No full-text available
To read the full-text of this research,
you can request a copy directly from the authors.
Model-based fault detection and diagnosis for centrifugal chillers
Abstract
Faulty operations of Heating, Ventilation and Air Conditioning (HVAC) chiller systems can lead to discomfort for the users, energy wastage, system unreliability and shorter equipment life. Faults need to be early diagnosed to prevent further deterioration of the system behaviour and energy losses. In this paper a model-based approach is used in order to detect important chiller systems faults. First, a linear dynamic black-box model is identified for each of the relevant characteristic features of the system during the normal functioning of the chiller. Then, an on-line correlogram method verifies the whiteness property of the residuals in order to distinguish anomalies from normal operations. A decision table, that matches the influence of anomalies with the characteristic features, allows to identify chiller faults. The proposed fault detection and diagnosis approach is assessed by using real chiller data provided by the ASHRAE research project RP-1043.
... Tran et al. [70] proposed a least-squares SVR method, a reformulation of the SVR [71], combined with differential evolution and EWMA to improve the performance of the EWMA-SVR method by Zhao et al. [68,69]. Other reference modelling methods have been proposed, such as RBF and Kriging (KRG) models [72,73], autoregressive with exogenous inputs (ARX) and autoregressive moving average with exogenous inputs (ARMAX) [74], and deep learning techniques [75]. However, these methods offer minimal improvement in the detection accuracy of faults like ConFou and RefLea which showed low FD accuracies with the SVR-based methods. ...
... Reddy [63] Simple linear regression t-statistic Lab Tran et al. [72] MLR, KRG, RBF EWMA Lab Tran et al. [73] RBF EWMA Lab Tran et al. [70] SVR EWMA Lab Beghi et al. [74] ARX and ARMAX Residual whiteness test Lab Bonvini et al. [96] Gray-box dynamic modelling N/A Lab & Field Zhu et al. [75] Deep learning N/A Lab ...
... Their method was found to improve the FDD performance, especially for fluid faults such as RedEva, RedCon, ExcOil, RefOve, RefLea, and NonCons. Lastly, Beghi et al.[74] used the whiteness test threshold determination method which is a hypothesis test if the residuals from ARX and ARMAX model predictions have a zero mean and finite variance, with a given confidence level. ...
... If a fault is present, deviations between the reference model and the process are to be expected on a larger scale, allowing faults to be detected [102]. Several methods are known for deriving such a model, including state estimation, parameter estimation, parity space evaluation, or a combination of these [92]. ...
... A major challenge in the development of chiller CBM models, or FDD in particular, constitutes the high system complexity as well as the multitude of different system architectures and operation conditions, as this complicates the nature of faults [102]. In particular, chillers utilised for industrial purposes are often custom-built [26, p. 464], while they are developed for specific application requirements. ...
... An MLR based reference model was also introduced in [24] and compared with a simple linear regression and a decoupling based FDD approach, whereby the latter was found to yield the highest fault detection performance. Another approach is presented by Beghi el al. [102], where various models for statistical analysis of time series data (ARMAX, ARX, etc.) were empirically evaluated and applied for predicting the values of some preselected features. ...
Download:https://art.torvergata.it/handle/2108/323763
The automatic assessment of the degradation state of industrial refrigeration systems is
becoming increasingly important and constitutes a key-role within predictive maintenance
approaches. Lately, data-driven methods especially became the focus of research in this
respect. As they only rely on historical data in the development phase, they offer great
advantages in terms of flexibility and generalisability by circumventing the need for specific
domain knowledge. While most scientific contributions employ methods emerging from
the field of machine learning (ML), only very few consider their applicability amongst
different heterogeneous systems. In fact, the majority of existing contributions in this field
solely apply supervised ML models, which assume the availability of labelled fault data for
each system respectively. However, this places restrictions on the overall applicability, as
data labelling is mostly conducted by humans and therefore constitutes a non-negligible
cost and time factor. Moreover, such methods assume that all considered fault types
occurred in the past, a condition that may not always be guaranteed to be satisfied.
