is carried out for a new plant.

FIG.15 is a flowchart showing a method of optimizing equipment spare part inventory for a new plant according to an embodiment of the present invention. The method 1500 is carried out by the inventory optimization module 128 of the data processing system 100 using the failure rate data 138 stored in the failure library 136.

The method 1500 uses a OEM spare part interchangeability record (SPIR) list 1502 which sets out correspondences between spare part definitions. This is combined with a central cataloger 1504 which sets out the parts of the equipment in question. The combination provides a cataloged SPIR 1506. The failure library 136 is then used to determine the estimated failure rates for the parts of the equipment Based on the estimated failure rates, an initial spare part recommendation 1508 is generated. Other inventory optimization tools use warehouse consumptions as their main input to arrive to an optimized inventory stocking level (inventory norms). However, consumption from warehouse can be coming from both failure of the parts, as well as opportunity to replace a part when some other part in the same equipment is being serviced. This scenario will inflate Inventory stocking where Just-in-Case requirement is being mashed up with Just-In-Time requirement. Embodiments of the present invention on the other hand, incorporate probability of material consumption for corrective and planned maintenance with cost benefit analysis. It involves Monte Carlo simulation, considering the following main attributes: -

1. Consumption time associated with historical material consumption for corrective maintenance or planned/proactive maintenance

2. Consumption quantity per elapse time for planned maintenance 3. Consumption quantity per equipment failure rectification

4. Lead time for material restocking

5. Cost i.e. Material cost, holding cost, administrative cost and cost of production downtime due to unavailability of spare parts. This model focus on determining the stock level required to ensure that no stock-outs occur (at a defined level of reliability) over a selected interval of time, typically the time required to receive a component on site, after an order has been placed.

FIG.16 shows material consumption pattern associated with planned and corrective maintenance activities in an example scenario. FIG.16 shows the stocking level of a spare part 1610 over time. The numbers in circles 1 - 5 indicate holding time of spare parts in warehouse, up until stocks deplete below min level. The letters a - d indicate acquisition cost of purchasing spare parts once stocks level depletes below min. “Lead time” indicates spare parts restocking duration once stock level reach min. FIG.16 also shows the status 1620 of equipment A and the status 1630 to equipment B. Where the status 1620 and the status 1630 is high this indicates that the respective equipment is operational and when the status 1620 and the status 1630 is low, this indicates that the respective equipment is inoperable. The mark V indicates duration of equipment downtime due to stock out. TTF (time to failure) represents duration of equipment running until it fails.

Historical consumption quantity for all corrective and planned maintenance will be analyzed to anticipate quantity of consumption associated to probability of future demand for maintenance. For corrective maintenance, distribution will be based on consumed quantity/corrective maintenance. TTF will be calculated for data classified as Failure (F) only. For planned maintenance, distribution will be based on consumed quantity/elapse time. TTF will be calculated for data classified as Suspension (S) only. Both failure and suspensions are results coming from step 12 (Failure or Suspension Identification) of the method 10 shown in FIG.1.

The model of the present disclosure works by considering all cost element related to spare parts inventory. In general, these costs are (i) material cost, (ii) acquisition costs, (iii) inventory holding costs, and (iv) stock-out or shortage costs.

Material costs consider the purchase cost of the material themselves,

(i) material cost

Acquisition costs consider the ordering costs associated with the processing of a purchase, from creation to receipt.

The distribution will be identified on calculation of total acquisition cost as below:-

Here i is the index of the purchase and as shown in FIG.16 there are 4 purchase events (a to d).

Inventory holding costs are related to the costs of managing the inventory and are regularly expressed per item, per unit time. Inventory holding costs are a function of inventory on hand and it is commonly assumed that their value ranges between 20% and 40% of the value of the components stocked per year. Referring to FIG.16, distribution will be identified based on calculation of inventory holding cost for 5 event of material consumed until level depletes below min level will be as below:-

Finally, stock-out or shortage costs are incurred whenever demand cannot be routinely satisfied from inventory, due to lack of spares. In the maintenance environment, shortage costs are often large, if a stock-out of the component results in lost production or valuable downtime of a system or piece of equipment Shortage costs are also regularly expressed per item, per unit time.

