JP7039737B2 - Methods and systems for diagnosing freezer performance abnormalities - Google Patents

Description

The present invention relates as a whole to computer mounting methods and systems for diagnosing chiller performance anomalies, and more specifically to chiller performance anomalies diagnosis based on root cause analysis (RCA).

As the main cooling equipment, refrigerators are widely used in comfortable cooling and industrial process cooling plants. Root cause analysis (RCA) methods based on Theory of Constraints have been developed to diagnose chiller performance anomalies such as reduced chilled water supply capacity and chilled water supply temperature spikes. In performance anomaly diagnosis, RCA may analyze the correlation between the performance of the chiller, such as cold water supply (CHW), and major constraints, such as cooling water flow rate, cold water return temperature, and the like. The constraint that most strongly correlates with the freezer performance during the anomaly is identified as the bottleneck or root cause of the freezer performance anomaly. RCA diagnostics have been useful in reducing maintenance costs for refrigerators and have been applied to various types of refrigerators, including vapor absorption chillers (SACs) and electric chillers (ECs).

However, in the conventional performance abnormality diagnosis based on RCA, various problems regarding low quality have been found. For example, the conventional performance abnormality diagnosis based on RCA may not be able to capture the abnormality in the performance of the refrigerator and identify the root cause of the abnormality.

Therefore, for refrigerating performance abnormality diagnosis, more specifically for RCA-based performance abnormality diagnosis, which seeks to overcome or at least improve one or more of the defects of the conventional refrigerator performance abnormality diagnosis method. Methods and systems are needed. The present invention has been developed in opposition to this background technique.

According to the first aspect of the present invention, a computer mounting method for diagnosing a freezer performance abnormality is provided, and the method is described as.
Steps to set up a first correlation model between the performance parameters of the refrigerator and the first plurality of operating parameters,
For the anomalies identified by the performance parameters, the steps to determine the optimal window size for performing root cause analysis diagnosis based on the first correlation model,
In order to determine the root cause of the anomaly identified by the performance parameters of the refrigerator, the root cause analysis diagnosis is a step to carry out the root cause analysis diagnosis based on the first correlation model and the optimum window size for the anomaly. Includes steps performed on a dataset containing measurement data for performance parameters collected over a period of time and measurement data for a first plurality of operating parameters, including anomaly measurement data with anomalies associated with anomalies.

According to the second aspect of the present invention, a system for diagnosing a freezer performance abnormality is provided, and the system is a system.
With memory
It comprises at least one processor communicably linked to a memory and configured to implement the method for diagnosing refrigerating performance anomalies according to a first aspect of the invention.

According to a third aspect of the invention, one or more including instructions that can be executed by at least one processor to implement the method for diagnosing refrigerator performance anomalies according to the first aspect of the invention. Computer program products are provided that are embodied in the form of non-temporary computer-readable storage media.

Embodiments of the present invention will be well understood and readily apparent by those skilled in the art by reading the following description in conjunction with the drawings as a mere example.
It is a schematic flow chart which shows the method for diagnosing the refrigerator performance abnormality by various embodiments of this invention. It is a schematic block diagram which shows the system for diagnosing the refrigerator performance abnormality by various embodiments of this invention. It is a schematic diagram which shows the window size optimization system for carrying out the root cause analysis diagnosis by various embodiments of this invention. FIG. 6 is a schematic diagram illustrating an example system architecture for RCA diagnostics for a refrigerator, including window size calculations, according to various exemplary embodiments of the invention. Presented to and utilized by an operator (eg, an operator device) to define an RCA model (correlation model) and anomalous duration to perform a root cause analysis diagnosis according to various exemplary embodiments of the invention. FIG. 5 is a diagram illustrating an example graphical user interface (GUI). It is a schematic flow diagram which shows the method of the refrigerator abnormality diagnosis based on the RCA model (having the multi-layered RCA model ability) by various exemplary embodiments of this invention. It is a figure which shows the two-layer RCA model which includes the upper layer and the lower layer by various exemplary embodiments of this invention. FIG. 6 is a schematic flow diagram illustrating a method of determining the optimal window size for performing RCA diagnosis for anomalies according to various exemplary embodiments of the invention. It is a figure which shows the sliding window technique of an example over an abnormal period by various exemplary embodiments of this invention.

More specifically, various embodiments of the present invention seek to overcome or at least improve one or more of the defects associated with conventional refrigerator performance abnormality diagnostic methods for refrigerating performance abnormality diagnosis. A method (computer mounting method) and a system for diagnosing a refrigerator performance abnormality based on root cause analysis (RCA) are provided. In connection with various embodiments, a chiller refers to any device or system configured to cool or freeze a fluid (eg, to produce cold water) to provide a cooling effect for a variety of purposes. It will be appreciated by those skilled in the art that it may be and is not limited to any particular type of freezer. For example, embodiments of the present invention include, but are not limited to, a steam absorption chiller (SAC), an electric chiller (EC), a centrifugal chiller (CC), a screw chiller (SC), or a cooling plant. It may be applied or implemented to perform performance anomaly diagnosis for various types of refrigerators, such as in heating, ventilation and air conditioning (HVAC) systems.

Various embodiments of the present invention have identified various problems with low quality found in conventional RCA-based freezer performance anomaly diagnostics. For example, the conventional RCA-based freezer performance abnormality diagnosis may not be able to capture the abnormality in the refrigerator performance and identify the root cause of the abnormality.

Hereinafter, as a mere example, for better understanding, a conventional RCA-based refrigerator performance abnormality diagnosis will be described.

For example, Okitsu et al. , "Root Case Analysis on Changes in Chiller Performance usage Linear Regression", International Conference on Computer and Information Science (Cience, Kuala Lumpur) , "A Case Study of Electric Chiller Performance Bottleneck Diagnosis by Root Cause Analysis", IEEE International Conferences on Information Technology, Information Systems and Electrical Engineering (ICITISEE 2016), Yogyakarta, Indonesia, 23-24 August 2016 (hereinafter referred to as Zhang literature ), Conventional RCA diagnostic methods have been developed based on constraint theory, which has been applied to diagnostic performance abnormalities of refrigerating machines including SAC and EC (both in the Okitsu and Zhang literatures). The contents are incorporated herein by reference in their entirety for all purposes). During conventional RCA diagnostics, the RCA, through multiple linear regression, has a refrigerator performance parameter (eg, cold water supply (CHW)) and its operating parameters (sometimes interchangeably referred to as "constraints") (eg, cooling water supply). A correlation model (RCA model) with temperature, cooling water flow rate, and chilled water regression temperature) may be constructed. The contribution of each constraint to the performance parameters is calculated based on the correlation coefficient associated with the correlation model. Therefore, the constraint with the highest contribution may be selected as a bottleneck or anomalous root cause in the freezer performance.

As a mere example, for better understanding, conventional methods for RCA diagnosis of CHW performance (performance parameters) of air-cooled EC may include the following basic steps:

Step 1: Build a heavy linear correlation model for the EC CHW performance and related constraints.
CHW'(t) = _CR (t) x CHW _r (t) + CT ( _t ) x TMP (t) + _CE (t) x ELE (t) (Equation 1)
During the ceremony
CHW'(t) represents the calculated cold water supply performance at time t estimated by the correlation model.
CHW _r (t) represents the constraint of the cold water return temperature at time t.
TMP (t) represents the constraint of ambient air temperature at time t.
ELE (t) represents the constraint on the power consumption of the refrigerator at time t.
_CR (t) represents the correlation coefficient of _CHWr (t) at time t generated from multiple linear regression.
CT ( _t ) represents the correlation coefficient of TMP (t) at time t generated from multiple linear regression.
_CE (t) represents the correlation coefficient of ELE (t) at time t generated from multiple linear regression.

Step 2: Calculate the contribution of each constraint to CHW performance. For example, in the case of the constraint _ELE (t), the contribution rate RE (t) may be calculated as follows.
_RE (t) = | _CE (t) | / P (t) (Equation 2)
P (t) = | _CE (t) | + | _CR (t) | + | _CT (t) | (Equation 3)
The contribution rate _RT (t) of the constraint TMP (t) and the contribution rate _RR (t) of the constraint CHW _r (t) may be calculated by the same or similar method.

Step 3: Select the constraint with the highest contribution rate as the bottleneck of CHW performance. During the period of performance anomalies, bottlenecks are considered the root cause of the anomaly.

From steps 1 to 3 above, the RCA diagnostic results are generated by multiple linear regression using the Gaussian distribution family and identity link functions, with correlation coefficients _CR (t), _CT (t), and _CE . It can be seen that it largely depends on (t). At time t, a regression process is performed on the data sample during the period [tw, t] (where w is the window size and is interchangeably referred to as the "sample size" or the like). According to the Okitsu and Zhang literatures, conventional RCA methods utilize a uniform (fixed, predetermined) window size for the regression process. In this regard, embodiments of the present invention have found that such conventional RCA-based performance abnormality diagnostic methods have various problems with low quality, and have identified the lack of window (sample) size control as a factor leading to this. .. For example, conventional RCA-based performance anomaly diagnostic methods use uniform window sizes for time-dependent correlation analysis, but according to embodiments of the invention, for anomalies of different sizes (eg, different durations). It has been found to be inefficient. For example, in a non-limiting embodiment, when the RCA window size is much larger than the anomalous size, the anomalous data is buried in normal data (ie, the portion of the non-abnormal data) and the RCA diagnosis is frozen. It was identified that one or more anomalies in machine performance could not be captured and the root cause of the anomalies could not be identified. On the other hand, if the window size is too small, RCA diagnostics can become unstable, for example, due to correlation overfitting problems that can be caused by too little sample data and too many constraints.