Therefore, this dissertation proposes a predictive maintenance model for industrial
refrigeration systems by especially addressing its transferability onto different but related heterogeneous systems. In particular, it aims at solving a sub-problem known as
condition-based maintenance (CBM) to automatically assess the system’s state of degradation. To this end, the model does not only estimate how far a possible malfunction
has progressed, but also determines the fault type being present. As will be described
in greater detail throughout this dissertation, the proposed model also utilises techniques
from the field of ML but rather bypasses the strict assumptions accompanying supervised
ML. Accordingly, it assumes the data of the target system to be primarily unlabelled
while a few labelled samples are expected to be retrievable from the fault-free operational
state, which can be obtained at low cost. Yet, to enable the model’s intended functionality, it additionally employs data from only one fully labelled source dataset and, thus,
allows the benefits of data-driven approaches towards predictive maintenance to be further
exploited.
After the introduction, the dissertation at hand introduces the related concepts as
well as the terms and definitions and delimits this work from other fields of research.
Furthermore, the scope of application is further introduced and the latest scientific work
is presented. This is then followed by the explanation of the open research gap, from which
the research questions are derived. The third chapter deals with the main principles of the
model, including the mathematical notations and the individual concepts. It furthermore
delivers an overview about the variety of problems arising in this context and presents the
associated solutions from a theoretical point of view. Subsequently, the data acquisition
phase is described, addressing both the data collection procedure and the outcome of the
test cases. In addition, the considered fault characteristics are presented and compared
with the ones obtained from the related publicly available dataset. In essence, both
datasets form the basis for the model validation, as discussed in the following chapter. This
chapter then further comprises the results obtained from the model, which are compared
with the ones retrieved from several baseline models derived from the literature. This
work then closes with a summary and the conclusions drawn from the model results.
Lastly, an outlook of the presented dissertation is provided.
... FDD methods can generally be classified into two categories: model-based and data-driven [9][10][11][12][13]. The major difference between both methods is that the former relies on a dynamical reference model of the observed system, whereas data-driven ones make use of historical data [1] for machinery health assessment. ...
... This still constitutes a major challenge for VCRS as modelling multi-phase flows inside the refrigerant cycle, especially for complex geometries, can be challenging. Blackbox models [10,[14][15][16][17] in turn, do not depend on the underlying physical structure of a system, but rather are based on learning the relationships between input and output variables of using statistical and machine learning methods. As shown in Figure 2, another widely applied approach is to apply data-driven models enabling learning of fault patterns on the basis of historical machine data. ...
... Their research particularly investigated the FDD strategies in the presence of multiple simultaneous faults. Another contribution to the topic is given by Beghi et al. [10], where different multiple-input-single-output models (ARMAX, ARX, etc.) were empirically identified for each selected CF respectively. ...
Vapor compression refrigeration systems are subject to performance degradation over time due to the presence of faults. However, latest work in the field of condition-based maintenance shows promising results in the automatic early detection of anomalous behaviour as well as in accurate machine diagnostics and can, therefore, increase the overall system reliability by simultaneously preventing machine downtimes. In this paper, the latest research works carried out within the last decade are reviewed and the approaches are classified regarding their working principles. Furthermore, the work at hand depicts the current research trend in this field and outlines current obstacles. 2 Introduction Vapor compression refrigeration systems (VCRS) are used in many industrial and commercial applications and are considered as large energy consumers. Particularly in the food industry, but also in other branches, such as the chemical or pharmaceutical industries, these systems are subject to high reliability standards and can account for 20%-40% of a facility's energy consumption [2] depending on the application scenario. Consequently, this sector offers particularly high potential for optimisation in terms of energy savings and reliability considerations. As the overall system performance is often decreased due to the presence of faults, energy efficient systems should be equipped with a condition monitoring system in order to enable the online system assessment of the respective appliance. Based on this measure, system faults may automatically be detected and diagnosed, thus, preventing energy waste and high maintenance costs [3]. Therefore, Condition-Based Maintenance (CBM) techniques have attracted attention for decades, and many approaches are well described across the literature. Due to CBM, maintenance actions can be scheduled based on the actual system condition [4] rather than on predetermined time intervals leading to lower operating costs [5]. Moreover, as VCRS can lose about 30% efficiency due to degradation processes by simultaneously seeming fully functional [6], CBM enables to detect and diagnose faults at an early stage of development and can, therefore , avoid energy waste. The main objective of a CBM systems is the automated detection and classification of faults with the lowest possible degree of severity, so that maintenance measures can be initiated in time. Besides the many subsystems of a CBM model, such as data-filtering or feature extraction, most researchers focussed on the development of automatic VCRS degradation assessment strategies. Especially fault detection and diagnosis (FDD) has been treated recently, which is indeed the most critical part in the CBM model development. Therefore, the work at hand reviews research contributions to the field with a special view to fault detection and
... In formula [5], is the heat exchange heat (w), is the heat exchange coefficient (w / m2 / k), is the engine and coolant heat exchange area (mm2), is the engine block temperature (k), is the engine cooling liquid inlet temperature (k), is the engine coolant outlet temperature (k). ...
