Improving future reliability predictions based on system operation and performance data modeling

Abstract

The present application relates to improving future reliability predictions based on system operation and performance data modeling. More specifically, to systems, methods, and apparatus for improving future reliability predictions of a measurable system by receiving operation and performance data, such as maintenance cost data, first principles data, and asset reliability data, via an input interface associated with the measurable system. A plurality of category values can be generated that categorize the maintenance cost data at specified intervals using maintenance criteria generated from one or more comparative analysis models associated with the measurable system. An estimated future reliability of the measurable system is determined based on the asset reliability data and the plurality of category values, and a result of the future reliability is displayed on an output interface.

Description

Improving future reliability predictions through system operation and performance data modeling

[0001] This application is a divisional application of Chinese patent application 201580027842.9, filed on April 11, 2015, entitled "Improving Future Reliability Prediction Based on System Operation and Performance Data Modeling".

[0002] Cross-reference to related applications

[0003] This application claims the benefit and priority of U.S. Provisional Patent Application Serial No. 61 / 978,683, filed April 11, 2014, entitled “System and Method for the Estimation of Future Reliability Based on Historical Maintenance Spend,” the disclosure of which is incorporated herein by reference in its entirety.

[0004] Statement regarding federally sponsored research or development

[0005] not applicable.

[0006] References to the appendix on microfilm

[0007] not applicable. Technical Field

[0008] This disclosure generally relates to the field of modeling and predicting the future reliability of measurable systems based on operational and performance data, such as current and historical data regarding production and / or costs associated with maintenance equipment. More specifically, but not as a limitation, embodiments in this disclosure perform comparative performance analyses and / or determine model coefficients used to model and estimate the future reliability of one or more measurable systems. Background Technology

[0009] Generally, for repairable systems, there is a general correlation between the methods and processes used to maintain the repairable system and the future reliability of the system. For example, individuals who own or operate bicycles, motor vehicles, and / or any other transport vehicles are generally aware that the operating conditions and reliability of the transport vehicle can depend to some extent on the degree and quality of maintaining the transport vehicle's operation. However, while a correlation may exist between maintenance quality and future reliability, quantifying and / or modeling this relationship can be difficult. Similar relationships and / or correlations can also exist for a wide variety of measurable systems, beyond repairable systems, where operational and / or performance data is available or otherwise used to evaluate the system.

[0010] Unfortunately, the value or amount of maintenance expenditure may not be a precise indicator for predicting the future reliability of a repairable system. Individuals can accumulate maintenance costs on tasks that have a relatively minimal impact on improving future reliability. For example, excessive maintenance expenditure may stem from actual system failures rather than tasks related to preventative maintenance. Generally, system failures, crashes, and / or unplanned maintenance will cost more than preventative and / or predictive maintenance items using a comprehensive maintenance plan. Accordingly, improvements are needed to enhance the accuracy of modeling and predicting the future reliability of measurable systems. Summary of the Invention

[0011] The following is a brief overview of the disclosed subject matter to provide a basic understanding of some aspects of the subject matter disclosed herein. This overview is not an exhaustive summary of the technology disclosed herein. It is not intended to identify key or essential elements of the invention or to outline the scope of the invention. Its sole purpose is to present some concepts in a simplified form as a prelude to the more detailed description discussed later.

[0012] In one embodiment, a system for modeling the future reliability of a facility based on operational and performance data includes an input interface configured to: receive maintenance cost data corresponding to the facility; receive first-principle data corresponding to the facility; and receive asset reliability data corresponding to the facility. The system may also include a processor coupled to a non-transitory computer-readable medium, wherein the non-transitory computer-readable medium includes instructions that, when executed by the processor, cause the apparatus to: obtain one or more comparative analysis models associated with the facility; obtain maintenance criteria that generate multiple category values, which classify the maintenance cost data at specified intervals based at least on the maintenance cost data, the first-principle data, and the one or more comparative analysis models; and determine an estimated future reliability of the facility based on the asset reliability data and the multiple category values. The computer node may also include a user interface for displaying the results of the future reliability assessment.

[0013] In another embodiment, a method for modeling the future reliability of a measurable system based on operational and performance data includes: receiving maintenance cost data via an input interface associated with the measurable system; receiving first-principle data via an input interface associated with the measurable system; receiving asset reliability data via an input interface associated with the measurable system; generating multiple category values ​​using a processor, which classify the maintenance cost data at specified intervals using maintenance criteria, wherein the maintenance criteria are generated from one or more comparative analysis models associated with the measurable system; determining an estimated future reliability of the measurable system using the processor based on the asset reliability data and the multiple category values; and outputting the estimated future reliability results using an output interface.

[0014] In another embodiment, an apparatus for modeling the future reliability of equipment assets based on operational and performance data includes: an input interface including a receiving device configured to: receive maintenance cost data corresponding to the equipment assets; receive first principle data corresponding to the equipment assets; receive asset reliability data corresponding to the equipment assets; a processor coupled to a non-transitory computer-readable medium, wherein the non-transitory computer-readable medium includes instructions that, when executed by the processor, cause the apparatus to: generate a plurality of category values ​​that classify the maintenance cost data according to maintenance criteria at specified intervals; and determine an estimated future reliability of a facility including estimated future reliability data based on the asset reliability data and the plurality of category values; and an output interface including a transmission device configured to transmit a processed dataset including the estimated future reliability data to a control center for comparing different equipment assets based on the processed dataset. Attached Figure Description

[0015] Figure 1 is a flowchart of an embodiment of a data analysis method that receives data from one or more various data sources associated with a measurable system (such as a power plant);

[0016] Figure 2 is a schematic diagram of an embodiment of the data compilation table generated in the data compilation of the data analysis method described in Figure 1;

[0017] Figure 3 is a schematic diagram of an embodiment of a classification maintenance table generated from time-based maintenance data of the data analysis method described in Figure 1;

[0018] Figure 4 is a schematic diagram of an embodiment of a classified reliability table generated from time-based reliability data using the data analysis method described in Figure 1.

[0019] Figure 5 is a schematic diagram of an embodiment of the future reliability data table generated in the future reliability prediction of the data analysis method described in Figure 1;

[0020] Figure 6 is a schematic diagram of an embodiment of the future reliability statistics table generated in the future reliability prediction of the data analysis method described in Figure 1;

[0021] Figure 7 is a schematic diagram of an embodiment of a user interface input screen configured to display information that a user may need to input in order to determine future reliability predictions using the data analysis method described in Figure 1;

[0022] Figure 8 is a schematic diagram of an embodiment of the user interface input screen configured for EFOR prediction using the data analysis method described in Figure 1;

[0023] Figure 9 is a schematic diagram of an embodiment of a computing node for implementing one or more embodiments;

[0024] Figure 10 is a flowchart of an embodiment of a method for determining model coefficients used in comparative performance analysis of measurable systems, such as power plants;

[0025] Figure 11 is a flowchart of an embodiment of a method for determining the main first-principles characteristics as shown in Figure 10;

[0026] Figure 12 is a flowchart of an embodiment of a method for developing a method for solving constraints used in the comparative analysis model shown in Figure 10;

[0027] Figure 13 is a schematic diagram of an embodiment for determining the model coefficient matrix as shown in Figures 10-12;

[0028] Figure 14 is a schematic diagram of an embodiment of the model coefficient matrix for determining the model coefficients used in the comparative performance analysis shown in Figures 10-12 for the fluidized catalytic cracking unit (Cat Cracker).

[0029] Figure 15 is a schematic diagram of an embodiment for determining the model coefficient matrix for pipelines and tank farms used in the comparative performance analysis shown in Figures 10-12;

[0030] Figure 16 is a schematic diagram of another embodiment of a computing node used to implement one or more embodiments.

[0031] While certain embodiments will be described in conjunction with preferred illustrative embodiments shown herein, it should be understood that this is not intended to limit the invention to those embodiments. Rather, it is intended to cover all substitutions, modifications, and equivalents that may be included within the spirit and scope of the invention as defined by the claims as included in this disclosure. In the non-total-scale drawings, the same reference numerals are used throughout the description and in the drawings for parts and elements having the same structure, and superscript reference numerals are used for parts and elements having similar function and structure to parts and elements having the same reference numerals without superscript reference numerals. Detailed Implementation

[0032] It should be understood that while illustrative implementations of one or more embodiments are provided below, various specific embodiments can be implemented using any number of techniques known to those skilled in the art. This disclosure should not be limited in any way to the illustrative embodiments, drawings, and / or techniques shown below, including the exemplary designs and implementations shown and described herein. Furthermore, this disclosure can be modified within the full scope of the appended claims, together with their equivalents.

[0033] This document discloses one or more embodiments for estimating the future reliability of a measurable system. Specifically, the one or more embodiments can obtain model coefficients used in comparative performance analysis by determining one or more target variables and one or more characteristics for each target variable. The target variables can represent different parameters for the measurable system. The characteristics of the target variables can be collected and ordered according to data collection classification. This data collection classification can be used to quantitatively measure differences in the characteristics. After collecting and validating the data, a comparative analysis model can be developed to compare the predicted target variables with the actual target variables for one or more measurable systems. The comparative analysis model can be used to obtain a set of complexity factors that attempt to minimize the differences between the predicted and actual target variable values ​​within the model. The comparative analysis model can then be used to develop representative values ​​for activities performed periodically on the measurable system to predict future reliability.