Therefore, various embodiments of the present invention advantageously improve the quality or effectiveness of the chiller performance anomaly diagnosis (eg, stability, accuracy, and / or accuracy), as described above for conventional refrigeration. RCA for refrigerator performance anomaly diagnosis, which seeks to overcome or at least improve one or more of the defects associated with conventional refrigerator performance anomaly diagnostic methods, such as machine performance anomaly diagnostic methods. A method (computer mounting method) and a system for diagnosing an abnormality in refrigerator performance based on the present invention are provided. In particular, the various embodiments of the invention advantageously adapt to variable time dependent anomalies, and more specifically, optimal windows for performing RCA diagnostics for anomalies identified in the performance parameters of the refrigerator. Includes RCA window size management techniques for diagnosing refrigerating performance anomalies to determine size.

FIG. 1 shows a schematic flow chart of a method (computer mounting method) 100 for diagnosing an abnormality in refrigerator performance according to various embodiments of the present invention. Method 100 is 102 with respect to the step of setting a first correlation model between the performance parameters of the refrigerator and the first plurality of operating parameters, and 104 with respect to the anomalies identified by the performance parameters. Steps to determine the optimal window size for performing a root cause analysis (RCA) diagnosis based on a correlation model, and at 106 to determine the root cause of the anomaly identified by the chiller performance parameters. Includes step 1 to perform RCA diagnosis based on the correlation model and optimal window size. In this regard, RCA diagnostics are performed on a dataset containing performance parameters collected over a period of time and measurement data for a first plurality of operational parameters, including anomaly measurement data with anomalies associated with the anomaly. To.

Accordingly, various embodiments of the invention preferably include RCA window size management techniques, and more particularly determine the optimum window size for performing RCA diagnosis for anomalies identified in the performance parameters of the refrigerator. Provides a method for diagnosing abnormal refrigerator performance. As a result, the quality or effectiveness of the freezer performance anomaly diagnosis according to embodiments of the present invention, such as stability, accuracy, and / or accuracy, can be significantly improved, thereby anomalies in the performance of the refrigerator. The root cause can be better and more efficiently identified / detected.

Refrigerator performance parameters include, but are not limited to, cold water supply (CHW), coefficient of performance (COP), or cold water supply temperature (CHWs), which are desirable or appropriate to be considered or measured. It will be appreciated by those skilled in the art that any type of performance may be used with respect to.

Also, the operating parameters may be of any type of operating parameters that are selected or determined to be associated with or related to the performance parameters being considered or measured, eg, to the performance parameters being considered or measured. It will be appreciated by those skilled in the art that there may be multiple operating parameters selected or determined as acting or influencing. As a mere example, the operating parameters are, but not limited to, a group consisting of chilled water return temperature, chilled water flow rate, cooling water supply temperature, cooling water flow rate, steam flow rate, vapor pressure, cooling water return temperature, power consumption, and ambient air temperature. May be selected from.

In various embodiments, the first correlation model may be a multiple linear correlation model between the performance parameters of the refrigerator and the plurality of operating parameters.

It will be appreciated by those skilled in the art that in various embodiments, performance and operating parameters may be measured by one or more suitable sensors known in the art. For example, non-limitingly, in a cooling plant, various sensors may be installed inside or around, as desired or as appropriate, to measure or monitor various performance and operating parameters. , Relevant sensor data (or measurement data) will be collected over time and collected over time in a manner known in the art and not required herein for brevity. Those skilled in the art can recognize that the data may be stored in the data storage medium. For example, the above-mentioned data set containing the performance parameters collected over a period of time and the measurement data related to the first plurality of operating parameters may be obtained from such sensor data collected over time.

In various embodiments, the above-mentioned steps for determining the optimum window size include a step of acquiring a plurality of candidate window sizes and an RCA diagnosis based on a first correlation model for anomalies among the plurality of candidate window sizes. Includes a step of evaluating each of the plurality of candidate window sizes with respect to the first correlation model to determine the optimal candidate window size to perform. For example, a set or range of candidate window sizes may be determined, and each candidate window size for that set or range is then RCA diagnostics using the candidate window size based on the first correlation model. It is evaluated (or assessed) to determine how well it is performed.

In various embodiments, the steps described above for evaluating each of the plurality of candidate window sizes are generated for each candidate window size based on the candidate window size and include at least anomalous data points belonging to the anomaly measurement data. A step of generating a set of anomalous data samples from a dataset, a step of performing a multiple linear regression analysis on a set of anomalous data samples based on a first correlation model, and a step of generating anomalous data samples. Includes a step to determine the overall quality measure in the fit of the first correlation model to the set. Therefore, for each candidate window size to be evaluated, a set of anomalous data samples is generated from the above dataset. In this regard, each anomalous data sample in the set is generated based on the candidate window size and contains at least anomalous data points that belong to the anomaly measurement data (eg, therefore referred to as anomalous data samples). For example, each anomalous data sample in a set is the same size (same period length) determined based on the candidate window size, such as, but not limited to, the candidate window size x the size equal to, but not limited to, the data sampling time interval of the dataset. ) May have. For example, each anomaly data sample in a set may contain different quantities or different parts / parts of anomalous measurement data. Next, a multiple linear regression analysis is performed to obtain each first quality measure (eg, the square of the multiple correlation coefficient (R ² )) in the fit of the first correlation model to the anomalous data sample. It may be performed for each anomalous data sample in the set based on the correlation model of. The overall quality measure (eg, confidence interval (CI)) in the fit of the first correlation model to the resulting set of anomalous data samples may then be determined. As a result, the suitability or performance of each candidate window size out of multiple candidate window sizes is evaluated, and then the optimal candidate window size for performing a root cause analysis diagnosis for the anomaly based on the first correlation model. May be determined.

In various embodiments, a set of anomalous data samples (for each candidate window size) is generated based on a sliding window technique performed over an anomalous period.

In various embodiments, the sliding window technique follows from the step of generating a first anomalous data sample containing the start data points of the anomaly measurement data to the final anomaly data sample containing the final data points of the anomaly measurement data. Includes a step of generating anomalous data samples that follow for a time interval to generate a set of anomalous data samples. For example, if the time interval of the dataset (eg, the data sampling time interval) is 5 minutes, subsequent anomalous data samples may be generated every 5 minutes until the final anomalous data sample.

In various embodiments, the first anomalous data sample has an anomalous sample period ending at the start data point of the anomaly measurement data and the final anomaly data sample has an anomaly sample period beginning at the final data point of the anomaly measurement data. .. In other words, the window (eg, anomalous sample period) may slide from the start point or leftmost of the anomaly measurement data to the end point or rightmost of the anomaly measurement data.

In various embodiments, each anomalous data sample in a set of anomalous data samples has an anomalous sample duration determined based on the candidate window size. For example, as described above, each anomaly data sample in a set is the same anomaly determined based on the candidate window size, such as, but not limited to, the candidate window size x the size equal to, but not limited to, the data sampling time interval of the dataset. It may have a sample period length.

In various embodiments, the above step of performing a multiple linear regression analysis performs a multiple linear regression analysis on the anomalous data samples based on the first correlation model for each anomalous data sample in the set of anomalous data samples. And for each anomalous data sample set of anomalous data samples, determine the first quality measure in the fit of the first correlation model to the anomalous data sample, and then determine the first set of anomalous data sample sets. Includes a step to generate a quality measure of 1. In this regard, therefore, the overall quality scale described above in the fit of the first correlation model to the set of anomalous data samples may be determined based on the set of first quality measures. As mentioned above, the first quality scale may include the square of the multiple correlation coefficient (R ² ) of the first correlation model for the anomalous data sample, and the overall quality scale may be the determined first quality scale. May include a confidence interval (CI) in the set of. In this regard, the multiple linear regression analysis on the anomalous data sample based on the first correlation model may generate the correlation coefficient of the operating parameters, respectively, contained in the first correlation model. The first quality measure may then be determined based on the obtained correlation coefficient.

In various embodiments, the above step of evaluating each of the plurality of candidate window sizes selects one of the plurality of candidate window sizes based on each overall quality measure determined for the plurality of candidate window sizes. It also includes a step to determine the optimal window size. For example, the candidate window size that yields the best overall quality measure determined is determined to be the optimal window size for performing RCA analysis based on the first correlation model for the anomalies identified by the performance parameters. May be done.

In various embodiments, method 100 is a synthetic motion parameter in which the root cause determined from the RCA diagnosis for anomalies based on the first correlation model and optimal window size is synthesized from a second plurality of motion parameters. Further includes determining whether or not. In other words, the RCA diagnosis (eg, first root cause analysis diagnosis) may output the root cause determined for the anomaly (eg, the first root cause), and the determined root cause is. It is determined whether or not the synthetic operation parameter is synthesized (based on a combination thereof) from the second plurality of operation parameters. As a mere example, the synthetic operating parameter may be, but is not limited to, cold water return quality (CHW _q ), cooling water supply quality (COW _q ), or steam supply quality (STM _q ). As an example, CHW _q may be synthesized from CHW _r and CHW _f (eg, CHW _r (t) × CHW _f (t)).

In various embodiments, method 100 comprises a step of determining that the root cause is a synthetic motion parameter (eg, if the root cause (first root cause) is determined to be a synthetic motion parameter) and a synthetic motion. To set up a second correlation model between the performance parameters associated with the parameters and the second plurality of operational parameters, and to perform a second RCA diagnosis for anomalies based on the second correlation model. Based on the second correlation model and the second optimal window size for the anomaly, to determine the second optimal window size step and to determine the second root cause of the anomaly identified in the refrigerator performance parameters. It further comprises the step of performing a second RCA diagnosis. In this regard, the second RCA diagnosis is a second dataset that includes performance parameters collected over a period of time (eg, the period described above) and measurement data for a second plurality of operating parameters, including anomalous measurement data. Is carried out against.

The second optimal window size may be determined in the same or similar manner as the first optimal window size and therefore does not need to be repeated herein for brevity and clarity. The second RCA diagnosis may also be performed in the same or similar manner as the first RCA diagnosis and therefore does not need to be repeated herein for brevity and clarity.