... The validation of the thermal management simulation platform was done in two stages: a comparison during steady-state conditions and a comparison of operation using data from an existing 4 strokes engine with 4 cylinders. The average part simulation error is below 5%. Figure 4 Model of the engine cooling system Because ambient temperature is one of the main influence parameters [5], the simulation is running under NEDC working condition, at different ambient temperatures from -7°C, 20°C to 40°C. In Figure 4, heat losses dynamic response to different conditions. ...
New challenges are arising in vehicle thermal managing systems as air-pollution and fuel consumption rules become more rigid. Therefore, automobile manufactures have to face more competitive pressure. The use of software and hardware-in-loop (HIL) simulation is essential to develop and validate new real-time applications of engine, air conditioner and battery thermal management systems. This paper presents a simulation platform developed to work in a joint-simulation approach of thermal system and control system. The joint-simulation platform functions in a closed loop integrating KULI and Matlab/Simulink, which make it possible to validate both the capacity of thermal system and new control policies without a real system controller at the same time. The steady-state performance of the simulator correlates well with the provided manufacturer’s curve. The simulation platform could provide substantial improvements to the thermal management system performance.
... Kết quả là, không dễ phân loại ba loại lỗi này vì triệu chứng là tương tự nhau. Có rất nhiều tham số đặc tính từng được chọn nhằm phát triển phương pháp chẩn đoán sự số hệ thống chiller [6] [12]. Dựa theo công bố của Comstock và Braun [13] và các nghiên cứu [14][15], trong nghiên cứu này đề xuất các tham số đặc tính gồm TE (chênh lệch nhiệt độ nước và ra tại thiết bị bay hơi), TC (chênh lệch nhiệt độ nước và ra tại thiết bị ngưng tụ), LMTDcd là chênh lệch nhiệt độ trung bình logarit tại thiết bị ngưng tụ (logarithm mean temperature difference of condenser), Toil (nhiệt độ dầu) và sc (hệ số hiệu suất trao đổi nhiệt của thiết bị quá lạnh) được xác định bởi biểu thức (1). ...
Một mô hình tham số đặc tính chính xác sẽ có vai trò then chốt trong việc nâng cao tỷ lệ chính xác của quá trình phát hiện và chẩn đoán lỗi trong hệ thống chiller. Vì vậy trong nghiên cứu này đã thực hiện so sánh đánh giá 3 phương pháp MLR, GRNN và RBFNN đóng vai trò là mô hình tham số để mô hình hóa các đặc tính hoạt động của chiller. Hai chỉ số thống kê là R2 và RMSE được sử dụng là tiêu chí đánh giá mô hình ở giai đoạn huấn luyện mô hình. Sau đó, kết hợp với phương pháp t-test cùng với quy luật chẩn đoán để nghiên cứu khảo sát và đánh giá khả năng phát hiện chẩn đoán của 3 mô hình. Bộ dữ liệu thực nghiệm thường được sử dụng hầu hết cho hướng nghiên cứu phát hiện chẩn đoán sự cố trong hệ thống chiller của ASHRAE RP-1043 đã được sử dụng trong nghiên cứu này. Nghiên cứu tiến hành khảo sát đánh giá 3 mô hình với 3 trường hợp tiêu biểu là “Chiller hoạt động bình thường” và 2 sự cố thường xuất hiện trong hệ thống chiller “Thiếu môi chất lạnh”, “Tắc thiết bị ngưng tụ”. Kết quả của nghiên cứu cho thấy rằng, RBFNN và GRNN là một chiến lược rất thiết thực và có độ chính xác cao.