[0034] Figure 1 is a flowchart of an embodiment of a data analysis method 60 that receives data from one or more various data sources associated with a measurable system (such as a power plant). The data analysis method 60 can be implemented by a user, a computing node, or a combination thereof to estimate the future reliability of the measurable system. In one embodiment, the data analysis method 60 can automatically receive updated available data (such as updated operational and performance data) from various data sources, update one or more comparative analysis models using the received updated data, and subsequently provide an update to the estimate of the future reliability of one or more measurable systems. A measurable system is any system associated with performance data, conditioned data, operational data, and / or other types of measurable data (e.g., quantitative and / or qualitative data) used to assess the state of a system. For example, a measurable system may be monitored using various parameters and / or performance factors associated with one or more components of the measurable system, such as in a power plant, facility, or commercial building. In another embodiment, a measurable system may be associated with available performance data such as stock prices, security records, and / or corporate finances. The terms “measurable system,” “facility,” “asset,” or “plant” are used interchangeably throughout this disclosure.

[0035] As shown in Figure 1, data from various data sources can be applied at different computational stages to model and / or improve future reliability predictions based on data available for the measurable system. In one embodiment, the available data can be current and historical maintenance data related to one or more measurable parameters of the measurable system. For example, in terms of maintenance and repairable equipment, one way to describe the quality of maintenance is to calculate the annual or periodic maintenance costs of a measurable system (such as equipment assets). Annual or periodic maintenance figures represent the amount spent over a given period of time, which may not necessarily accurately reflect future reliability. For example, a car owner may spend money weekly to wash and clean their vehicle, but spend relatively little or no money on maintenance that could potentially improve the future reliability of the car, such as changing tires and / or oil or filter changes. While the annual maintenance cost for washing and cleaning a car can be a considerable figure when performed frequently, the maintenance task and / or activity of washing and cleaning may have a relatively small or no impact on improving the reliability of the car.

[0036] Figure 1 illustrates how data analysis method 60 can be used to predict the future equivalent forced outage rate (EFOR) estimate for power plants based on the Rankine and Brayton cycles. EFOR is defined as the number of hours of unit failure (e.g., unplanned outage hours and equivalent unplanned degradation hours) given as a percentage of the total hours of unit availability (e.g., unplanned outage, unplanned degradation, and service hours). As shown in Figure 1, during the first data collection phase, data analysis method 60 can initially obtain asset maintenance cost data 62 and asset unit first-principles data or other asset-level data 64 related to measurable data systems (such as power plants). Asset maintenance cost data 62 for various facilities can typically be obtained directly from the plant facilities. Asset maintenance cost data 62 can represent the costs associated with maintaining the measurable system over a specified time period (e.g., in seconds, minutes, hours, months, and years). For example, asset maintenance cost data 62 can be annual or periodic maintenance costs for one or more measurable systems. Asset unit first-principle data or other asset-level data 64 can represent the physical or fundamental characteristics of a measurable system. For example, asset unit first-principle data or other asset-level data 64 can be operational and performance data corresponding to one or more measurable systems, such as turbine inlet temperature, asset age, size, horsepower, amount of fuel consumed, and actual power output compared to the nameplate.

[0037] Data acquired during the first data collection phase can then be received or input to generate maintenance criterion 66. In one embodiment, maintenance criterion 66 may be an annualized maintenance criterion, wherein the user provides in advance one or more modeling equations for calculating the annualized maintenance criterion. The results can be used to normalize asset maintenance cost data 62 and provide a benchmark indication to measure expenditure appropriateness relative to other power plants of similar types. In one embodiment, the divisor or criterion may be calculated based on first-principles data of the asset unit or other asset-level data 104, as explained in more detail in Figures 10-12. Alternative embodiments may generate maintenance criterion 66, for example, from a simple regression analysis of data from available plant-related target variables.

[0038] Maintenance costs for replacing components that typically wear out over time can occur at different time intervals, resulting in variations in periodic maintenance costs. To address this potential problem, data analysis method 60 can generate maintenance criteria 66 to develop representative values ​​for periodic maintenance activities. For example, to generate maintenance criteria 66, data analysis method 60 can normalize maintenance costs to a different time period. In another embodiment, data analysis method 60 can generate a periodic maintenance expenditure divisor to normalize actual periodic maintenance expenditures. This measures expenditures that are too low (actual cost / divisor ratio < 1) or too high (actual cost / divisor ratio > 1). The maintenance expenditure divisor can be a value calculated from the data using asset maintenance cost data 62, asset unit first-principles data, or other asset-level data 64 (e.g., asset characteristics) and / or recorded expert opinions. In this embodiment, asset unit first-principles data or other asset-level data 64 (such as plant size, plant type, and / or plant output) can be used, in conjunction with the calculated annualized maintenance costs, to calculate the standard maintenance cost (divisor) value for each asset in the analysis, as described in U.S. Patent 7,233,910, filed July 18, 2006, entitled "System and Method for Determining Equivalency Factors for use in Comparative Performance Analysis of Industrial Facilities," which is incorporated herein by reference as if reproduced in its entirety. The calculation can be performed using historical datasets, including those based on the assets currently being analyzed. The maintenance standard calculation can be applied as a model comprising one or more equations for modeling future reliability predictions of a measurable system. The data used to calculate the maintenance standard divisor can be provided by a user, transmitted from a remote storage device, and / or received via a network from a remote network node (such as a server or database).

[0039] Figure 1 illustrates that data analysis method 60 may receive asset reliability data 400 during the second data collection phase. Asset reliability data 70 may correspond to each measurable system. Asset reliability data 70 is any data corresponding to determining the reliability, failure rate, and / or unexpected downtime of a measurable system. Once data analysis method 60 receives the asset reliability data 70 for each measurable system, it can be compiled and linked to the maintenance expenditure ratio of the measurable system, which can be correlated with other measurable system- and time-specific data or displayed on the same line. For power plants, asset reliability data 70 can be obtained from the Electric Reliability Corporation's Generation Availability Database (NERC-GADS). Other types of measurable systems may also obtain asset reliability data 70 from similar databases.

[0040] During data compilation 68, data analysis method 60 compiles the calculated maintenance standard 66, asset maintenance cost data 62, and asset reliability data 70 into a common file. In one embodiment, data analysis method 60 may add additional columns to the data layout in the common file. These additional columns may represent the ratio of the actual annualized maintenance cost for each measurable system to the calculated standard value. Data analysis method 60 may also add another column during data editing 68, which categorizes the maintenance expenditure ratio by dividing it by some percentile intervals or categories. For example, data analysis method 60 may use nine different intervals or categories to categorize the maintenance expenditure ratio.

[0041] In the categorized time-based maintenance data 72, data analysis method 60 can place maintenance category values ​​into a matrix (such as a 2×2 matrix) defining each measurable system (such as a power plant and time unit). In the categorized time-based reliability data 74, data analysis method 60 uses the same matrix structure described in the categorized time-based maintenance data 72 to assign reliability for each measurable system. In future reliability prediction 76, data from the categorized time-based maintenance data 72 and the time-based categorized reliability data 74 are statistically analyzed to calculate averages and / or other statistical calculations to determine the future reliability of the measurable system. The calculated future time period or number of years can be a function of available data, such as asset maintenance cost data 62, asset reliability data 70, and asset unit first-principles data or other asset-level data. For example, the future interval can be one year in advance due to available data, but other embodiments can rely on the available dataset to use a choice of two or three years in the future. Moreover, depending on the granularity of the available data, other embodiments can use time periods other than years, such as seconds, minutes, hours, days, and / or months.

[0042] It should be noted that while the discussion relating to Figure 1 is specific to power plants and that industry, data analysis method 60 can also be applied to other industries where similar maintenance and reliability databases exist. For example, in the refining and petrochemical industries, maintenance and reliability data for refineries and / or other measurable systems have existed for many years. Therefore, data analysis method 60 can also use the ratio of current to previous year's maintenance expenditures to predict the future reliability of refineries and / or other measurable systems. Other embodiments of data analysis method 60 can also be applied to the pipeline industry as well as the maintenance of buildings (e.g., office buildings) and other structures.

[0043] Those skilled in the art will recognize that other industries may use various metrics or parameters for asset reliability data 70 that differ from the power industry EFOR metric applied in Figure 1. For example, other suitable asset reliability data 70 that may be used in data analysis method 60 include, but are not limited to, “unavailability,” “availability,” “commercial unavailability,” and “mean time between failures.” These metrics or parameters may have definitions often specific to a given situation, but their general interpretation is known to those skilled in the art of reliability analysis and reliability prediction.

[0044] Figure 2 is a schematic diagram of an embodiment of a data compilation table 250 generated in the data compilation 68 of the data analysis method 60 described in Figure 1. The data compilation table 250 can be displayed using an output interface (such as a graphical user interface) or sent to an output interface (such as a printing device). Figure 2 shows that the data compilation table 250 includes a client number column 252 indicating the asset owner, a plant name column 254 indicating the measurable system and / or the plant from which data is being collected, and a study year column 256. As shown in Figure 2, each asset owner in table 200 owns a single measurable system. In other words, each measurable system is owned by a different asset owner. Other embodiments of the data compilation table 250 may have multiple measurable systems owned by the same asset owner. The study year column 256 refers to the time period from which data is collected or analyzed from the measurable system.