Therefore, in various embodiments, if the root cause determined by the RCA diagnosis for anomalies is a synthetic motion parameter, the additional root cause determined may be interchangeably referred to as a non-synthetic motion parameter (replaceably referred to as a live motion parameter). ), The steps described above may be repeated to determine the further root cause of the anomaly (eg, set up a further correlation model, determine a further optimal window size, and further. Including the step of performing further RCA diagnosis based on the correlation model). As a result, raw or primary operating parameters may be identified as anomalous bottlenecks or root causes that have been found to further improve the quality or effectiveness of the freezer performance anomaly diagnosis.

Therefore, various embodiments of the present invention advantageously improve the quality or effectiveness of the chiller for diagnosing chiller performance anomalies and maintain the chiller more effectively and efficiently. Provided is a method for diagnosing an abnormality in the performance of a refrigerator based on RCA.

FIG. 2 shows a refrigerator according to various embodiments of the present invention, such as those corresponding to the method 100 for diagnosing a refrigerator performance abnormality as described above with reference to FIG. 1. The schematic block diagram of the system 200 for the performance abnormality diagnosis is shown. The system 200 may be configured primarily to perform the refrigerator performance abnormality diagnosis, or as one of a plurality of other functions that can be performed by the system 200, the refrigerator performance abnormality diagnosis is performed. Those skilled in the art will recognize that they may be configured to perform.

The system 200 is communicably connected to the memory 202 and according to various embodiments of the present invention according to the method 100 for diagnosing a refrigerator performance abnormality as described above with reference to FIG. It comprises at least one processor 204 configured to carry out a method for refrigerating performance anomalies diagnosis.

At least one processor 204 is configured to perform a required function or operation through a set of instructions (eg, a software module) that can be performed by at least one processor 204 to perform the required function or operation. Those skilled in the art will recognize that it may be acceptable. Therefore, as shown in FIG. 2, the system 200 is configured to set a first correlation model between the performance parameters of the refrigerator and the first plurality of operating parameters. Optimal window sizing module or circuit configured to determine the optimal window size for performing RCA diagnostics based on the first correlation model for the anomalies identified in the performance parameters with the configuration module or circuit 206. RCA diagnosis is performed on the anomalies based on the first correlation model and the optimal window size to determine the root cause of the anomalies identified in the 208 (also called the window size optimization system) and the performance parameters of the refrigerator. It may further comprise an RCA module or circuit (or referred to as an RCA engine) 210 configured to do so. In this regard, RCA diagnostics is a dataset containing measurement data for performance parameters and first plurality of operating parameters collected over a period of time, including anomaly measurement data with anomalies associated with the anomaly (eg, memory 202). Is remembered in).

The modules described above do not necessarily have to be separate modules, but one or more modules will be one functional module (eg, as desired or appropriate, as desired or appropriate, without departing from the scope of the invention. It will be appreciated by those skilled in the art that it may be implemented by a circuit or software program) or compiled as a single functional module. For example, the first correlation model setting module 206, the optimal window size determination module 208, and the RCA module 210 are stored, for example, in memory 202 to perform the functions / operations described herein according to various embodiments. In addition, it may be implemented as one executable software program (eg, a software application, or simply referred to as an "app"), which may be runnable by at least one processor 204 (eg, compiled with each other). good.

In various embodiments, the system 200 corresponds to method 100 as described above with reference to FIG. 1, and therefore the various functions or operations configured to be performed by at least one processor 204 vary. The various steps or actions of the method 100 described herein above may be accommodated according to certain embodiments and therefore need not be repeated for system 200 for clarity and brevity. In other words, the various embodiments described herein in relation to the method are approximately valid for each system or device and vice versa. For example, in various embodiments, the memory 202 may store a first correlation model setting module 206, an optimal window sizing module 208, and / or an RCA module 210, which are described herein, respectively. Corresponds to the various steps or parts of Method 100 as described above that can be performed by at least one processor 204 to perform the corresponding function / operation as described.

A computing system, controller, microcontroller, or any other system that provides processing power may be provided in accordance with various embodiments of the present disclosure. Such a system may be considered to include one or more processors and one or more computer-readable storage media. For example, the system 200 described herein may include, for example, a processor (or controller) 204 and a computer-readable storage medium (or memory) 202, which are used in various processes performed as described herein. And may be included. The memory or computer-readable storage medium used in various embodiments may be volatile memory, such as DRAM (Dynamic Random Access Memory), or non-volatile memory, such as Programmable Read-Only Memory, EPROM (Erasable DRAM). It may be an EEPROM (Electrically Erasable PROM) or a flash memory, such as a floating gate memory, a charge trap memory, an MRAM (magnetic resistance random access memory), or a PCRAM (phase change random access memory).

In various embodiments, the "circuit" may be understood as any kind of logical implementation entity, which may be a dedicated circuit stored in memory or processor execution software, firmware, or any combination thereof. .. Thus, in one embodiment, the "circuit" is a hardwired logic circuit, or a programmable logic circuit such as a programmable processor, eg, a microprocessor (eg, a complex instruction set computer (CISC) processor, or a reduced instruction set computer (RISC)). It may be a processor). The "circuit" may also be a computer program that uses processor execution software, such as any kind of computer program, such as virtual machine code (eg, Java). Any other type of implementation in each of the functions described in more detail below may also be understood as a "circuit" according to various alternative embodiments. Similarly, a "module" may be part of a system according to various embodiments of the invention, may include a "circuit" as described above, or as any kind of logical implementation entity thereby. May be understood.

Some parts of this disclosure are presented explicitly or implicitly with respect to algorithms and functional or schematic representations of behavior for data in computer memory. These algorithmic descriptions and functional or schematic representations are the means used by those skilled in the art of data processing to most efficiently convey the substance of their research to others. The algorithm is also conceived here and generally as a series of self-consistent steps leading to the desired result. A step is a step that requires physical manipulation of a physical quantity, such as an electrical signal, a magnetic signal, or an optical signal, which can be stored, transmitted, coupled, compared, or otherwise manipulated.

Unless otherwise specified and, as will be apparent from the following, throughout the specification, "set", "determine", "implement", "acquire", "evaluate", Considerations using terms such as "generate" manipulate data represented as physical quantities in a computer system and are similarly represented as physical quantities in a computer system or other information storage device, transmitting device, or display device. It will be recognized that it refers to the operation and process of a computer system or similar electronic device that translates into other data.

The present specification also discloses a system, device, or apparatus for carrying out the operation / function of the methods described herein. Such systems, devices, or devices may comprise a general purpose computer or other device that may be specially constructed for a required purpose or that may be selectively actuated or reconfigured by a computer program stored in the computer. You may. The algorithms presented herein are not inherently relevant to any particular computer or other device. Various general purpose machines may be used in conjunction with computer programs as taught herein. Alternatively, it may be appropriate to build a more specialized device to carry out the required method steps.

In addition, the specification is also a computer program or software / software in that it will be apparent to those of skill in the art that the individual steps of the methods described herein may be performed by computer code. Disclose functional modules at least implicitly. Computer programs are not limited to any particular programming language and its implementation. It will be appreciated that various programming languages and their coding may be used to implement the teachings of the disclosures contained herein. Further, the computer program is not limited to any particular control flow. There are many other variants of computer programs that can use different control flows without departing from the spirit or scope of the invention. The various modules described herein (eg, first correlation model setting module 206, optimal window sizing module 208, and / or RCA module 210) can be performed by a computer processor to perform the required functions. It may be a software module implemented by a computer program or instruction set, or it may be a hardware module, which is a functional hardware unit designed to perform the required functions. It will be recognized by those skilled in the art. It will also be recognized that a combination of hardware and software modules may be realized.

Moreover, one or more of the steps in the computer programs / modules or methods described herein may be performed in parallel rather than sequentially. Such computer programs may be stored on any computer readable medium. Computer-readable media may include storage devices such as magnetic or optical discs, memory chips, or other storage devices suitable for interacting with general purpose computers. The computer program is loaded into such a general purpose computer and, when executed there, effectively provides a device that implements the steps of the method described herein.

In various embodiments, instructions (eg, first) that can be executed by one or more computer processors to implement method 100 for refrigerating performance abnormality diagnosis as described above with reference to FIG. Embodied in the form of one or more computer-readable storage media (non-temporary computer-readable storage media), including the Correlation Modeling Module 206, Optimal Window Sizing Module 208, and / or RCA Module 210). Computer programming products are provided. Accordingly, the various computer programs or modules described herein are received by a computer system or electronic device therein, as they are performed by at least one processor of the system or device to perform the required or desired function. It may be stored in a possible computer program product.

The software or functional modules described herein may also be implemented as hardware modules. More specifically, in the hardware sense, a module is a functional hardware unit designed for use with other components or modules. For example, the module may be implemented using discrete electronic components, or may form part of an entire electronic circuit, such as an application specific integrated circuit (ASIC). There are many other possibilities. Those skilled in the art will recognize that the software or functional modules described herein can also be implemented as a combination of hardware and software modules.

It will be appreciated by those skilled in the art that the terminology used herein is merely for the purpose of describing various embodiments and is not intended to limit the invention. As used herein, the singular forms "a," "an," and "the" shall also include the plural, unless otherwise stated in the context. Further, the terms "equipped" and / or "equipped", as used herein, specify the existence of defined mechanisms, integers, steps, actions, elements, and / or components. It is understood that does not preclude the existence or addition of one or more other mechanisms, integers, steps, actions, elements, components, and / or groups thereof. Will be.