... For example, expert knowledge, maintenance records, building information models (BIM), and real-time occupancy data. Expert knowledge has been integrated into data-driven FDD approaches to (1) detect outliers in the data preprocessing step [83], (2) develop, select, and interpret the characteristic features of faults [107,170,181,202], and (3) support the selection of layers, nodes, and parameters in the BN-based or treestructured FDD algorithms [108,146,158]. Maintenance records were utilized to label the ground truth of field measurements, i.e., whether the collected data contain faults or not [116,148]. ...
With the wide adoption of building automation system, and the advancement of data, sensing, and machine learning techniques, data-driven fault detection and diagnostics (FDD) for building heating, ventilation, and air conditioning systems has gained increasing attention. In this paper, data-driven FDD is defined as those that are built or trained from data via machine learning or multivariate statistical analysis methods. Following this definition, this paper reviews and summarizes the literature on data-driven FDD from three aspects: process, systems studied (including the systems being investigated, the faults being identified, and the associated data sources), and evaluation metrics. A data-driven FDD process is further divided into the following steps: data collection, data cleansing, data preprocessing, baseline establishment, fault detection, fault diagnostics, and potential fault prognostics. Literature reported data-driven methods used in each step of an FDD process are firstly discussed. Applications of data-driven FDD in various HVAC systems/components and commonly used data source for FDD development are reviewed secondly, followed by a summary of typical metrics for evaluating FDD methods. Finally, this literature review concludes that despite the promising performance reported in the literature, data-driven FDD methods still face many challenges, such as real-building deployment, performance evaluation and benchmarking, scalability and transferability, interpretability, cyber security and data privacy, user experience, etc. Addressing these challenges is critical for a broad real-building adoption of data-driven FDD.
... To increase chiller reliability and performance, fault-detection and diagnosis (FDD) systems are used [3]. Existing FDD techniques in chillers can be broadly classified as: i) model-based, ii) data-driven, and iii) hybrid [4]. Model-based FDD uses physical laws to capture the component dynamics followed by signal processing operations to determine the presence of faults [5]. ...
We investigate the role of explainable Artificial Intelligence (XAI) for building trust in data-driven fault detection and diagnosis (FDD). We examine use cases for XAI-FDD on a building in Singapore that has six chillers.
... For example, in [16] a simplified chiller white-box model has been developed for fault detection tasks by defining performance indicators based on thermodynamic dependencies. Other authors proposed black-box models as developed by Tran et al. [17][18][19], Y. Zhao et al. [20], 40 Beghi et al. [21] or X. Zhao et al. [22]. In model-based FDD, faults are commonly identified by use of decision tables that require appropriate expert knowledge. ...
Automatic fault diagnosis is becoming increasingly important for assessing a chiller’s degradation state and plays a key role in modern maintenance strategies. Data-driven approaches have already become well established for this purpose as they rely on historical data and are therefore more generally applicable compared to their model-based counterparts. Existing chiller fault diagnosis models, however, require labelled data from the target system, which are often not available. Therefore, in this paper, a data-driven fault diagnosis model is proposed that deploys domain adaptation techniques to enable the transfer of knowledge amongst heterogeneous chillers. In particular, the model utilizes transfer component analysis (TCA) and a support vector machine with adapting decision boundaries (SVM-AD) to diagnose faults by aggregating labelled source and unlabelled target domain data in the training phase. Furthermore, it is demonstrated how the model parameters can be tuned to ensure effective classification performance, which is then evaluated by use of fault data stemming from different chiller types. Experimental results show that with the proposed approach faults can be diagnosed with high accuracy for cases when labelled target domain data are not available.