[0045] Data compilation table 250 may include additional columns calculated using data analysis method 60. The calculated maintenance (Mx) standard column 258 may include data values ​​representing the calculated results of the maintenance standards described in maintenance standard 66 as shown in FIG1. ​​Recall that in one embodiment, maintenance standard 66 may be generated as described in U.S. Patent 7,233,910. Other embodiments may calculate the results of maintenance standards known to those skilled in the art. The actual annualized Mx cost column 260 may include calculated data values ​​representing normalized actual maintenance data based on the maintenance standards described in maintenance standard 66 as shown in FIG1. ​​Actual maintenance data may be effective annual costs over several years (e.g., approximately 5 years). The ratio actual (Act)Mx / Std (Std)Mx column 262 may include data values ​​representing a normalized maintenance expenditure ratio used to assess the relationship between the appropriateness or effectiveness of maintenance expenditures and future reliability. The last column (EFOR column 266) includes data values ​​representing the reliability for the current time period, or in this example, the unreliability value. The data value in EFOR column 266 is the sum of unplanned downtime and degraded hours divided by the hours in the operating period. In this example, the definition of EFOR follows the notation described in the NERC-GADS literature. For example, an EFOR value of 9.7 indicates that the measurable system was actually down for approximately 9.7% of its operating time due to unplanned outage events.

[0046] Act Mx / Std Mx: Decimal column 264 can include data values ​​representing maintenance expenditure ratios that are categorized into intervals of values ​​related to the different ranges discussed in Data Compilation 68 in Figure 1. Two decimals, decimals, quartiles, quintiles, or quartiles can be used; however, in this example, the data is divided into nine categories based on the percentile ranking of the maintenance expenditure ratio data values ​​found in Act Mx / Std Mx column 262. The number of intervals or categories used to divide the maintenance expenditure ratio can depend on the dataset size, where statistically possible more detailed divisions can be generated for relatively large dataset sizes. Various methods or algorithms known to those skilled in the art can be used to determine the number of intervals based on the dataset size. The conversion of maintenance expenditure ratios to ordered categories can serve as a reference for assigning future EFOR reliability values ​​to actual implementations.

[0047] Figure 3 is a schematic diagram of an embodiment of a categorized maintenance table 350 generated from time-based maintenance data 72 categorized by the data analysis method 60 described in Figure 1. The categorized maintenance table 350 can be displayed using an output interface (such as a graphical user interface) or sent to an output interface (such as a printing device). Specifically, the categorized maintenance table 350 is a transformation of the maintenance expenditure ratio ordinal category data values ​​found in the data compilation table 250 of Figure 2. Figure 3 shows that the plant name column 352 can identify different measurable systems. The year columns 354-382 represent different years or time periods for each measurable system. Using Figure 3 as an example, plants 1 and 2 have data values ​​from 1999 to 2013, and plants 3 and 4 have data values ​​from 2002 to 2013. The type of data found in the year columns 354-382 is substantially similar to the type of data in the Act Mx / Std Mx: decimal column 264 in Figure 2. Specifically, the data types in the year columns 354-382 represent intervals related to different ranges of maintenance expenditure ratios and are generally referred to as maintenance expenditure ratio ordinal categories. For example, for 1999, Plant 1 has a maintenance expenditure ratio classified as "5", and Plant 2 has a maintenance expenditure ratio classified as "1".

[0048] Figure 4 is a schematic diagram of an embodiment of a categorized reliability table 400 generated from the categorized time-based reliability data 74 of the data analysis method 60 described in Figure 1. The categorized reliability table 400 can be displayed using an output interface (such as a graphical user interface) or sent to an output interface (such as a printing device). The categorized reliability table 400 is a transformation of the EFOR data values ​​found in the data compilation table 250 of Figure 2. Figure 4 shows that the plant name column 452 can identify different measurable systems. Year columns 404-432 represent the different years used for each measurable system. Using Figure 4 as an example, plants 1 and 2 have data values ​​from 1999 to 2013, and plants 3 and 4 have data values ​​from 2002 to 2013. The type of data found in year columns 354-382 is substantially similar to the type of data in EFOR column 266 in Figure 2. Specifically, the type of data in year columns 354-382 represents EFOR values ​​indicating the percentage of unplanned outage events. For example, in 1999, Plant 1 had an EFOR of 2.4, which indicates that Plant 1 was out of service for approximately 2.4% of its operating time due to unplanned outage events, and Plant 2 had an EFOR of 5.5, which indicates that Plant 2 was out of service for approximately 5.5% of its operating time due to unplanned outage events.

[0049] Figure 5 is a schematic diagram of an embodiment of the future reliability data table 500 generated in the future reliability prediction 76 of the data analysis method 60 described in Figure 1. The future reliability data table 500 can be displayed using an output interface (such as a graphical user interface) or sent to an output interface (such as a printing device). The process of calculating future reliability begins with selecting a future reliability interval, for example, approximately two years in Figure 5. After selecting the future reliability interval, the data shown in Figure 3 is scanned horizontally or row-by-row in a categorized maintenance table 350 to identify rows that are separated by only approximately one year. For example, using Figure 3, the row associated with plant 1 will satisfy the approximately one-year data separation, but plant 11 will not, because plant 11 has a data gap between 2006 and 2008 in the categorized maintenance table 350. In other words, plant 11 lacks data for 2007, and therefore, the entries for plant 11 are not separated by approximately one year. Other embodiments may use different time intervals, measured in seconds, minutes, hours, days, and / or months, to select the future reliability interval. The time interval used to determine future reliability depends on the level of data granularity.

[0050] The maintenance expenditure ratio ordinal category used for each separated row can then be paired with the time-forward EFOR value from the categorized reliability data table 400 to form ordered pairs. The resulting ordered pairs consist of the maintenance expenditure ratio ordinal category and the time-forward EFOR value. Since the selected future reliability interval is approximately two years, the years associated with the maintenance expenditure ratio ordinal category and EFOR value in the generated ordered pairs can be two years apart. Some examples of analyzing these ordered pairs approximately two years in advance for the same plant or the same row are:

[0051] First ordinal pair (ordinal category of maintenance expenditure ratio in 1999, EFOR value in 2001)

[0052] Second ordinal pair: (2000 maintenance expenditure ratio ordinal category, 2002 EFOR value)

[0053] Third ordinal pair: (2001 maintenance expenditure ratio ordinal category, 2003 EFOR value)

[0054] Fourth order pair: (Ordinal category of maintenance expenditure ratio in 2002, EFOR value in 2004)

[0055] As shown above, in each order pair, the years separating the maintenance expenditure ratio ordinal category and the EFOR value are based on a future reliability interval, which is approximately two years. To form the order pairs, the metrics in Figures 3 and 4 can be scanned to find possible data pairs separated by two years (e.g., 1999 and 2001). In this case, the intermediate year data (e.g., 2000) is not used for the data pair. This process can be repeated for other future reliability intervals (e.g., one year prior to the user-determined maintenance ratio ordinal value and expected information from the analysis). Furthermore, the order pair examples above plot the maintenance expenditure ratio ordinal category and EFOR value pair incrementing by one for each order pair. For example, the first order pair has the maintenance expenditure ratio ordinal category in 1999, and the second order pair has the maintenance expenditure ratio ordinal category in 2000.

[0056] Different maintenance expenditure ratio ordinal category values ​​are used to place the corresponding forward EFOR values ​​into the correct columns within the future reliability data table 500. As shown in Figure 5, column 502 includes EFOR values ​​with maintenance expenditure ratio ordinal category "1"; column 504 includes EFOR values ​​with maintenance expenditure ratio ordinal category "2"; column 506 includes EFOR values ​​with maintenance expenditure ratio ordinal category "3"; column 508 includes EFOR values ​​with maintenance expenditure ratio ordinal category "4"; column 510 includes EFOR values ​​with maintenance expenditure ratio ordinal category "5"; column 512 includes EFOR values ​​with maintenance expenditure ratio ordinal category "6"; column 514 includes EFOR values ​​with maintenance expenditure ratio ordinal category "7"; column 516 includes EFOR values ​​with maintenance expenditure ratio ordinal category "8"; and column 518 includes EFOR values ​​with maintenance expenditure ratio ordinal category "9".

[0057] Figure 6 is a schematic diagram of an embodiment of a future reliability statistics table 600 generated in the future reliability prediction 76 of the data analysis method 60 described in Figure 1. The future reliability statistics table 600 can be displayed using an output interface (such as a graphical user interface) or sent to an output interface (such as a printing device). In Figure 6, the future reliability statistics table 600 includes maintenance expenditure ratio ordinal category columns 602-618. As shown in Figure 6, each column in the maintenance expenditure ratio ordinal category columns 602-618 corresponds to a maintenance expenditure ratio ordinal category. For example, maintenance expenditure ratio ordinal category column 602 corresponds to maintenance expenditure ratio ordinal category "1", and maintenance expenditure ratio ordinal category column 604 corresponds to maintenance expenditure ratio ordinal category "2". The compiled data in each maintenance ratio ordinal value column 602-618 is analyzed using data from the future reliability data table 500 to calculate various statistics indicating future reliability information. As shown in Figure 6, rows 620, 622, and 624 represent the mean, median, and 90th percentile distribution values ​​used for the future reliability data for each maintenance ratio ordinal value. In Figure 6, future reliability information is interpreted as a forecast of future reliability, or EFOR, for a measurable system with a specific maintenance expenditure ratio in the current year.