FIG. 3 is a schematic diagram of a window size optimization system 208 for performing RCA diagnostics according to various embodiments of the invention, such as those corresponding to the steps described above for determining the optimal window size described for Method 100. Shows. The window size optimization system 208 has a candidate acquisition module or circuit 302 configured to acquire a plurality of candidate window sizes, and a plurality of candidate window sizes based on a first correlation model for anomalies. To determine the optimal candidate window size for performing the RCA diagnosis, a candidate evaluation module or circuit 304 configured to evaluate each of the plurality of candidate window sizes with respect to the first correlation model and the determined optimal. It comprises an output module or circuit 306 configured to output the window size.

In various embodiments, the window size optimization system 208 corresponds to the steps described above for determining the optimal window size described for method 100 and thus, when performed by a processor (eg, processor 204), window size optimization. The various functions or behaviors configured to be performed by Chemical System 208 may correspond to the various steps or parts of the above-mentioned steps that determine the optimal window size described for Method 100 and are therefore clear and clear. For the sake of brevity, there is no need to repeat for the window size optimization system 208. The window size optimization system 208 implements a software module (eg, a computer program) that can be executed by a processor, or implements it, in order to implement the method of determining the optimum window size described above according to various embodiments of the present invention. It will be appreciated that it may be implemented as a combination of hardware modules (eg, circuits) configured to do so, or hardware and software modules configured to implement it.

To facilitate understanding and practicing the invention, various exemplary embodiments of the invention are described below, but not exclusively, by way of example only. However, it will be appreciated by those skilled in the art that the invention may be embodied in a variety of different forms or configurations and should not be construed as being limited to the exemplary embodiments described below. Let's do it. Rather, these exemplary embodiments are provided so that the present disclosure is thorough and complete, and that the scope of the invention is fully communicated to those of skill in the art.

Various exemplary embodiments of the invention provide RCA window size control methods in refrigerator diagnostics for adapting to variable time dependent anomalies. In various exemplary embodiments, the method addresses, for example, unstable diagnostic problems that can be caused by small RCA window sizes, two-layer (or multi-layer in the case of two or more layers) diagnostics. Form a model (RCA or correlation model). For example, in the case of a two-layer RCA model, the upper layer (or first layer) of the RCA model can be combined by combining the relevant raw sensor constraints (non-synthetic constraints) on the lower layer (or second layer) of the RCA model. A synthetic refrigerating machine constraint may be built on top, which effectively reduces the number of constraints used in the RCA correlation analysis, so that smaller window sizes fit better and are more stable for correlation analysis. The diagnosis is provided. The method may further determine the optimal RCA window size for performing an RCA diagnosis based on the RCA model for anomalies, eg, based on the correlation coefficient (determined for the correlation model) and the confidence interval (CI). good. The method increases diagnostic accuracy by, for example, rejecting window sizes and anomalous samples that have a low correlation to the performance of the refrigerator being considered, and by selecting the optimal window size with the narrowest CI. It has been found to increase accuracy.

Accordingly, various exemplary embodiments of the invention provide quantitative window size management methods for RCA diagnostics, including multi-layer (eg, bi-layer) diagnostic models and window size estimation or optimization methods. Various exemplary embodiments of the invention address the poor quality of RCA-based refrigerator anomaly diagnostics and control RCA window sizes based on correlation coefficients and multi-layer (eg, bi-layer) diagnostic models. , Methods and computer systems.

FIG. 4 shows a schematic representation of an example system architecture 400 for RCA diagnostics for a refrigerator, including window size calculation / optimization, according to various exemplary embodiments of the invention. The system architecture 400 may include an operator (eg, an operator device) 401, a refrigerating machine 402, a server 403, and workstations 411 to 412, all of which are local area networks (LANs) or wide area networks (LANs). WAN) and the like, but not limited to them, may be connected via a network 415. The operator 401 may be one or more terminals used by the refrigerator operator (eg, maintenance manager) to estimate the optimal window size and initiate the RCA diagnosis, eg, a personal computer. It may be a computing device such as a tablet or a smartphone. The refrigerator 402 may provide operation data of the refrigerator used in the network 415 through a sensor data collection system such as, for example, an OSIsoft PI system. The server 403 stores instructions that can be executed by the processor in the server 403 to determine the optimal window size (eg, software module (program logic) 404) and perform RCA diagnostics based on the optimal window size. It may be implemented by a computer system (eg, corresponding to the system 200 described above with reference to FIG. 2 according to various embodiments) (eg, RCA engine 405). Server 403 also provides the operator 401 with refrigerating machine operation data 406 (eg, corresponding to the "data set containing measurement data" described above according to various embodiments) and a graphic user interface (GUI). It may store instructions that can be executed by the internal processor.

The operator 401 may display various refrigerator operation data from the server 403 through the GUI presented on the operator 401, where the refrigerator operator defines an RCA model (correlation model) and sets the anomaly period. It may be specified and submitted to server 403 for window size estimation and RCA diagnostic requests. In various exemplary embodiments of the invention, the operator 401 may include multiple terminals throughout the network 415, which allows remote access to the server 403 from various locations.

For example, the refrigerator 402 may represent the SAC or EC of a cooling plant, such as, but not limited to, an OSIsoft PI system that provides a uniform data access interface to the server 403 and other components through network 415. A data collection system may be used. More than one chiller (ie, multiple chillers) may be operating in the cooling plant, so the server 403 provides window size estimation or optimization and RCA diagnostics for each chiller. It will be appreciated by those skilled in the art that it may be configured as such. In other words, the methods and systems for diagnosing refrigerating performance abnormalities according to various embodiments of the present invention may be applied to any one or more as desired or as appropriate. , Will be recognized by those skilled in the art.

In various exemplary embodiments, upon receiving an RCA window size estimation request from the operator 401, the server 403 may execute program logic 404 and return the determined optimal window size to the operator 401. If the request from the operator 401 was for a refrigerator anomaly diagnosis, the server 403 was based on a correlation model (RCA model) defined and an optimal window size determined by running the RCA engine 405. You may proceed to the implementation of RCA diagnosis. The RCA engine 405 is a technique known in the art, such as that described in the Okitsu or Zhang literatures, provided that it is determined according to various exemplary embodiments for improving the quality of RCA diagnostics. The optimal window size may be used to configure RCA diagnostics. The server 403 may locally host the (stored) refrigerator operation data 406, which can be automatically or manually collected from the refrigerator 402, or may be stored remotely (eg, in a storage server). It will be appreciated by those skilled in the art that the refrigerator operation data 406 may be acquired.

In various exemplary embodiments, the server 403 and operator 401 may be implemented on different computers through network 415, in which case the transmission of requests and results between them is a RESTful web service or remote procedure call ( It may be realized through RPC). If the computational load is heavy, the server 403 may distribute some window size computation tasks to one or more workstations 411-412. In various exemplary embodiments, various distributed, concurrent, or cloud-based computing systems, such as Hadoop, reduce window size estimation time and facilitate real-time diagnostics on server 403 and workstation 411. It may be used between ~ 412. In various other exemplary embodiments, for example, for smaller applications, the operator 401 and the server 403 may both be implemented on the same computer.

In various exemplary embodiments, the program logic 404 configured to calculate the optimal window size for RCA diagnosis is implemented by a suitable computer language known in the art, such as an editing language or an interpreting language. You may. In various exemplary embodiments, program logic 404 may be stored in server 403 and executed there. In various exemplary embodiments, the program logic 404 may be distributed within one or more of servers 403 and workstations 411-412 in a distributed computing schema that speeds up computation. In such cases, the server 403 and one or more workstations 411 to 412 may both correspond to the system 200 described above with reference to FIG. 2 in various embodiments.

FIG. 5 is an exemplary GUI presented to and utilized by operator 401 to define an RCA model 502 and an anomalous period of RCA window size estimation according to various exemplary embodiments of the invention. Shows 500. For example, the operator 401 may define or select the refrigerator performance that should be considered with respect to the refrigerator, and the relevant constraints in the RCA model 502. As a mere example, sensors relating to refrigerator performance and constraints may be selected through their respective pull-down menus 503, without limitation. For example, for chiller performance, performance parameters may be selected from three possible options, including cold water supply (CHW), coefficient of performance (COP), and cold water supply temperature (CHWs). For constraints, operator 401 is two or three constraints in each column of constraints (eg, those represented in FIG. 5 as "constraint 1", "constraint 2", and "constraint 3"). May define a plurality of constraints (eg, corresponding to the "first plurality of operating parameters" described above according to various embodiments), each of which is one in each column. It may be defined by one or more sensors. For example, if the sequence of constraints contains only one sensor, the constraints may be represented by the data (operation parameters) of that sensor in the RCA model. If the constraint column contains more than one sensor, the constraint is a composite constraint based on a combination of data from two or more sensors (two or more operating parameters), such as multiplication of data from two or more sensors. (For example, corresponding to the "second plurality of operating parameters" described above according to various embodiments). As a mere example, FIG. 5 shows that constraint 1 and constraint 3 are synthetic constraints based on the two sensors, respectively.

In various exemplary embodiments, one or more appropriate weights or offsets are applied to the data of one or more sensors, respectively, obtained taking into account various metrics, data compatibility, etc., respectively. May be applied. For example, the text field 506 is for each sensor to allow measurement / sensor data to be adjusted / modified as desired or appropriately, such as to form a more complex model. It may be provided adjacent to the sensor. As an example, an offset of "50" may be applied to the sensor data from sensor 1, thereby constructing constraint 1 by the combination of (50-sensor 1 sensed value) and sensor 2 sensed value. ..

As shown in FIG. 5, the GUI 500 also plots the time series performance of the selected chiller based on the chiller raw data (measurement / sensor data) obtained by the operator 401 in the performance field 510. It may be displayed in. The GUI 500 may automatically update the time series performance of the refrigerator according to the type of performance selected in the RCA model 502. In various exemplary embodiments, the operator 401 may select or identify a portion 512 of the measurement data relating to the anomaly, which portion 512 may be referred to as anomaly measurement data having an anomaly duration associated with the anomaly. The selection may be performed manually by the user based on user input (eg, via a mouse device), or may be automatically identified / detected by operator 401 or server 403. Next, the period of the anomaly may be displayed in the anomaly timetable 514. The operator 401 may also directly change / adjust the time shown in the timetable 514, whereby T1 represents the abnormal start time and T2 represents the abnormal end time. Any change in the anomaly period in the timetable 514 may be automatically synchronized in the display of the period of the anomaly measurement data 512 in the GUI 500.