(iv) = x * stock out cost/time

Loss of production is also accounted for in the calculation where cost of loss (RM) per day is being taken into consideration should the amount of spare simulated on hand is not able to satisfy the request for the downtime. The production loss per day cost is taken from company’s Equipment Criticality Assessment (EGA).

Total cost associated to a material for in a period of time is calculated as following: Total Cost (TC) = (i) + (ii) + (iii) + (iv)

Monte Carlo simulation is applied in this model to test all possibilities and come out with approximately 20 years of spare parts demand with associated cost, resulting in undoubted recommendation of spare parts optimum min and max level.

Optimization will be at lowest total cost. Availability will be used as additional metric to review. Optimum spare parts min and max level will be finalized only if Matching Spare Parts Availability is more than 99%.

As discussed above, embodiments of the present invention provide spare part inventory optimization based on the following:

1. Automation of Failure or Suspension identification via Text Analytics 2. Establishment of Failure library based on clustering of material and equipment as a reference for failure rates and probability of failure within a given time/period. Use cases to determine cluster type.

3. Ability to recommendation of non-moving spares where no consumption data is available by leveraging on failure rates dictionary.

4. Ability to recommend initial spare stocking from Original Equipment Manufacturer proposal by leveraging on failure rates dictionary.

5. Cost-benefit analysis of Inventory minimum and maximum level (Norms) based on plant downtime cost and double jeopardy situation leveraging on Equipment Criticality Assessment (EGA).

Whilst the foregoing description has described exemplary embodiments, it will be understood by those skilled in the art that many variations of the embodiments can be made within the scope and spirit of the present invention.

Claims

1. A method of processing equipment maintenance report data to optimize spare part inventory for equipment, the method comprising: identifying equipment maintenance reports relating to failed parts in the equipment report data; calculating a time-to-failure for each failed part in the equipment maintenance reports relating to failed parts and thereby generating time-to-failure data for a plurality of equipment-part combinations; for each equipment-part combination of the plurality of equipment-part combinations, fitting the time-to-failure data to a statistical distribution; estimating failure rates for each equipment-part combination of the plurality of equipment-part combinations using the respective statistical distributions; and optimizing spare part inventory for equipment using the estimated failure rates.

2. A method according to claim 1, wherein identifying equipment maintenance reports relating to failed parts in the equipment report data comprises classifying equipment maintenance reports as relating to failed parts corresponding to a scenario when a part is replaced due to failure or as relating to suspension corresponding to a scenario when a part is replaced prior to failure.

3. A method according to claim 2, wherein classifying equipment maintenance reports as relating to failed parts corresponding to a scenario when a part is replaced due to failure or as relating to suspension corresponding to a scenario when a part is replaced prior to failure comprises analyzing text of the maintenance reports.

4. A method according to claim 3, wherein analyzing text of the maintenance reports comprises determining if a word indicating failure is present in a maintenance report and classifying the maintenance report as relating to failed parts if a word indicating failure is present

5. A method according to any preceding claim wherein fitting the time-to-failure data to a statistical distribution comprises selecting a statistical distribution type from a plurality of statistical distribution types and estimating parameters for the selected statistical distribution type.

6. A method according to claim 5, wherein selecting a statistical distribution type comprises fitting the time-to-failure data to each of a plurality of statistical distribution types, determining an Akaike information criterion for each statistical distribution type and selecting the statistical distribution type with the lowest Akaike information criterion.

7. A method according to any preceding claim wherein estimating failure rates for each equipment-part combination of the plurality of equipment-part combinations using the respective statistical distributions comprises performing a Monte Carlo simulation on the respective statistical distributions.

8. A method according to any preceding claim wherein estimating failure rates for each equipment-part combination of the plurality of equipment-part combinations using the respective statistical distributions comprises grouping equipment-part combinations and estimating failure rates for the grouped equipment-part combinations.