[0058] Future EFOR forecasts can be calculated using the maintenance expenditure ratios from the current and previous years. For multi-year scenarios, the maintenance expenditure ratio is calculated by summing the annualized costs for those years and then dividing by the sum of the maintenance standards used for the previous years. In this way, the expenditure ratio reflects performance relative to a general standard over several years, which is a summation of standards calculated for each included year.

[0059] Figure 7 is a schematic diagram of an embodiment of a user interface input screen 700, configured to display information that a user might need to input to determine future reliability predictions 76 using the data analysis method 60 described in Figure 1. The user interface input screen 700 includes a measurable system selection column 702, which the user can use to select the type of measurable system. Using Figure 7 as an example, the user can select a “Coal-Rankine” plant as the type of generator set or measurable system. Other options shown in Figure 7 include “Gas-Rankine” and “Combustion Turbine”. Once the type of measurable system is selected, the user interface input screen 700 can generate the required data items 704 associated with the type of measurable system selected by the user. The data items 704 appearing in the user interface input screen 700 can vary depending on the measurable system selected in the measurable system selection column 702. Figure 7 shows that the user has selected a Coal-Rankine plant and the user can enter all fields displayed as blank spaces with underlines. This may also include annualized maintenance costs for a specific year. In other embodiments, blank fields can be entered using information received from a remote data storage device or via a network. If desired, the current model also allows the user to enter data from the previous year to add more information for future reliability predictions. Other embodiments may import and obtain additional information from storage media or via a network.

[0060] Once this information is entered, the calculation fields 706 at the bottom of the user interface input screen 700 (such as the Annual Maintenance Standard ($) field and the Risk Modification Factor field) can be automatically populated based on the user's input. The Annual Maintenance Standard ($) field can be calculated in a manner substantially similar to the MX Standard 258 shown in Figure 6. The Risk Modification Factor field can represent the overall risk modification factor used for comparative analysis models and can be the ratio of the calculated average EFOR for the next year to the overall average EFOR. In other words, the automatically generated data result in the Risk Modification Factor field represents the relative reliability risk of a specific measurable system compared to the overall average.

[0061] Figure 8 is a schematic diagram of an embodiment of a user interface input screen 800 configured for EFOR prediction using the data analysis method 60 described in Figure 1. In Figure 8, several results are presented for the user to consider. Curve 802 is a ranking curve representing the distribution of the maintenance expenditure ratio, and triangle 804 on curve 802 indicates the location of the currently measurable system or the measurable system the user is considering (e.g., the “Coal-Rankine” plant selected in Figure 7). The user interface input screen 800 shows the user the range of known performance and where the specific measurable system under consideration falls within that range. The numbers below this curve are the quintile values ​​of the maintenance expenditure ratio, where the maintenance expenditure ratio is categorized into five distinct value intervals. The data results shown in Figure 8 in this embodiment are calculated using the quintiles; however, other divisions are possible based on the amount of data available and the goals of the analyst and user.

[0062] Histogram 806 represents the average one-year future EFOR dependent on the specific quintiles to which the maintenance expenditure ratio falls. For example, the lowest one-year future EFOR occurs for plants with a maintenance expenditure ratio in the second quintile, or with a maintenance expenditure ratio of approximately 0.8 and approximately 0.92. This expenditure level indicates that the unit is successfully managing its assets in accordance with better practices to ensure long-term reliability. It should be noted that plants in the first quintile, or with a maintenance expenditure ratio of approximately zero to approximately 0.8, actually exhibit higher EFOR values, suggesting that the operator is not performing the necessary or sufficient maintenance to generate long-term reliability. If a plant falls into the fifth quintile, one explanation for this is that the operator may be overspending due to breakdowns. A high maintenance expenditure ratio produces a high EFOR value because the cost of maintenance from unplanned maintenance events is greater than the cost of planned maintenance.

[0063] Dotted line 810 represents the average EFOR used for analysis of all data on the currently measurable system. Diamond 812 represents the actual one-year future EFOR estimate positioned above triangle 804, which represents the maintenance expenditure ratio. These two symbols associate or connect the current maintenance expenditure level (triangle 804) to the one-year future EFOR estimate (diamond 812).

[0064] Figure 10 is a flowchart of an embodiment of a method 100 for determining model coefficients used in comparative performance analysis of a measurable system, such as a power plant. Method 100 can be used to generate one or more comparative analysis models used in the maintenance standard 66 described in Figure 1. Specifically, method 100 determines useful characteristics and model coefficients associated with one or more comparative analysis models that demonstrate the correlation between maintenance quality and future reliability. Method 100 can be implemented using a user and / or a computing node configured to receive input data for determining the model coefficients. For example, the computing node can automatically receive data and update the model coefficients based on received updated data.

[0065] Method 100 begins at step 102 and selects one or more target variables (“Target Variables”). Target variables are quantifiable attributes associated with a measurable system, such as total operating costs, financial results, capital costs, operating costs, staffing, product output, emissions, energy consumption, or any other quantifiable performance attribute. Target variables can be found in manufacturing, refining, chemical (including petrochemical, organic and inorganic chemistry, plastics, agricultural chemicals, and pharmaceuticals), olefin plants, chemical manufacturing, pipelines, power generation, power distribution, and other industrial facilities. Other embodiments of target variables can also be used for various environmental aspects, building and other structural maintenance, and other forms and types of industrial and commercial sectors.

[0066] In step 104, method 100 identifies first-principle characteristics. First-principle characteristics are physical or fundamental properties of the measurable system or process for which the target variable is expected. In one embodiment, a first-principle characteristic may be first-principle data of an asset unit as shown in Figure 1 or other asset-level data 64. Common brainstorming or team knowledge management techniques may be used to develop a first list of possible characteristics for the target variable. In one embodiment, all characteristics of an industrial facility that might cause changes in the target variable when compared to different measurable systems (such as industrial facilities) are identified as first-principle characteristics.

[0067] In step 106, method 100 determines the principal first-principal characteristic from all the first-principal characteristics identified in step 104. As those skilled in the art will understand, many different options are available for determining the principal first-principal characteristic. One such option is shown in Figure 11, which will be discussed in more detail below. Method 100 then moves to step 108 to classify the principal characteristic. Potential classifications for principal characteristics include discrete, continuous, or ordinal. Discrete characteristics are those that can be measured using a choice between two or more states, such as a binary determination like "yes" or "no". An example discrete characteristic could be "repeating equipment". The determination of "repeating equipment" is "yes, the equipment has repeating equipment" or "no, there is no repeating equipment". Continuous characteristics are those that are directly measurable. An example of a continuous characteristic could be "feed capacity" because it is directly measured as a continuous variable. Ordinal characteristics are those that are not easily measurable. Instead, ordinal characteristics can be scored along an ordinal scale that reflects physical differences that cannot be directly measured. It is also possible to create ordinal characteristics for measurable or binary variables. An example of ordinal characteristics would be the refinery configuration among three typical major industrial options.

[0068] These are given in ordinal scales in terms of unit complexity:

[0069] 1.0 Atmospheric distillation

[0070] 2.0 Catalytic Cracking Unit

[0071] 3.0 Coking Unit

[0072] The measurable systems listed above are ranked based on ordinal variables and generally do not include information about any quantifiable measurement quality. In the examples above, the difference in complexity between a 1.0 measurable system or atmospheric distillation unit and a 2.0 measurable system or catalytic cracking unit is not necessarily equal to the difference in complexity between a 3.0 measurable system or coking unit and a 2.0 measurable system or catalytic cracking unit.

[0073] Variables placed on ordinal scales can be converted to interval scales for the development of model coefficients. The conversion from ordinal to interval variables can use scales developed to describe differences between units on measurable scales. The process of developing interval scales for ordinal characteristic data can rely on a team of experts' scientifically driven understanding of the characteristics. The expert team can first determine the type of relationship between different physical characteristics and the target variable based on their understanding of the measured process and scientific principles. The relationship can be linear, logarithmic, power function, quadratic function, or any other mathematical relationship. Then, the experts can optionally estimate a complexity factor to reflect the relationship between variations in the characteristics and the target variable. The complexity factor can be an exponential power used to make the relationship between the ordinal and target variables linear, thus resulting in an interval variable scale. Furthermore, in the absence of data, the determination of key characteristics can be based on expert experience.

[0074] In step 110, method 100 can develop a data collection classification layout. Method 100 can quantify characteristics classified as continuous, ensuring data is collected in a consistent manner. For characteristics classified as binary, a simple yes / no questionnaire can be used to collect data. It may be necessary to develop a defined system to collect data in a consistent manner. For characteristics classified as ordinal, measurement scales as described above can be developed.

[0075] To develop a measurement scale for ordinal properties, method 100 can employ at least four methods to develop a consensus function. In one embodiment, an expert or team of experts can be used to determine the type of relationship between the property and the change in the target variable. In another embodiment, the ordinal property can be scaled (e.g., 1, 2, 3...n for n configurations). By plotting the relationship between the target value and the configurations, the configurations are placed in an asymptotic order of influence. When using an arbitrary scaling method, the determination of the relationship between the target variable value and the ordinal property is forced into the optimization analysis, as described in more detail below. In this case, the general optimization model described in Equation 1.0 can be modified to accommodate potential nonlinear relationships. In another embodiment, the ordinal measurement can be scaled as discussed above and then regressed against the data to make the curve of the target variable versus the ordinal property as close to linear as possible. In yet another embodiment, a combination of the foregoing embodiments can be used to leverage available expert experience, as well as the available data quality and quantity.