As shown in FIG. 5, an RCA start button 520 may also be provided that allows the operator 401 to send an RCA request to the server 403 and initiates an RCA diagnosis when started. The RCA request may include information about the defined RCA model 502 and anomalous time information from the specified anomaly period. For example, upon receiving an RCA request, the server 403 starts the RCA engine 405 with the defined RCA model 502 and anomaly time information and RCA for anomalies 512 identified with the selected performance parameters. Diagnosis may be performed. In this regard, the server 403 that receives the information in the RCA model 502 may thereby be configured to use the RCA model in subsequent processing, as defined in the information in the received RCA model 502. ..

As shown in FIG. 5, the diagnostic result field 522 contains two layers of the RCA model (two-layer RCA model) 502 (eg, "first correlation" as described above according to various embodiments of the invention. Optimal window size for the upper layer or first layer model corresponding to the "model" and the lower layer or second layer model corresponding to the "second correlation model", and the bottleneck of the refrigerator abnormality 512 (root cause). ) And various diagnostic outputs / results received from the server 403 may be displayed. The diagnostic results are determined by the window size of the upper model field 516, which displays the optimal window size of the dual tier RCA model 502 as determined by the server 403 for the upper tier model, and by the server 403 for the lower tier model. It may also include the window size of the lower model field 517, which displays the optimal window size. For example, the window size of the lower model field 517 may display N / A or 0 hours if there is no compositing constraint identified as a bottleneck. The diagnostic result may further include a bottleneck field 519 indicating the sensor (eg, sensor ID or name) determined by the server 403 as the root cause of the refrigerator anomaly 512.

Accordingly, in various exemplary embodiments, there is provided a method of diagnosing refrigerating machine anomalies based on a two-layer RCA model, which may be implemented by program logic 404. FIG. 6 shows a schematic flow diagram of a method 600 for diagnosing an abnormality in a refrigerator based on an RCA model (having multi-layer RCA model capability) according to various exemplary embodiments of the present invention.

In step 604, method 600 follows the defined RCA model 502 (eg, a multi-layer RCA model such as a two-layer RCA model) with a higher or first layer of the RCA model (as described above according to various embodiments). Form or set (corresponding to the first correlation model). For purposes of illustration only, without limitation, FIG. 7 shows a two-layer RCA model 700 comprising an upper layer or first layer 704 and a lower layer or second layer 706. The lower layer 706 has a cold water return temperature (CHW _r ), a cold water flow (CHW _t ), a cooling water supply temperature (COW _s ), a cooling water flow rate (COW _f ), a steam flow rate (STM _r ), and a steam pressure (STM _p ). ), Cooling water return temperature (COW _r ), power consumption (ELE), ambient temperature (AMB _t ), and other refrigerating-related (eg, corresponding to each sensor) multiple "raw" operating parameters (non-synthetic). There are operating parameters (constraints). In the upper layer 704, one or more synthetic operation parameters may be constructed by combining the related operation parameters from the lower layer 706. If each operating parameter (constraint) selected in the RCA model 502 is a non-synthetic operating parameter (eg, including only one sensor), the multi-layer (eg, double-layer) RCA model is reduced to a single-layer RCA model. Those skilled in the art will recognize that it may be acceptable.

Multi-layer (eg, two-layer) RCA models have advantages over conventional single-layer RCA models disclosed in the Okitsu and Zhang literatures. For example, traditional RCA models may have to choose constraints from too many raw sensors and miss important sensors if not enough are selected, or especially if the window size is small. Too many constraints in the RCA model can lead to unstable correlation results (for example, sometimes referred to as an overfit problem). By replacing the raw sensor constraints in the RCA model with the relevant synthetic constraints, the multi-layer RCA model effectively reduces the number of constraints in the correlation analysis without losing the scope of the sensor information, thereby advantageously. Maintain the stability of correlation analysis and diagnosis.

By way of example only, but not limited to, three exemplary synthetic constraints may be defined as set forth in Equations 4, 5, and 6 to diagnose the most frequently reported anomalies.
CHW _q (t) = CHW _r (t) × CHW _f (t) (Equation 4)
During the ceremony
CHW _q (t) represents a synthetic constraint that represents cold water return quality.
CHW _r (t) represents a raw sensor constraint representing the cold water return temperature at time t.
CHW _f (t) represents a raw sensor constraint representing the cold water flow rate at time t.

In Equation 4, CHW _q (t) is defined as the product of CHW _r (t) and CHW _f (t), which, for example, has higher cold water return quality due to higher cold water return temperature or higher cold water flow. It may mean that. In the RCA diagnosis, if CHW _q is selected as one of the constraints in the RCA model and identified as a bottleneck in refrigerator performance, then for lower or additional layers of the RCA model with CHW _r and CHW _f constraints. Further RCA diagnosis is performed to determine that one of the two constraints is the root cause of the refrigerator performance or anomaly.
COW _q (t) = (T _top -COW _s (t)) * COW _f (t) (Equation 5)
During the ceremony
COW _q (t) represents a synthetic constraint that represents the cooling water supply quality.
T _top represents a constraint that represents a temperature higher than any cooling water supply temperature.
CHW _s (t) represents a raw sensor constraint that represents the cooling water supply temperature at time t.
CHW _f (t) represents a raw sensor constraint representing the cooling water flow rate at time t.

In Equation 5, COW _q (t) is proportional to COW _f (t) but inversely proportional to COW _s (t), which is, for example, due to more cooling water flow or lower cooling water supply temperature. May mean that is higher. Similarly, in RCA diagnostics, if CHW _q is selected as one of the constraints in the RCA model and identified as a bottleneck in refrigerator performance, it is a sublayer of the RCA model with COW _s and COW _f constraints or even more. Further RCA diagnosis is performed on the layer to determine that one of the two constraints is the root cause of the refrigerator performance or anomaly.
STM _q (t) = STM _f (t) * STM _p (t) (Equation 6)
During the ceremony
STM _q (t) represents a synthetic constraint that represents steam supply quality.
STM _f (t) represents a raw sensor constraint that represents the steam flow rate at time t.
STM _p (t) represents a raw sensor constraint representing the vapor pressure at time t.

In Equation 6, STM _q (t) is defined for SAC. Equal to the product of STM _f (t) and STM _p (t), which may mean, for example, higher steam flow rates or higher vapor pressures result in higher steam supply quality. Similarly, in RCA diagnostics, if STM _q is selected as one of the constraints in the RCA model and identified as a bottleneck in refrigerator performance, it will be added to the lower layer RCA model with STM _f and STM _p constraints. RCA diagnosis is performed to determine that one of the two constraints is the root cause of the refrigerator performance or anomaly.

In step 606, the RCA model provided by step 604 (eg, the model of the upper layer or the first layer) or the RCA model provided by step 612 (eg, the lower layer or the second layer, if created). The optimum window size is calculated for the model). Step 606 will be described in more detail with reference to FIG.

At step 608, method 600 calls the RCA diagnostic program and uses the optimal window size determined from step 606. For example, the RCA diagnostic program may be hosted on server 403 and implemented according to the RCA diagnostic method as described in the Okitsu or Zhang literature, using the optimal window size determined from step 606.

Step 610 determines if further RCA diagnosis is needed for the lower layers (or more layers) of the RCA model to discover the root cause of the anomaly. If the bottleneck diagnosed by the RCA diagnostic program is a synthetic constraint, the method may proceed to step 612 to perform further RCA diagnostics on the lower layer (or additional layer) model with respect to the synthetic constraint. .. Otherwise, if the bottleneck is already a "raw" sensor (a "raw" operating parameter), the method may simply proceed to step 614.

Step 612 builds a lower layer (or further layer) of the RCA model according to the definition of the synthetic constraint diagnosed in step 608 as a bottleneck. For example, in the lower layer, only "raw" sensors within the bottleneck constraint are selected to form new constraints on the model, and each constraint contains only one "raw" sensor. Step 612 then sends the newly constructed lower layer of the RCA model (eg, corresponding to the "second correlation model" as described above according to various embodiments) to step 606 to estimate the window size. Or perform optimization.

At step 614, the bottleneck diagnosed by the RCA diagnostic program is identified as the root cause of the anomaly. Next, the diagnostic result is sent to the operator 401 and displayed in the diagnostic result field 522 of the GUI 500.

FIG. 8 shows a schematic flow diagram 800 of Method 800 for calculating the optimal window size for performing an RCA diagnosis for anomalies based on a correlation model with respect to various exemplary embodiments of the invention (eg, FIG. 8). Corresponding to step 104 for determining the optimum window size described above with reference to FIG. 1 and step 606 for calculating the optimum window size 606 described above with reference to FIG. 6). In the example shown in FIG. 8, the method 800 for calculating the optimal window size, without limitation, may be based on R ² (square of the multiple correlation coefficient) and CI for more precise and accurate RCA diagnosis. ..

In step 804, a plurality of sets of anomalous data samples are generated from the refrigerator operation data 406, and each set is generated for one of the plurality of candidate window sizes. For example, Equation 7 below may define a possible range or set of candidate window sizes (W), thereby defining the window size with respect to the number of intervals (Δ).
W = i-1
4 × C ≦ i ≦ 2 × (A + 1) (Equation 7)
φ = {W}
During the ceremony
i represents the number of the data point in the abnormal data sample.
C represents the number of constraints in the RCA diagnostic model (eg C ∈ {2,3}).
Δ represents the time interval of data sampling (for example, 5 minutes).
A represents the abnormal size measured by the number of Δ.
φ represents a set of W.