9. A method of estimating an initial equipment spare part inventory requirement for a plant, the method comprising: identifying equipment maintenance reports relating to failed parts in the equipment report data; calculating a time-to-failure for each failed part in the equipment maintenance reports relating to failed parts and thereby generating time-to-failure data for a plurality of equipment-part combinations; for each equipment-part combination of the plurality of equipment-part combinations, fitting the time-to-failure data to a statistical distribution; estimating failure rates for each equipment-part combination of the plurality of equipment-part combinations using the respective statistical distributions; generating a failure library using the estimated failure rates; looking up estimated failure rates for equipment of the plant in the failure library; and estimating the initial spare part inventory requirement for the plant using the estimated failure rates.

10. A computer readable medium storing processor executable instructions which when executed on a processor cause the processor to carry out a method according to any one of claims 1 to 9.

11. A data processing system for processing equipment maintenance report data to optimize spare part inventory for equipment, the system comprising a processor and a data storage device storing computer program instructions operable to cause the processor to: identify equipment maintenance reports relating to failed parts in the equipment report data; calculate a time-to-failure for each failed part in the equipment maintenance reports relating to failed parts and thereby generating time-to-failure data for a plurality of equipment-part combinations; for each equipment-part combination of the plurality of equipment-part combinations, fit the time-to-failure data to a statistical distribution; estimate failure rates for each equipment-part combination of the plurality of equipment-part combinations using the respective statistical distributions; and optimize spare part inventory for equipment using the estimated failure rates.

12. A data processing system according to claim 11, wherein wherein the data storage device further stores computer program instructions operable to cause the processor to: identify equipment maintenance reports relating to failed parts in the equipment report data by classifying equipment maintenance reports as relating to failed parts corresponding to a scenario when a part is replaced due to failure or as relating to suspension corresponding to a scenario when a part is replaced prior to failure.

13. A data processing system according to claim 12, wherein the data storage device further stores computer program instructions operable to cause the processor to: classify the equipment maintenance reports as relating to failed parts corresponding to a scenario when a part is replaced due to failure or as relating to suspension corresponding to a scenario when a part is replaced prior to failure by analyzing text of the maintenance reports.

14. A data processing system according to claim 13, wherein the data storage device further stores computer program instructions operable to cause the processor to: analyze text of the maintenance reports by determining if a word indicating failure is present in a maintenance report and classifying the maintenance report as relating to failed parts if a word indicating failure is present

15. A data processing system according to any one of claims 11 to 14, wherein the data storage device further stores computer program instructions operable to cause the processor to: fit the time-to-failure data to a statistical distribution by selecting a statistical distribution type from a plurality of statistical distribution types and estimating parameters for the selected statistical distribution type.

16. A data processing system according to claim 15, wherein the data storage device further stores computer program instructions operable to cause the processor to: select a statistical distribution type by fitting the time-to-failure data to each of a plurality of statistical distribution types, and determine an Akaike information criterion for each statistical distribution type and select the statistical distribution type with the lowest Akaike information criterion.

17. A data processing system according to any one of claims 11 to 16, wherein the data storage device further stores computer program instructions operable to cause the processor to: estimate failure rates for each equipment-part combination of the plurality of equipment-part combinations using the respective statistical distributions by performing a Monte Carlo simulation on the respective statistical distributions.

18. A data processing system according to any one of claims 11 to 17, wherein the data storage device further stores computer program instructions operable to cause the processor to: estimate failure rates for each equipment-part combination of the plurality of equipment-part combinations using the respective statistical distributions by grouping equipment-part combinations and estimating failure rates for the grouped equipment-part combinations.

19. A data processing system for estimating an initial equipment spare part inventory requirement for a plant, the system comprising a processor and a data storage device storing computer program instructions operable to cause the processor to: identify equipment maintenance reports relating to failed parts in the equipment report data; calculate a time-to-failure for each failed part in the equipment maintenance reports relating to failed parts and thereby generating time-to-failure data for a plurality of equipment-part combinations; for each equipment-part combination of the plurality of equipment-part combinations, fit the time-to-failure data to a statistical distribution; estimate failure rates for each equipment-part combination of the plurality of equipment-part combinations using the respective statistical distributions; generate a failure library using the estimated failure rates; look up estimated failure rates for equipment of the plant in the failure library; and estimate the initial spare part inventory requirement for the plant using the estimated failure rates.