[0076] Once method 100 establishes the relationship, it can develop a measurement scale in step 110. For example, a single characteristic can take the form of five different physical configurations. The characteristic with the least impact on changes in the target variable can be assigned a scale setting score. This value can be assigned to any non-zero value. In this example, the assigned value is 1.0. The characteristic with the second largest impact on changes in the target variable will be a function of the scale setting value, as determined by a consistency function. The consistency function is achieved by utilizing the measurement scale described above for ordinal characteristics. This is repeated until a scale for the applicable physical configuration is developed.

[0077] In step 112, method 100 uses the classification system developed in step 110 to collect data. The data collection process can begin with the development of data input forms and instructions. In many cases, data collection training workshops are conducted to assist in data collection. Training workshops can improve the consistency and accuracy of data submissions. Considerations in data collection may involve defining the boundaries of the measurable system (such as an industrial facility) being analyzed. Data input instructions can provide definitions of what costs and staffing of the measurable system should be included in the data collection. Data collection input forms can provide worksheets for many reporting categories to aid in the preparation of the data used for input. The collected data can originate from several sources, including existing historical data, newly collected historical data from existing facilities and processes, simulation data from one or more models, or synthesized empirical data from experts in the field.

[0078] In step 114, method 100 can validate the data. Many data checks can be programmed in step 114 of method 100 such that method 100 can accept data that passes the validation checks, or the checks can be rewritten by appropriate authorities. Validation routines can be developed to validate the data as it is collected. Validation routines can take many forms, including: (1) the range of acceptable data is the ratio of one data point to another; (2) where applicable data is cross-checked against all other similar data submitted to identify outlier data points for further investigation; and (3) data is cross-referenced to any previous data submissions by experts for their judgment. After all input data validations are satisfied, the data is examined relative to all data collected in an extensive “cross-study” validation. This “cross-study” validation can highlight more areas that need to be examined and can lead to changes in the input data.

[0079] In step 116, method 100 may develop constraints for use in solving the comparative analysis model. These constraints may include constraints on the values ​​of the model coefficients. These may be minimum or maximum values, or constraints on grouping values, or any other form of mathematical constraint. One method for determining constraints is shown in Figure 12, which is discussed in more detail below. Then, in step 118, method 100 solves the comparative analysis model by applying a self-selected optimization method (such as linear regression) to the collected data to determine the optimal set of factors that associate the target variable with the characteristics. In one embodiment, a generalized gradient-reducing nonlinear optimization method may be used. However, method 100 may utilize many other optimization methods.

[0080] In step 120, method 100 can determine the developed characteristic. The developed characteristic is the result of any mathematical relationship existing between one or more first-principles characteristics and can be used to express the information represented by that mathematical relationship. Furthermore, if a linear general optimization model is used, nonlinear information in the characteristic can be captured in the developed characteristic. The determination of the form of the developed characteristic is accomplished through discussions with experts, modeling expertise, and through trial and error. In step 122, method 100 applies the optimization model to the primary first-principles characteristics and the developed characteristic to determine the model coefficients. In one embodiment, if the developed characteristic is used, steps 116 through 122 can be repeated iteratively until method 100 reaches a level of model accuracy.

[0081] Figure 11 is a flowchart of an embodiment of a method 200 for determining the principal first principle characteristic 106 as described in Figure 10. In step 202, method 200 determines the effect of each characteristic on the change in a target variable across measurable systems. In one embodiment, the method may be iteratively repeated, and a comparative analysis model may be used to determine the effect of each characteristic. In another embodiment, method 200 may use a correlation matrix. The effect of each characteristic may be expressed in the initial dataset as a percentage of the total change in the target variable. In step 204, method 200 may rank each characteristic from highest to lowest based on its effect on the target variable. Those skilled in the art will recognize that method 200 may use other ranking criteria.

[0082] In step 206, features may be grouped into one or more categories. In one embodiment, features are grouped into three categories. The first category includes features that affect the target variable by a percentage less than a lower threshold (e.g., approximately 5%). The second category may include one or more features having a percentage between the lower percentage and a second threshold (e.g., approximately 5% and approximately 20%). The third category may include one or more features having a percentage exceeding the second threshold (e.g., approximately 20%). Other embodiments of method 200 in step 2006 may include additional or fewer categories and / or different ranges.

[0083] In step 208, method 200 may remove features from the feature list that have an average change in the target variable below a certain threshold. For example, method 200 may remove features that include the first category described above in step 206 (e.g., features with a percentage of less than approximately five percent). Those skilled in the art will recognize that other thresholds may be used, and multiple categories may be removed from the feature list. In one embodiment, if a feature is removed, the process may be repeated in step 202 above. In another embodiment, features are not removed from the list until another covariance is determined to exist, as described in step 212 below.

[0084] In step 210, method 200 determines the relationships between intermediate-level characteristics. Intermediate-level characteristics are those that have some level of influence on the target variable but do not significantly affect it individually. Using descriptive categories, those characteristics in the second category are intermediate-level characteristics. Example relationships between characteristics are covariance, dependency, and independence. A covariance relationship occurs when modifying one characteristic causes a change in the target variable, but only if the other characteristic is present. For example, in the case where a change in characteristic "A" causes a change in the target variable, "A" and "B" are covariant only if characteristic "B" is present. A dependency relationship occurs when one characteristic is derived from or directly related to another characteristic. For example, when characteristic "A" exists only if characteristic "B" is present, A and B are dependent. Characteristics that are not covariant or dependent are classified as having an independent relationship.

[0085] In step 212, method 200 may remove dependencies and highly correlated features to decompose features that show dependence on each other. Several potential methods for decomposing dependencies exist. Some examples include: (i) grouping multiple dependent features into a single feature, (ii) removing all dependent features except for one, and (iii) retaining one dependent feature and creating a new feature that differentiates the retained feature from the others. After method 200 removes dependencies, the process can be repeated starting from step 202. In one embodiment, if the differential variable is not significant, it can be removed from the analysis in the repeated step 208.

[0086] In step 214, method 200 may analyze the characteristics to determine the degree of their interrelationship. In one embodiment, if any previous step results in a repetition of the process, the repetition should be performed before step 214. In some embodiments, the process may be repeated multiple times before proceeding to step 214. At 216, characteristics that cause a change in the target variable less than a minimum threshold in the effect of another characteristic on the change of the target variable are removed from the list of potential characteristics. The illustrative threshold may be approximately 10%. For example, if the change in the target variable caused by characteristic "A" increases when characteristic "B" is present, the percentage increase in the change of the target variable caused by the presence of characteristic "B" must be estimated. If the change in characteristic "B" is estimated to increase the change in the target variable by less than approximately 10% of the increase caused by characteristic "A" alone, then characteristic "B" can be removed from the list of potential characteristics. Characteristic "A" may also be considered to have an insignificant effect on the target variable. The remaining characteristics are considered primary characteristics.

[0087] Figure 12 is a flowchart of an embodiment of a method 300 for developing constraints used in solving the comparative analysis model described in step 116 of Figure 10. In step 302, constraints are developed for the model coefficients. In other words, constraints are any limitations placed on the model coefficients. For example, the model coefficients could be constrained with respect to having a maximum contribution of approximately 20% to the target variable. In step 354, the objective function of method 300 (described below) is optimized to determine an initial set of model coefficients. In step 306, method 300 can calculate the percentage contribution of each characteristic to the target variable. Several methods exist for calculating the percentage contribution of each characteristic, such as the "averaging method" described in, for example, U.S. Patent 7,233,910.

[0088] Having developed the percentage contributions, method 300 proceeds to step 308, where each percentage contribution is compared with expert knowledge. Domain experts may have an intuitive or experiential sense of the relative impact of key features on the overall target value. The contribution of each feature is judged against this expert knowledge. In step 310, method 300 may make a decision regarding the acceptability of each contribution. If a contribution is found to be unacceptable, method 300 continues to step 312. If a contribution is found to be acceptable, method 300 continues to step 316.

[0089] In step 312, method 300 makes a decision on how to address or handle unacceptable results from the individual contributions. In step 312, options may include adjusting the constraints affecting the solution of the model coefficients, or determining that the selected set of characteristics cannot be helped by constraint adjustment. If the user decides to accept the constraint adjustment, method 300 proceeds to step 316. If a decision is made to achieve acceptable results through constraint adjustment, method 300 continues to step 314. In step 314, the constraints are adjusted to increase or decrease the influence of the individual characteristics in an effort to obtain acceptable results from the individual contributions. Method 300 continues to step 302 using the revised constraints. In step 316, peer and expert review of the developed model coefficients may be performed to determine the acceptability of the developed model coefficients. If the factors pass the expert and peer review, method 300 continues to step 326. If the model coefficients are found to be unacceptable, method 300 continues to step 318.

[0090] In step 318, method 300 may obtain additional methods and suggestions for modifying the developed characteristics by collaborating with experts in the specific field. This may include the creation of new or updated developed characteristics, or the addition of new or updated first-principles characteristics to the analysis dataset. In step 320, a determination is made regarding the existence of data from an investigation supporting methods and suggestions for modifying the characteristics. If the data exists, method 300 proceeds to step 324. If the data does not exist, method 300 proceeds to step 322. In step 322, method 300 collects additional data to attempt to make the corrections required to obtain a satisfactory solution. In step 324, method 300 modifies the characteristic set based on the new methods and suggestions. In step 326, method 400 may document the reasoning behind the characteristic selection. This document may be used to interpret the results of the use of model coefficients.