Based on Equation 7, the duration of the anomalous sample may be calculated as W × Δ, where W is, for example, in the range 4 × C-1 to 2 × A + 1. For example, the lower bound of the window size may be determined to ensure that the anomalous sample has minimal data that meets the constraint correlation analysis requirements in RCA, otherwise the RCA may have an unstable correlation or May have diagnostic results. The upper limit of the window size is to prevent the window size from becoming too large, thereby including too much normal data (non-abnormal data) in all anomalous data samples, making the anomalous data inconspicuous or buried in the normal data. It may be decided. Various exemplary embodiments have only one window size choice if the upper limit is lower than the lower limit, for example, in the case of a very narrow freezer performance spike anomaly, ie with a lower limit such as size 4xC-1. Notice that there are things. The lower bounds and upper bounds described above are for illustration purposes only, and the lower bounds and upper bounds may be defined differently, as desired or as appropriate, without departing from the scope of the invention. Good things will be recognized by those skilled in the art.

In various exemplary embodiments, anomalous samples may be constructed for each window size W ∈ φ by selecting the operation data of the refrigerator through a sliding window over an anomalous period (sliding window technique). An example sliding window technique is shown in FIG. In various exemplary embodiments, the first anomalous sample 902 is formed by selecting refrigerator data from T ₀ to T ₀ + W × Δ. The data at T ₀ + W × Δ is the first data point 903 of the anomaly measurement data having an anomaly period. For example, this ensures that all anomalous samples contain at least one anomalous data point. Further, the abnormal period may be expressed as T ₀ + W × Δ to T ₀ + (W + A) × Δ. For illustration purposes only, Table 1 below shows an example data format of a data sample at each data point of the anomalous sample, including multiple fields. For example, the "time" field may indicate the data sampling time, the "performance" may indicate the selected performance parameter, and a plurality of constraints including "Constraint_1", "Constraint_2", and "Constraint_3". May indicate the constraints selected for the RCA model. In this example, the performance of the selected refrigerator is CHW with the unit RT, the selected Constraint_1 is CHWr with the unit of temperature (° C), the selected Constraint_1 is the AMBt with the unit ° C, selected. Temperature_3 is an ELE having the unit KW.

続いて、ウィンドウは、第２の異常サンプル９０４を構築するため、［Ｔ_０，Ｔ_０＋Ｗ×Δ］から［Ｔ_０＋Δ，Ｔ_０＋（Ｗ＋１）×Δ］までスライドされてもよい。次に、Ｔ_０＋（Ｗ＋Ａ）におけるデータ点が異常期間の最終データ点９１０である、ウィンドウ［Ｔ_０＋（Ｗ＋Ａ）×Δ，Ｔ_０＋（２Ｗ＋Ａ）×Δ］を有する最終異常サンプル９０８まで、後に続く各時間間隔に対してプロセスを繰り返してもよい。

Subsequently, the window may be slid from [T ₀ , T ₀ + W × Δ] to [T ₀ + Δ, T ₀ + (W + 1) × Δ] to construct the second anomalous sample 904. Next, up to the final anomalous sample 908 with the window [T ₀ + (W + A) × Δ, T ₀ + (2W + A) × Δ], where the data point at T ₀ + (W + A) is the final data point 910 of the anomaly period. , The process may be repeated for each subsequent time interval.

For each W of φ, step 806 forms a refrigerator performance correlation model through multiple linear correlation regression for the anomalous sample of W (a set of anomalous data samples) according to Equation 1, and then for that set. For each anomalous sample, R ² (square of multiple correlation coefficient) (eg, corresponding to the "first quality measure" described above according to various embodiments) is calculated. By way of example only, without limitation, Equation 8 below provides an example ^R2 calculation according to various exemplary embodiments for a first anomalous sample s ₁ with a window [T ₀ , T ₀ + W × Δ]. show. In this regard, R ² ranges from 0 to 1, and the closer the value is to 1, the better the RCA correlation model fits the anomalous sample.

（式８）
式中、
ｉは、異常サンプルにおけるデータ点の連続番号を表し、
Ｗは、異常サンプルに対するウィンドウサイズ（Ｗ∈φ）を表し、
ｓ_１は、ウィンドウサイズＷの下で生成された第１の異常サンプルを表し、
Δは、異常サンプルにおける２つの隣接したデータ点間の時間間隔を表し、
Ｔ_０は、異常サンプルにおける第１のデータ点の時間スタンプを表し、
ＳＳＥは、残差平方和を表し、
ＳＳＴＯは、総平方和を表し、
ＣＨＷ（ｔ）は、冷水流量および冷水温度差から導き出した、時間ｔにおける、観察された冷水供給性能を表し、
ＣＨＷ′（ｔ）は、時間ｔにおける相関モデルによる、計算された冷水供給性能を表し、
オーバーライン付きのＣＨＷは、異常サンプルＣＨＷ（ｔ）（ｔ＝Ｔ_０＋Δ，…，Ｔ_０＋Ｗ×Δ）の平均を表す。

(Equation 8)
During the ceremony
i represents the serial number of the data points in the abnormal sample.
W represents the window size (W ∈ φ) for the anomalous sample.
s ₁ represents the first anomalous sample generated under window size W.
Δ represents the time interval between two adjacent data points in the anomalous sample.
T ₀ represents the time stamp of the first data point in the anomalous sample.
SSE represents the residual sum of squares
SSTO stands for total sum of squares
CHW (t) represents the observed cold water supply performance at time t, which is derived from the cold water flow rate and the cold water temperature difference.
CHW'(t) represents the calculated cold water supply performance by the correlation model at time t.
The CHW with an overline represents the average of the abnormal sample CHW (t) (t = T ₀ + Δ, ..., T ₀ + W × Δ).

Similarly, the ^R2 value of other anomalous samples generated under the same window size W may be calculated. For each window size W, a sampling distribution of R ² values is obtained, and each R ² corresponds to anomalous samples generated during the same window sliding process.

Step 808 may calculate the mean value of R ² for a set of anomalous samples generated for a particular window size W and represent it as R ² (W) with an overline.

（式９）
式中、
Ｎは、ウィンドウサイズＷを有する異常サンプルの数を表し、
Ｒ^２（ｓ_ｉ）は、ウィンドウサイズＷを有するｉ番目の異常サンプルに対するＲ^２を表す。

(Equation 9)
During the ceremony
N represents the number of anomalous samples with a window size W.
R ² (s _i ) represents R ² for the i-th anomalous sample having a window size W.

Step 810 determines whether the window size is acceptable or not according to the correlation strength of the performance model. For example, if R ² (W) with an overline on the left side of Equation 9 is less than the threshold K, then the relevant performance model based on the corresponding window size W has a weak linear correlation with the actual anomalous data. May be considered to reduce the accuracy of further RCA diagnosis. Therefore, the corresponding window size W may be removed from the current set φ in step 812, and then method 800 may proceed to step 818. Otherwise, the corresponding window size W may be allowed or maintained in the current set φ, and then method 800 may also proceed to step 818. As a non-limiting example, the default value of the threshold value K may be set as 0.5, but may be adjusted as desired or appropriately based on a specific refrigerator, abnormal conditions, and the like.

Step 818 determines whether or not the loop started in step 806 ends. For example, if all Ws in the set φ have been processed by steps 806-810, then method 800 may proceed to step 814, otherwise method 800 may return to step 806.

For the remaining candidate window sizes W of the current set φ, for each window size W, step 814 estimates the quality of R ² determined for the set to represent the true value. Calculate the confidence interval (CI) of R ² of the anomalous sample determined for the set of anomalous samples obtained for the corresponding window size W. For example, a 95% CI may be calculated for ^R2 of anomalous samples determined for a set of anomalous samples. The 95% CI may be expressed as a range of ^R2 values, that is, the probability that the CI will contain the true value of ^R2 for the entire set of anomalies is 95%. Therefore, the narrower the 95% CI, the higher the accuracy of R ² and the more stable the result. By way of example only, but not limited to, the following shows an exemplary calculation of 95% CI for ^R2 .

In various exemplary embodiments, when the number of anomalous samples (N) in the set is large (eg, N ≧ 30), the distribution of R ² approximates normally, so the 95% CI of R ² is the formula below. It may be calculated as indicated by 10.

（式１０）
式中、
Ｎは、ウィンドウサイズＷを有する異常サンプルの数を表し、
Ｒ_ｉ ^２は、ウィンドウサイズＷを有するｉ番目の異常サンプルに対するＲ^２を表し、
オーバーライン付きのＲ^２は、ウィンドウサイズＷを有する異常サンプルに対するＲ^２の中間値（平均）を表し、
１．９６は、９５％ＣＩに対応するＺスコア数を表し、
σ_Ｒ ^２は、ウィンドウサイズＷを有する異常サンプルに対するＲ^２の標準偏差を表す。

(Equation 10)
During the ceremony
N represents the number of anomalous samples with a window size W.
R _i ² represents R ² for the i-th anomalous sample having a window size W.
R ² with an overline represents the median value (mean) of R ² for anomalous samples with a window size W.
1.96 represents the number of Z scores corresponding to 95% CI.
σ _R ² represents the standard deviation of R ² for anomalous samples with a window size W.

When the number of anomalous samples (N) is small (eg, N <30), the distribution of R ² is t distribution instead of normal distribution, so the calculation of 95% CI for R ² is shown in Equation 11 below. It may be changed to be.

（式１１）
式中、
ｔ_ｓは、９５％ＣＩに対応するｔスコア数を表す。例えば、Ｎ＝２０のとき、ｔ_ｓ＝２０９３である。

(Equation 11)
During the ceremony
t _s represents the number of t scores corresponding to 95% CI. For example, when N = 20, ts = ₂₀₉₃ .