[0091] Figure 13 is a schematic diagram of an embodiment of the model coefficient matrix 10 used to determine the model coefficients as described in Figures 10-12. While the model coefficient matrix 10 can be expressed in various configurations, in this particular example, it can be interpreted on one axis using first principle characteristics 12 and first development characteristics 14, and on another axis using different facilities 16 for which data is collected. For each first principle characteristic 12 in each facility 16, there exists an actual data value 18. For each first principle characteristic 12 and development characteristic 14, there exists a model coefficient 22 that will be calculated using the optimization model. Constraints 20 limit the range of the model coefficients 22. Constraints can be minimum or maximum values, or other mathematical functions or algebraic relationships. Moreover, constraints 20 can be grouped and further constrained. Additional constraints 20 can also be employed regarding facility data, as well as constraints 20 similar to those used in the data validation step regarding relationships between data points, and constraints 20 can be employed regarding any mathematical relationship of the input data. In one embodiment, the constraints 20 to be satisfied during optimization apply only to the model coefficients.

[0092] Column 24, Target Variable (Actual), includes the actual values ​​of the target variables as measured for each facility. Column 26, Target Variable (Predicted), includes the values ​​for the target values ​​as calculated using the determined model coefficients. Column 28, Error, includes the error values ​​for each facility as determined by the optimization model. Error Sum 30 is the sum of the error values ​​in Error Sum 30. The optimization analysis, including the target variable equation and objective function, solves for the model coefficients to minimize Error Sum 30. In the optimization analysis, the model calculates the coefficient α. j The error ∈ is calculated as the minimum error for all facilities. i The nonlinear optimization process determines the set of model coefficients that minimize this equation for a given set of first-principles properties, constraints, and selected values.

[0093] The target variable can be calculated as a function of the characteristics and the model coefficients to be determined. The target variable equation is expressed as:

[0094] Target variable equation: TV i The target variable is represented by α; the characteristic variable represents the first-principle characteristic; f is the value of the first-principle characteristic or the developed principle characteristic; i represents the facility number; j represents the characteristic number; α j This represents the j-th model coefficient, which is consistent with the j-th principle characteristic; and ε i The error representing the TV prediction of the model is defined as the difference between the actual target variable value and the predicted target variable value for facility i.

[0095] The objective function has a general form:

[0096] Objective function: Where i is the facility; m represents the total number of facilities; and p represents the selected value.

[0097] A common use of the general form of the objective function is to minimize the absolute sum of errors by using p=1, as shown below:

[0098] Objective function:

[0099] Another common use of the general form of the objective function is to use the least squares version corresponding to p=2, as shown below:

[0100] Objective function: Because the analysis involves a finite number of first-principles properties and the objective function's form corresponds to a mathematical norm, the analytical results do not depend on a specific value of p. Analysts can choose a value based on the specific problem to be solved or an additional statistical application of the objective function. For example, p=2 is frequently used due to its statistical application in measurement data and the prediction error of the target variable.

[0101] The third form of the objective function is to solve for the simple sum of the squared errors given in Equation 5 below.

[0102] Objective function: While several forms of objective functions have been shown, other forms of objective functions can also be used for specific purposes. Based on optimization analysis, the determined model coefficients are those that, after iteratively moving the model through each facility and characteristic, result in the minimum difference between the sum of the objective variables and the actual values ​​after multiplying each constrained latent model coefficient by the data values ​​for the corresponding characteristic and summing over the specific facility.

[0103] For illustrative purposes, more specific examples of one or more embodiments used to determine the model coefficients employed in the comparative performance analysis shown in Figures 10-12 are discussed below. A catalytic cracker is a processing unit found in most refineries. A catalytic cracker cracks long molecules into shorter, lighter molecules within the gasoline boiling point range. This process typically takes place at relatively high temperatures in the presence of a catalyst. During the cracking feed, coke is generated and deposited on the catalyst. The coke is burned off the catalyst to recover heat and reactivate it. A catalytic cracker has several main components: a reactor, a regenerator, a main fractionator, and emission control equipment. Refineries may wish to compare the performance of their catalytic crackers with that of catalytic crackers operated by their competitors. Examples comparing different catalytic crackers are illustrative but do not represent the actual results of applying this method to catalytic crackers or any other industrial facility. Moreover, the catalytic cracker example is merely one example among many potential embodiments used to compare measurable systems.

[0104] Using Figure 10 as an example, Method 100 begins at step 102 and determines that the target variable will be the “cash operating cost” or “cash OPEX” in the catalytic cracking facility. In step 104, the first principle characteristics that may affect the cash operating cost of the catalytic cracker may include one or more of the following: (1) feed quality; (2) regenerator design; (3) personnel experience; (4) location; (5) cell age; (6) catalyst type; (7) feed capacity; (8) personnel training; (9) union; (10) reactor temperature; (11) regeneration equipment; (12) reactor design; (13) emission control equipment; (14) main fractionator design; (15) maintenance practices; (16) regenerator temperature; (17) feed preheating level; (18) staffing level.

[0105] To determine the primary characteristics, method 100 may determine the impact of the first characteristic in step 106. In one embodiment, method 100 may implement step 106 by determining the primary characteristics as shown in FIG11. In FIG11, in step 202, method 200 may assign a percentage change to each characteristic. In step 204, method 200 may rate and rank the characteristics from the catalytic cracker example. The following graphs in Table 1 show the relative impact and ranking of at least some example characteristics:

[0106]

[0107] Table 1

[0108] In this embodiment, the categories are shown in Table 2 below:

[0109]

[0110] Table 2

[0111] Other embodiments may have any number of categories, and the percentage values ​​for demarcation between categories may be changed in any way.

[0112] Based on the example ranking above, method 200 groups the characteristics according to category in step 206. In step 208, method 200 may discard characteristics in category 3 as minor. Method 200 may analyze the characteristics in category 2 to determine the type of relationship they exhibit with other characteristics in step 210. Method 200 may classify each characteristic in step 212 as exhibiting covariance, dependence, or independence. Table 3 is an example of classifying the characteristics of a catalytic cracking facility:

[0113]

[0114]

[0115] Table 3

[0116] In step 214, method 200 can analyze the degree of relationship between these characteristics. Using this embodiment for the catalytic cracker example: personnel configuration levels classified as having independent relationships can be retained in the analysis. The unit's age is classified as dependent on personnel training. Dependency means that the unit's age is a derivative of personnel experience, and vice versa. After further consideration, method 200 can decide to discard the unit's age characteristic from the analysis, and the broader characteristic of personnel training can be retained. Three characteristics classified as having covariant relationships—personnel training, emission equipment, and maintenance practices—must be examined to determine the degree of covariance.

[0117] Method 200 can determine that changes in cash operating costs caused by changes in personnel training can be modified by more than 30% by changes in maintenance practices. Similarly, changes in cash operating costs caused by changes in emission equipment can be modified by more than 30% by changes in maintenance practices, thus requiring maintenance practices, personnel training, and emission equipment to be retained in the analysis. Method 200 can also determine that changes in cash operating costs caused by changes in maintenance practices are not modified by changes in personnel experience exceeding a selected threshold of 30%, thus requiring personnel experience to be discarded from the analysis.

[0118] Continuing with the catalytic cracker example and returning to Figure 10, method 100 classifies the remaining characteristics in step 108 as continuous, ordinal, or binary type measurements, as shown in Table 4.

[0119]

[0120] Table 4

[0121] In this catalytic cracker example, maintenance practices can have an "economies of scale" relationship with cash operating costs (which is the objective variable). Improvements in the objective variable improve at a decreasing rate with improvements in maintenance practices. Based on historical data and experience, a complexity factor is assigned to reflect economies of scale. In this specific example, a factor of 0.6 is chosen. As an example of coefficients, complexity factors are often estimated to follow a power curve relationship. Using cash operating costs as an example that typically exhibits the characteristic of "economies of scale," the effect of maintenance practices can be described as follows:

[0122]

[0123] In step 110, method 100 can develop a data collection classification system. In this example, a questionnaire can be developed to measure how many of the ten key maintenance practices are frequently used in each facility. A definition system can be used to ensure that the data is collected in a consistent manner. The data on the number of frequently used maintenance practices is converted into maintenance practice scores using a 0.6 factor and an "economies of scale" relationship, as shown in Table 5.

[0124]

[0125] Table 5

[0126] For illustrative purposes regarding the example of the catalytic cracker, in step 112, method 100 may collect data, and in step 114, method 100 may verify the data as shown in Table 6:

[0127]

[0128] Table 6

[0129] A team of experts developed constraints for each feature to control the model so that the results are within a reasonable range of the solutions shown in Table 7.

[0130]

[0131] Table 7

[0132] In step 116, method 100 produces the results of the model optimization run, which are shown in Table 8 below.

[0133]

[0134] Table 8

[0135] This model indicates that emission equipment and maintenance practices are not significant drivers of changes in cash operating costs across different catalytic crackers. This can be indicated by finding approximately zero values ​​for the model coefficients for these two characteristics. Reactor design, personnel training, and emission equipment were found to be significant drivers. In the case of both emission equipment and maintenance practices, experts can agree that these characteristics may not be significant in driving changes in cash operating costs. Experts can identify dependency effects that may not have been previously identified to fully compensate for the influence of emission equipment and maintenance practices.

[0136] Figure 14 is a schematic diagram of an embodiment of the model coefficient matrix 10 for a catalytic cracker used to determine the model coefficients used in the comparative performance analysis shown in Figures 10-12. Figure 14 shows a sample model configuration for an illustrative example of a catalytic cracker. Data 18, actual values ​​24, and the resulting model coefficients 22 are shown. In this example, the error sum 30 is relatively minimal, therefore no developed characteristic is required in this case. In other examples, the sum of errors for different values ​​can be determined to be significant, thus necessitating the determination of developed characteristics.

[0137] For additional illustrative purposes, another example of the model coefficients used to determine the comparative performance analysis shown in Figures 10-12 is discussed below. This example will involve pipeline and tank farm terminals. Pipelines and tank farms are assets used by the industry to store and distribute liquid and gaseous feedstocks and products. This example illustrates the development of equivalence factors for: (1) pipelines and pipeline systems; (2) tank farm terminals; and (3) any combination of pipelines, pipeline systems, and tank farm terminals. This example is for illustrative purposes and does not represent the actual results of applying this method to any particular pipeline and tank farm terminal or any other industrial facility.

[0138] Using Figure 10 as an example, in step 102, method 100t selects the desired target variable as "cash operating cost" or "cash OPEX" in the pipeline assets. For step 104, the first-principles characteristics that may affect cash operating cost for pipeline-related characteristics may include: (1) the type of fluid being transported; (2) the average fluid density; (3) the number of input and output stations; (4) the total installed capacity; (5) the total main pump drive kilowatts (KW); (6) the length of the pipeline; (7) the pipeline height variation; (8) the total utilization capacity; (9) the pipeline replacement value; and (10) the pump station replacement value. The first-principles characteristics that may affect cash operating cost for tank-related characteristics may include: (1) the fluid type; (2) the number of tanks; (3) the total number of valves in the terminal; (4) the total rated tank capacity; (5) the annual turnover of the tank; and (6) the tank terminal replacement value.

[0139] To determine the primary first-principles characteristics, method 100 determines the impact of the first characteristics in step 106. In one embodiment, method 100 can implement step 106 by determining the primary characteristics as shown in FIG11. In FIG11, in step 202, method 100 can assign an impact percentage to each characteristic. This analysis shows that pipeline replacement values ​​and tank terminal replacement values ​​are widely used in industry and are characteristics that depend on more fundamental characteristics. Therefore, in this case, these values ​​are removed from consideration of the primary first-principles characteristics. In step 204, method 200 can rate and rank the characteristics. Table 9 shows the relative impact and ranking for example characteristics, where method 200 can assign a change percentage to each characteristic.

[0140]

[0141]

[0142] Table 9

[0143] In this embodiment, the categories are as shown in Table 10:

[0144]

[0145] Table 10

[0146] Other embodiments may have any number of categories, and the percentage values ​​for demarcation between categories may be changed in any way.

[0147] Based on the example ranking above, method 200 groups the features according to category in step 206. In step 208, method 200 discards features in category 3 because they are minor. Method 200 may further analyze the features in category 2 to determine the type of relationship they exhibit with other features in step 210. Method 200 classifies each feature as exhibiting covariance, dependency, or independence, as shown in Table 11 below:

[0148]

[0149] Table 11

[0150] In step 212, method 200 can decompose dependent features. In this example, there are no dependent features that method 200 needs to decompose. In step 214, method 200 can analyze the covariance of the remaining features and determine that no features are discarded. Method 200 can then treat the remaining variables as principal features in step 218.

[0151] Continuing with the pipeline and tank yard example and returning to Figure 10, method 100 can classify the remaining characteristics into continuous, ordinal, or binary type measurements in step 108, as shown in Table 12.

[0152]

[0153]

[0154] Table 12

[0155] In step 110, method 100 may develop a data collection classification system. In this example, a questionnaire may be developed to collect information about the aforementioned measurements from participating facilities. In step 112, method 100 may collect data, and in step 114, method 100 may validate the data, as shown in Tables 13 and 14.

[0156]

[0157] Table 13

[0158]

[0159] Table 14

[0160] In step 116, method 100 may be developed by an expert to impose constraints on the model coefficients, as shown in Table 15 below.

[0161]

[0162] Table 15

[0163] In step 116, method 100 produces the results of the model optimization run, which are shown in Table 16 below.

[0164]

[0165]

[0166] Table 16

[0167] In step 118, method 100 can determine the features that do not need to be developed in this example. The final model coefficients may include the model coefficients determined in the comparative analysis model step described above.

[0168] Figure 15 is a schematic diagram of an embodiment of the model coefficient matrix 10 for pipeline and tank farms used to determine the model coefficients used in the comparative performance analysis shown in Figures 10-12. This example illustrates only one of the many potential applications of the invention to the pipeline and tank farm industry. The methods described and illustrated in Figures 10-15 can be applied to many other different industries and facilities. For example, this method can be applied to the power generation industry, such as developing model coefficients for predicting operating costs for single-cycle and combined-cycle power plants that generate electricity from any combination of circulating boilers, steam turbine generators, gas turbine generators, and heat recovery steam generators. In another example, this method can be applied to developing model coefficients for ethylene manufacturers to predict the annual costs of complying with environmental regulations associated with continuous emissions monitoring and reporting from ethylene furnaces. In one embodiment, the model coefficients will be applied to both environmental and chemical industry applications.

[0169] Figure 9 is a schematic diagram of an embodiment of a computing node for implementing one or more embodiments described in this disclosure, such as methods 60, 100, 200, and 300 described in Figures 1 and 10-12, respectively. The computing node may correspond to or be part of a computer and / or any other computing device, such as a handheld computer, tablet computer, laptop computer, portable device, workstation, server, mainframe, supercomputer, and / or database. The hardware includes a processor 900 containing sufficient system memory 905 to perform the required numerical calculations. The processor 900 executes a computer program residing in the system memory 905, which may be a non-transitory computer-readable medium, to perform methods 60, 100, 200, and 300 as described in Figures 1 and 10-12, respectively. A video and storage controller 910 may be used to enable the operation of a display 915 to display various information, such as tables and user interfaces described in Figures 2-8. The computing node includes various data storage devices for data input, such as floppy disk unit 920, internal / external disk drive 925, internal CD / DVD 930, tape unit 935, and other types of electronic storage media 940. The data storage devices mentioned above are merely illustrative and exemplary.

[0170] The computing node may also include one or more other input interfaces (not shown in FIG. 9) comprising at least one receiving device configured to receive data via electrical, optical, and / or wireless connections using one or more communication protocols. In one embodiment, the input interface may be a network interface including multiple input ports configured to receive and / or transmit data via a network. In particular, the network may transmit operational and performance data via wired links, wireless links, and / or logical links. Other examples of input interfaces may include, but are not limited to, keyboards, universal serial bus (USB) interfaces, and / or graphical input devices (e.g., on-screen and / or virtual keyboards). In another embodiment, the input interface may include one or more measuring and / or sensing devices for measuring the first-principles data of the asset unit or other asset-level data 64 described in FIG. 1. In other words, the measuring and / or sensing devices may be used to measure various physical properties and / or characteristics associated with the operation and performance of the measurable system.

[0171] These storage media are used to input the dataset and outlier removal criteria into the computation nodes, store the dataset after outlier removal, store the calculated factors, and store the trend lines and iterative trend line plots generated by the system. The computations can be performed using statistical software packages, or can be performed using, for example, Microsoft... Data entered in spreadsheet format is used for execution. In one embodiment, a custom software program designed for a company-specific system implementation or by leveraging Microsoft... The computation can be performed using commercially available software compatible with other databases and spreadsheet programs. The computation node can also interface with dedicated or public external storage media 955 to link with other databases, thereby providing data to be used with future reliability based on current maintenance expenditure methods. Output interfaces include output devices for transmitting data. Output devices can be telecommunications equipment 945, transmission equipment, and / or devices used to transmit processed future reliability data, such as computation data spreadsheets, graphs, and / or reports, to other computation nodes, network nodes, control centers, printers 950, electronic storage media 920, 925, 930, 935, 940, and / or dedicated storage databases 960 via one or more networks, intranets, or the Internet. These output devices used herein are merely illustrative and exemplary.

[0172] In one embodiment, system memory 905 interfaces with a computer bus or other connection to transfer or send information stored in system memory 905 to processor 900 during the execution of a software program, such as an operating system, application program, device driver, and software modules including program code and / or computer-executable process steps incorporating the functionality described herein, such as methods 60, 100, 200, and 300. Processor 900 first loads the computer-executable process steps from storage, such as system memory 905, storage media, removable media drives, and / or other non-transitory storage devices. Processor 900 can then execute the stored process steps to perform the loaded computer-executable process steps. During the execution of the computer-executable process steps, processor 900 can access stored data (e.g., data stored by storage devices) to instruct one or more components within a computing node.

[0173] It is well known in the art to program and / or load executable instructions into system memory 905 and / or one or more processing units such as processors or microprocessors to transform computing node 40 into a non-general-purpose, specific machine or device that performs modeling for estimating the future reliability of a measurable system. Implementing instructions, real-time monitoring, and other functions by loading executable software into a microprocessor and / or processor can be translated into a hardware implementation and / or a general-purpose processor can be transformed into an application-specific processor using well-known design rules. For example, the decision between implementing a concept in software or hardware can depend on several design choices, including the stability of the design and the number of units to be produced, as well as the issues involved in transitioning from the software domain to the hardware domain. Designs can often be developed and tested in software form and subsequently transformed into an equivalent hardware implementation in an ASIC or dedicated hardware with hard-wired software instructions using well-known design rules. Just as a machine controlled by a new ASIC is a specific machine or device, a computer that has been programmed and / or loaded with executable instructions is also considered a non-general-purpose, specific machine or device.