From equations 9, 10, and 11, the value N is widely used in the calculation of the median value of ^R2 and 95% CI, N = W + A + 1 (as shown in FIG. 9), where N is by W and A. As determined, one of ordinary skill in the art will recognize that the window size determination based on the median of 95% CI and R ² already takes into account the information of the variable outlier size A.

Step 816 determines the optimal window size based on the 95% CI of the anomalous sample. If the width of 95% CI (W) is the narrowest for all window sizes in the current set φ and associated anomalous samples, then that particular window size W is the most accurate correlation model for the coefficient ^R2 . Since it is suggested that anomalous samples with can be constructed, the most stable diagnostic capability for RCA is provided and a particular window size W may be selected as optimal.

Therefore, various embodiments of the present invention are advantageous over conventional RCA diagnostic methods as they provide a quantitative method of controlling the window size of RCA diagnostics to adapt to variable refrigerator anomalies. In addition, various embodiments of the invention further improve diagnostic stability through multi-layer (eg, bi-layer) RCA modeling and improve diagnostic accuracy by rejecting window sizes and samples that are less correlated with anomalies. Improve diagnostic accuracy by choosing the optimal window size, such as the one with the narrowest confidence interval.

The following examples relate to various embodiments of the invention, including, for example, further exemplary embodiments.

In Example 1, a computer mounting method for diagnosing refrigerating performance anomalies is disclosed, the method comprising setting a first correlation model between a refrigerating machine performance parameter and a first plurality of operating parameters. For the anomalies identified by the performance parameters, the steps to determine the optimal window size for performing the root cause analysis diagnosis based on the first correlation model and the root causes of the anomalies identified by the performance parameters of the refrigerator. To determine, a step of performing a root cause analysis diagnosis based on a first correlation model and optimal window size for anomalies, wherein the root cause analysis diagnosis contains anomaly measurement data with anomalies associated with the anomaly. Includes steps performed on a dataset containing performance parameters collected over a period of time and measurement data for a first plurality of operating parameters.

In Example 2, the method for diagnosing an abnormality in refrigerating performance according to Example 1 is disclosed, and the above-mentioned steps for determining the optimum window size include a step of acquiring a plurality of candidate window sizes and a plurality of candidate window sizes. From, in order to determine the optimum candidate window size for performing the root cause analysis diagnosis based on the first correlation model for the anomaly, the step of evaluating each of the plurality of candidate window sizes for the first correlation model. include.

In Example 3, the method for diagnosing freezing performance abnormalities according to Example 2 is disclosed, and the above-mentioned steps for evaluating each of the plurality of candidate window sizes are performed for each candidate window size based on the candidate window size. A set of anomalous data samples generated from the dataset, each containing at least anomalous data points belonging to the anomalous measurement data, and a multiple linearity with respect to the set of anomalous data samples based on the first correlation model. It involves performing a regression analysis and determining an overall quality measure in the fit of the first correlation model to the set of generated anomalous data samples.

Example 4 discloses a method for diagnosing freezing performance anomalies according to Example 3, where a set of anomalous data samples is generated based on a sliding window technique performed over an anomalous period.

In Example 5, the method for refrigerating performance abnormality diagnosis according to Example 4 is disclosed, and the sliding window technique includes a step of generating a first abnormality data sample including a start data point of abnormality measurement data, and an abnormality measurement data. Includes a step of generating subsequent anomalous data samples for each subsequent data interval to generate a set of anomalous data samples, up to the final anomalous data sample containing the final data points of.

In Example 6, the method for refrigerating performance abnormality diagnosis according to Example 5 is disclosed, the first abnormality data sample has an abnormality sample period ending at the start data point of the abnormality measurement data, and the final abnormality data sample is. , Has an anomalous sample period starting at the final data point of the anomaly measurement data.

In Example 7, a method for diagnosing a freezing performance abnormality according to any one of Examples 4 to 6 is disclosed, and each abnormality data sample in the set of abnormality data samples is an abnormality determined based on the candidate window size. Has a sample period length.

In Example 8, a method for diagnosing anomalous refrigeration performance according to any one of Examples 3 to 7 is disclosed, and the above-mentioned step of performing a multiple linear regression analysis is each anomalous data sample of a set of anomalous data samples. For each anomalous data sample in the step of performing a multiple linear regression analysis on the anomalous data sample based on the first correlation model, and for each anomalous data sample, the first correlation model for the anomalous data sample. The overall fit of the first correlation model to the set of anomalous data samples, including the step of determining the first quality measure in the fit and generating a set of first quality measures for the set of anomalous data samples. The quality scale is determined based on a set of first quality scales.

Example 9 discloses a method for diagnosing refrigeration performance anomalies according to Example 9, wherein the first quality scale comprises the square of the multiple correlation coefficient of the first correlation model for the anomalous data sample, the overall quality scale. Includes a confidence interval for the determined set of first quality measures.

In Example 10, a method for diagnosing an abnormality in refrigerator performance according to any one of Examples 3 to 9 is disclosed, and the above-mentioned step of evaluating each of a plurality of candidate window sizes is performed for a plurality of candidate window sizes. Further includes the step of determining one of the plurality of candidate window sizes as the optimum window size based on each of the overall quality measures determined in the above.

In Example 11, a method for diagnosing a freezer performance abnormality according to any one of Examples 1 to 10 is disclosed, and a root cause analysis diagnosis for the abnormality is made based on the first correlation model and the optimum window size. Further includes a step of determining whether or not the root cause determined from the above is a synthetic operation parameter synthesized from the second plurality of operation parameters.

In Example 12, the method for diagnosing a refrigerating machine performance abnormality according to Example 11 is disclosed, a step of determining that the root cause is a synthetic operation parameter, a performance parameter associated with the synthetic operation parameter, and a second plurality. A step of setting a second correlation model with the operating parameters of the second and a step of determining a second optimal window size for performing a second root cause analysis diagnosis for anomalies based on the second correlation model. And in the step of performing a second RCA diagnosis based on a second correlation model and a second optimal window size for the anomaly to determine the second root cause of the anomaly identified in the refrigerating machine performance parameters. Therefore, a second root cause analysis diagnosis is performed on a second dataset that includes performance parameters collected over the period and measurement data for a second plurality of operating parameters, including anomaly measurement data. Including steps.

In Example 13, the method for diagnosing refrigerator performance anomalies according to Example 11 or 12 is disclosed, performance parameters are selected from the group consisting of chilled water supply, coefficient of performance, and chilled water supply temperature, and operating parameters are chilled water. Selected from the group consisting of return temperature, cold water flow rate, cooling water supply temperature, cooling water flow rate, steam flow rate, steam pressure, cooling water return temperature, power consumption, and ambient air temperature, the synthetic operating parameters are cold water return quality, cooling. It is selected from the group consisting of water supply quality and steam supply quality.

In Example 14, a system for diagnosing refrigerator performance anomalies is disclosed, wherein the system comprises a memory and at least one processor communicatively coupled to the memory, and is any one of Examples 1-13. It is configured to carry out the method for diagnosing the refrigerator performance abnormality by.

In Example 15, a computer program product embodied in the form of one or more non-temporary computer-readable storage media is disclosed for the refrigerator performance abnormality diagnosis by any one of Examples 1 to 13. Includes instructions that can be executed by at least one processor to implement the method.

In Example 16, a method for performance diagnosis based on a two-layer RCA model is disclosed, which includes a method for two-layer RCA modeling, a method for calculating the optimum window size, and an upper layer of the two-layer RCA model based on synthetic constraints. It is associated with the steps of forming layers, calculating the optimal window size using correlation coefficients and confidence intervals, performing the RCA diagnosis using the optimal window size, and the constraints determined as bottlenecks. Includes steps to build a lower layer of the two-layer RCA model based on the raw sensor.

In Example 17, the method for performance diagnosis according to Example 16 is disclosed, and the method is applied to the freezer diagnosis.

Example 18 discloses a method for performance diagnostics according to Example 16 or 17, via a user interface, where the steps of forming the upper layer provide a selection of refrigerator performance, raw sensor constraints, and synthetic constraints. Further includes the steps of forming an RCA model.

In Example 19, the method for performance diagnosis according to Example 16 or 17 is disclosed, and the step of calculating the optimum window size is such that each set is generated for each one of a plurality of candidate window sizes. , The step of constructing multiple sets of anomalous data samples, the step of forming a correlation model (RCA model) for the anomalous data samples and calculating the correlation coefficient for the correlation model, and the candidate window with weak correlation with refrigerating machine performance. It further includes a step of rejecting the size and sample, a step of calculating the confidence interval for R ² of the anomalous sample, and a step of determining the optimum window size based on the confidence interval of R ² .

In Example 20, the method for performance diagnosis according to Example 16 or 17 is disclosed, and the step of performing the RCA diagnosis further includes a step of calling the RCA program to perform the diagnosis using the optimum window size.

In Example 21, the method for performance diagnosis according to Example 16 or 17 is disclosed, and the step of constructing the lower layer is a step of constructing the lower layer RCA model using the refrigerator performance and the raw sensor constraint. The raw sensor constraint is included in the synthetic constraint determined as the bottleneck, further including steps.

In Example 22, the method for performance diagnosis according to Example 19 is disclosed, and the step of constructing a plurality of sets of anomalous data samples is based on a sliding window over an anomalous period, and each set of anomalies with respect to the corresponding candidate window size. Includes steps to build a data sample.

In Example 23, the method for performance diagnosis according to Example 19 is disclosed, and the step of forming the correlation model forms a correlation model for each anomalous sample through multiple linear correlation regression, and calculates the correlation coefficient ^R2 . Includes more steps.

In Example 24, the method for performance diagnosis according to Example 19 is disclosed, and the step of rejecting the window size further includes the step of rejecting the window size for constructing a sample having a small average value of R ² .

In Example 25, the method for performance diagnosis according to Example 19 is disclosed, and the step of calculating the confidence interval calculates the 95% confidence interval for ^R2 of the abnormal sample according to the window size, the abnormal size, and the like. Further includes steps to do.

In Example 26, the method for performance diagnosis according to Example 19 is disclosed, and the step of determining the optimum window size further includes the step of determining the optimum window size based on the 95% confidence interval of R ² . The window size with the narrowest confidence interval is selected as optimal.

Example 27 discloses a computer program product implemented for the method according to Example 16 or 17, forming an upper layer of a two-layer RCA model through synthetic constraints and optimally using correlation coefficients and confidence intervals. At least one processor that calculates the window size, performs an RCA diagnosis with the optimal window size, and builds a sublayer of the two-layer RCA model based on the raw sensors associated with the constraints determined as bottlenecks. Includes instructions that can be executed by.

In Example 28, the computer program product according to Example 27 is disclosed, and the steps of forming the upper layer form an RCA model via a user interface that provides a choice of refrigerator performance, raw sensor constraints, and synthetic constraints. Further includes steps to do.

In Example 29, the computer program product according to Example 27 is disclosed, and the step of calculating the optimum window size is a plurality of sets of anomalies in which each set is generated for each one of the plurality of candidate window sizes. A step to build a data sample, a step to form a correlation model for anomalous data samples and calculate the correlation coefficient for the correlation model, and a step to reject candidate window sizes and samples that are weakly correlated with refrigerating machine performance. It further includes a step of calculating the confidence interval for R ² of the anomalous sample and a step of determining the optimum window size based on the confidence interval of R ² .

In Example 30, the computer program product according to Example 27 is disclosed, and the step of performing the RCA diagnosis further includes a step of calling the RCA program in order to perform the diagnosis using the optimum window size.

In Example 31, the computer program product according to Example 27 is disclosed, and the step of constructing the RCA model is a step of constructing a lower layer of the RCA model using the refrigerator performance and the raw sensor constraint, and the raw sensor constraint. Includes steps included in the synthetic constraint determined as the bottleneck.

In Example 32, the computer program product according to Example 29 is disclosed, and the step of constructing a plurality of sets of anomalous data samples is based on a sliding window over an anomalous period, with each set of anomalous data samples for the corresponding candidate window size. Includes steps to build.

In Example 33, the computer program product according to Example 29 is disclosed, and the step of forming a correlation model includes a step of forming a correlation model for each anomalous sample through multiple linear correlation regression and calculating a correlation coefficient ^R2 . ..

In Example 34, the computer program product according to Example 29 is disclosed, and the step of rejecting the window size further includes the step of rejecting the window size for constructing a sample having a small mean value of R ² .

In Example 35, the computer program product according to Example 29 is disclosed, and the step of calculating the confidence interval is a step of calculating a 95% confidence interval for ^R2 of the abnormal sample according to the window size, the abnormal size, and the like. Further included.

In Example 36, the computer program product according to Example 29 is disclosed and the determination step further comprises determining the optimum window size based on the 95% confidence interval of ^R2 . The window size with the narrowest confidence interval is selected as optimal.

In Example 37, a computer program product embodied in the form of one or more non-temporary computer readable storage media is disclosed and the window size calculated and used for RCA diagnosis (eg, to an operator device). Includes instructions that can be executed by at least one processor to indicate (via the GUI displayed).
In Example 38, a computer program product for a dual-layer RCA model, embodied in the form of one or more non-temporary computer-readable storage media, is disclosed and both determined for the superior and inferior models. Includes instructions that can be executed by at least one processor to indicate the window size (eg, via a GUI displayed on the operator device).

Therefore, embodiments of the present invention may be applied to the abnormality diagnosis of various refrigerators in a cooling plant, including a steam absorption chiller, a centrifugal chiller, a screw chiller, and the like. For example, a chiller operator (eg, a maintenance manager) is based on various embodiments of the invention to reduce maintenance time and costs by discovering the root cause of anomalies more quickly and precisely. You may use the system.

Various embodiments of the present invention may also be applied to the operation optimization of the refrigerator. For example, the plant operator may apply various embodiments of the invention to identify bottlenecks in the performance of the refrigerator and optimize behavioral decisions to resolve the bottlenecks. In this regard, by improving the performance of the refrigerator, the energy consumption in the operation of the refrigerator is efficiently reduced.

For example, various embodiments of the present invention have been applied to the diagnosis of the entire cooling plant using a plurality of combined cooling devices such as a refrigerator, a cooling tower, a thermal energy storage (TES), and a heat exchanger (HEX). Also, in larger cogeneration plants, various embodiments of the invention may extend the diagnosis from cooling equipment to power generation equipment such as steam turbine generators, gas turbine generators and the like.

Although embodiments of the present invention have been specifically illustrated and described with reference to specific embodiments, the embodiments and details without departing from the spirit and scope of the invention as defined by the appended claims. It should be understood by those skilled in the art that various changes may be made. Thus, the scope of the invention is set forth by the appended claims, and thus all modifications within the equivalent meaning and scope of the claims shall be embraced.

Claims (15)

Steps to set up a first correlation model between the performance parameters of the refrigerator and the first plurality of operating parameters,
With respect to the anomalies identified by the performance parameters, the step of determining the optimum window size for performing the root cause analysis diagnosis based on the first correlation model, and
In order to determine the root cause of the abnormality identified by the performance parameter of the refrigerator, it is a step of carrying out the root cause analysis diagnosis based on the first correlation model and the optimum window size for the abnormality. , The root cause analysis diagnosis is in a dataset containing measurement data for the performance parameters and the first plurality of operating parameters collected over a period of time, including anomaly measurement data having an anomaly period associated with the anomaly. Computer mounting methods for refrigerating performance anomaly diagnosis, including, and steps performed against. The steps above to determine the optimal window size are:
Steps to get multiple candidate window sizes and
With respect to the first correlation model, in order to determine the optimum candidate window size for performing the root cause analysis diagnosis based on the first correlation model for the abnormality from the plurality of candidate window sizes. The method of claim 1, comprising the step of evaluating each of the plurality of candidate window sizes. The step of evaluating each of the plurality of candidate window sizes is for each candidate window size.
A step of generating a set of anomalous data samples from the dataset, each of which is generated based on the candidate window size and includes at least anomalous data points belonging to the anomaly measurement data.
A step of performing a multiple linear regression analysis on the set of anomalous data samples based on the first correlation model,
The method of claim 2, comprising the step of determining an overall quality measure in the fit of the first correlation model to the set of generated anomalous data samples. The method of claim 3, wherein the set of anomalous data samples is generated based on a sliding window technique performed over the anomalous period. At each subsequent time interval, the sliding window technique continues from the step of generating a first anomaly data sample containing the start data points of the anomaly measurement data to the final anomaly data sample containing the final data points of the anomaly measurement data. The method of claim 4, comprising the step of generating the anomalous data sample that follows to generate the set of the anomalous data samples. The first anomalous data sample has an anomalous sample period ending at the start data point of the anomaly measurement data, and the final anomalous data sample has an anomalous sample period beginning at the final data point of the anomaly measurement data. , The method according to claim 5. The method according to any one of claims 4 to 6, wherein each anomalous data sample in the set of anomalous data samples has an anomalous sample period length determined based on the candidate window size. The step of performing a multiple linear regression analysis is
For each anomalous data sample in the set of anomalous data samples, a step of performing a multiple linear regression analysis on the anomalous data sample based on the first correlation model, and
For each anomalous data sample in the anomalous data sample set, a first quality measure in the fit of the first correlation model to the anomalous data sample is determined, and a set of first anomalous data samples to the anomalous data sample set. Including the step of generating a quality scale of 1 and
13. Method. The first quality scale comprises the square of the multiple correlation coefficient of the first correlation model for the anomalous data sample, and the overall quality scale provides confidence intervals for the determined set of first quality scales. The method of claim 8, including. The step of evaluating each of the plurality of candidate window sizes sets one of the plurality of candidate window sizes to the optimum window size based on each of the overall quality measures determined for the plurality of candidate window sizes. The method according to any one of claims 3 to 9, further comprising the step of determining as. Whether or not the root cause determined from the root cause analysis diagnosis regarding the abnormality based on the first correlation model and the optimum window size is a synthetic operation parameter synthesized from a second plurality of operation parameters. The method according to any one of claims 1 to 10, further comprising a step of determining. The step of determining that the root cause is a synthetic operation parameter,
A step of setting a second correlation model between the performance parameter associated with the synthetic operation parameter and the second plurality of operation parameters.
A step of determining a second optimal window size for performing a second root cause analysis diagnosis for the anomaly based on the second correlation model.
The second RCA diagnosis based on the second correlation model and the second optimal window size for the anomaly to determine the second root cause of the anomaly identified in the performance parameter of the refrigerator. The second root cause analysis diagnosis includes measurement data relating to the performance parameters and the second plurality of operating parameters collected over the period, including the anomaly measurement data. 11. The method of claim 11, further comprising the steps performed on the dataset of. The performance parameters are selected from the group consisting of cold water supply, performance coefficient, and cold water supply temperature.
The operating parameters are selected from the group consisting of chilled water return temperature, chilled water flow rate, cooling water supply temperature, cooling water flow rate, steam flow rate, vapor pressure, cooling water return temperature, power consumption, and ambient air temperature.
The method of claim 11 or 12, wherein the synthetic operating parameter is selected from the group consisting of cold water return quality, cooling water supply quality, and steam supply quality. With memory
Includes at least one processor communicably linked to the memory and configured to perform the method for diagnosing a refrigerator performance abnormality according to any one of claims 1-13. A system for diagnosing abnormal refrigerator performance. One or more non-temporary computer readable including instructions executable by at least one processor to perform the method for diagnosing refrigerating performance anomalies according to any one of claims 1-13. A computer program product that is embodied in the form of a storage medium.

Images