[0174] Figure 16 is a schematic diagram of another embodiment of a computing node 40 for implementing one or more embodiments described in this disclosure, such as methods 60, 100, 200, and 300 described in Figures 1 and 10-12, respectively. The computing node 40 can be any form of computing device, including computers, workstations, handheld devices, mainframes, embedded computing devices, holographic computing devices, bio-computing devices, nanotechnology computing devices, virtual computing devices, and / or distributed systems. The computing node 40 includes a microprocessor 42, an input device 44, a storage device 46, a video controller 48, a system memory 50, and a display 54, all interconnected via one or more buses, wires, or other communication pathways 52, as well as a communication device 56. The storage device 46 can be a floppy disk drive, a hard disk drive, a CD-ROM, an optical drive, a bubble memory, or any other form of storage device. Furthermore, the storage device 42 can be capable of receiving floppy disks, CD-ROMs, DVD-ROMs, memory sticks, or any other form of computer-readable medium that can contain computer-executable instructions or data. The other communication device 56 can be a modem, network card, or any other device to enable the node to communicate with people or other nodes.

[0175] At least one embodiment has been disclosed, and variations, combinations, and / or modifications to one or more embodiments and / or features of one or more embodiments that can be made by those skilled in the art are also within the scope of this disclosure. Alternative embodiments resulting from combining, integrating, and / or omitting features of one or more embodiments are also within the scope of this disclosure. When numerical ranges or limitations are explicitly stated, such expressions of ranges or limitations can be understood to include iterative ranges or limitations of similar values ​​falling within the explicitly stated ranges or limitations (e.g., from about 1 to about 10, including 2, 3, 4, etc.; greater than 0.10, including 0.11, 0.12, 0.13, etc.). Unless otherwise stated, the term "about" is used to mean ±10% of the following figures.

[0176] The use of the term "optionally" with respect to any element of a claim means that the element is required, or alternatively, that the element is not required, both of which are within the scope of the claim. The use of broader terms such as including, comprising, and having can be understood to provide support for narrower terms such as consisting of, substantially consisting of, and substantially constituted by. Therefore, the scope of protection is not limited by the foregoing description, but is defined by the appended claims, encompassing all equivalents of the subject matter of the claims. Each claim is incorporated into the specification as further disclosure and the claims are embodiments of this disclosure(s).

[0177] While several embodiments have been provided in this disclosure, it will be understood that the disclosed embodiments may be embodied in many other specific forms without departing from the spirit or scope of this disclosure. The examples given are to be considered illustrative rather than restrictive, and the invention is not limited to the details given herein. For example, various elements or components may be combined or integrated in another system, or certain features may be omitted or not implemented. Well-known elements are given without detailed description so as not to obscure the invention with unnecessary detail. In most cases, details unnecessary for obtaining a full understanding of the invention are omitted, as such details are within the skill of one of ordinary skill in the art.

[0178] Furthermore, without departing from the scope of this disclosure, the various embodiments described and shown as discrete or separate in the various embodiments can be combined or integrated with other systems, modules, technologies, or methods. Other items shown or discussed as coupled or directly coupled or communicating with each other may be indirectly coupled or communicating through some interface, device, or intermediate component, whether electrically, mechanically, or otherwise. Other examples of changes, substitutions, and modifications can be identified by those skilled in the art without departing from the spirit and scope of this disclosure.

[0179] While the systems and methods described herein have been described in detail, it should be understood that various changes, substitutions, and modifications can be made without departing from the spirit and scope of the invention as defined by the appended claims. Those skilled in the art will be able to study preferred embodiments and identify other ways of practicing the invention not precisely described herein. The intent of this disclosure is that variations and equivalents of the invention are within the scope of the claims, while the description, abstract, and drawings are not intended to limit the scope of the invention. The invention is specifically intended to be as extensive as the following claims and their equivalents.

[0180] Finally, it should be noted that the discussion of any references is not an admission that they are prior art to the invention, especially any references with publication dates after the priority date of this application. Meanwhile, each of the following claims is incorporated into this specific description or specification as an additional embodiment of this disclosure.

Claims

1. A system comprising: At least one measurable system, said at least one measurable system comprising multiple equipment assets operated at the facility; At least one measuring device, the at least one measuring device being configured to generate measurement data of at least one performance attribute related to the operation or performance of multiple equipment assets of the at least one measurable system; A processor, operatively coupled to: i) the at least one measuring device, and ii) a non-transitory computer-readable medium, wherein the non-transitory computer-readable medium includes instructions that, when executed by the processor, cause the processor to: receive maintenance cost data of the at least one measurable system; receive first principle data corresponding to the facility; and receive asset reliability data of the at least one measurable system; receive one or more comparative analysis models associated with the at least one measurable system; and generate at least one maintenance criterion based on the maintenance cost data and the first principle data, using the one or more comparative analysis models. Based on the asset reliability data and category values ​​derived from the maintenance cost data using the at least one maintenance criterion, the estimated future reliability of the at least one measurable system is determined; Multiple category values ​​are generated, which divide the maintenance cost data into specified intervals based on the maintenance criteria. One or more of the comparative analysis models are updated based on the updated data received through the measurement device, and an update of the future reliability estimate for one or more measurable systems is subsequently provided. The estimated future reliability data identification involves performing one or more reliability-effective maintenance tasks on the at least one measurable system to affect the target variable of the at least one measurable system; Based on performing one or more reliable and effective maintenance tasks on the facility, current data of at least one major first principle characteristic of the at least one measurable system is obtained, and the current data is sent to the processor, which determines the change of the target variable of the at least one measurable system; The user interface is configured to display the estimated future reliability data.

2. The system of claim 1, wherein the asset reliability data is equivalent forced downtime rate data.

3. The system of claim 1, wherein when the instructions are executed by the processor, the processor further causes the processor to compile the at least one maintenance standard and the asset reliability data into a compiled data file.

4. The system of claim 3, wherein when the instructions are executed by the processor, the processor further causes the processor to: generate categorized time-based maintenance cost data based at least on the compiled data file; and generate categorized time-based asset reliability data based at least on the compiled data file.

5. The system of claim 4, wherein when the instructions are executed by the processor, the processor further causes the processor to generate classified time-based maintenance cost data by arranging category values ​​for a plurality of other facilities according to one or more time intervals.

6. The system of claim 4, wherein when the instructions are executed by the processor, the processor further causes the processor to generate categorized time-based asset reliability data by arranging asset reliability data values ​​for a plurality of other facilities according to one or more time intervals.

7. The system of claim 1, wherein the estimated future reliability interval is based on the amount of the maintenance cost data, the asset reliability data, and the first principle data.

8. The system of claim 1, wherein the at least one maintenance standard is used to normalize maintenance cost data.

9. The system of claim 8, wherein when the instruction is executed by the processor, the processor further causes the processor to normalize the maintenance cost data by generating a periodic maintenance expense divisor over a period of time.

10. The system of claim 1, wherein the estimated future reliability report includes a graph displaying asset reliability data according to multiple category values.

11. A method comprising: At least one performance attribute related to the operation or performance of a measurable system is measured by at least one measuring device, the measurable system comprising multiple equipment assets operated at a facility; The processor receives maintenance cost data associated with at least one measurable system; wherein the processor is operatively coupled to the at least one measuring device; the processor receives first principle data corresponding to the facility; and the processor receives asset reliability data associated with the at least one measurable system. The processor receives one or more comparative analysis models related to the at least one measurable system; the processor generates at least one maintenance standard based on the maintenance cost data and the first principle data using the one or more comparative analysis models; The processor determines the estimated future reliability of the at least one measurable system based on asset reliability data and category values ​​derived from the maintenance cost data using the maintenance criteria. The processor generates multiple category values, which divide the maintenance cost data into specified intervals based on the maintenance criteria. The processor updates one or more of the comparative analysis models based on updated data received through the at least one measurement device, and subsequently provides an update to the future reliability estimate of one or more measurable systems; The processor then uses an output interface to output the estimated future reliability and all its updates.

12. The method of claim 11, further comprising: The at least one maintenance standard is generated based at least in part on maintenance cost data and first principle data.

13. The method of claim 12, wherein the at least one maintenance criterion is used to generate normalized maintenance cost data from the maintenance cost data and the one or more comparative analysis models.

14. The method of claim 13, further comprising: The standardized maintenance cost data is generated by generating a divisor for periodic maintenance expenses.

15. The method of claim 11, further comprising: Compile the at least one maintenance standard and the asset reliability data into a compiled data file; At least based on the compiled data file, generate time-based maintenance cost data categorized by category; and at least based on the compiled data file, generate time-based asset reliability data categorized by category.

16. The method of claim 15, wherein generating categorized time-based maintenance cost data includes arranging category values ​​for a plurality of other facilities according to one or more time intervals.

17. The method of claim 15, wherein generating categorized time-based asset reliability data includes arranging asset reliability data values ​​for a plurality of other facilities according to one or more time intervals.

Images