Available at:
https://www.sciencedirect.com/science/article/pii/S0967066121000927
... Notwithstanding the effort required by the model building process, grey-box models maintain a certain appeal due the physical interpretability of the parameters, a feature that may prove worthy in certain industrial applications, e.g. fault detection [5]- [8]). ...
The optimal energy management of multiple chiller systems calls for the construction of mathematical models of chiller energy efficiency. The existing grey- or blackbox models include parameters that have to be estimated from experimental data. So far, the predictive capabilities of alternative models have been assessed and compared on data sets created by laboratory tests or provided by chiller manufacturers. In an Industry 4.0 context, the continuous monitoring and collection of field data discloses new opportunities but raises also robustness issues that are herein addressed. Herein, exploiting an extensive experimental dataset collected over a six-month period, four literature models and a new machine learning approach are compared. The second objective is assessing the robustness of the five models against covariate shifts, i.e. variations in the statistical distribution of the input variables that occur across different months. The grey-box Gordon-Ng model, though less accurate in nominal conditions than the Bi-quadratic and Multivariate polynomial models, proves however more robust against covariate shifts. The best performances, both in term of accuracy and robustness, are however provided by the two machine learning methods, with the Gaussian Process model performing better than the MLP artificial neural network.
... Examples of AHU-level faults from the literature are the return and supply air temperature sensor biases, fouling, leaking or stuck heating and cooling coils, stuck economizer dampers, efficiency degradations in fans, broken or fouling filters, inappropriate airflow and temperature setpoints, and coil valve control loop-tuning issues (Li & Wen, 2014;Trojanová, Vass, Macek, Rojiček, & Stluka, 2009;Wang, 2012;Yang, Cho, Tae, & Zaheeruddin, 2008). Common examples for plant-level faults are chiller condenser fouling, hot and chilled water circulating pumps' flow rate, chiller refrigerant over-or undercharge, leaking or stuck valves, and fouling in boiler tubes, and various sensor biases (Beghi, Cecchinato, Peterle, Rampazzo, & Simmini, 2016;Zhao, Wang, & Xiao, 2013a, 2013bZhao, Xiao, & Wang, 2013). As reported in two recent literature surveys (Kim & Katipamula, 2017;Yang et al., 2017), about 90% of the recent AFDD literature has been focused on these HVAC systems and components. ...
Operators’ work order descriptions in computerized maintenance management systems (CMMS) represent an untapped opportunity to benchmark a facility’s maintenance and operation performance. However, it is challenging to carry out analytics on these large and amorphous databases. This paper puts forward a text-mining method to extract information about failure patterns in building systems and components from CMMS databases. The method is executed in three steps. Step 1 is pre-processing to convert work order descriptions into a mathematical form that lends itself to a quantitative lexical analysis. Step 2 is clustering to focus on interesting sections of a CMMS database that contain work orders about failures in building systems and components – rather than less interesting routine maintenance and inspection activities. Step 3 is association rule-mining to identify the coexistence tendencies among the terms of cluster of interest (e.g. coexistence of the terms ‘radiator’ and ‘leak’). This text-mining method is demonstrated by using two data sets. One data set was from a central heating and cooling plant with four boilers and five chillers; the other data set was from a cluster of 44 buildings. The results provide insights into per equipment breakdown of failure events, top system and component-level failure modes, and their occurrence frequencies.
... Something else of particular relevance to the advancement of humanoid capital and its explanation is current research in the area of predictive maintenance (PdM). Predictive maintenance algorithms are being developed for a wide range of disparate technologies: medical equipment [35]; cargo trains [36]; packing and sealing technologies used in the food industries [37]; aeroengine components [38]; HVAC (heating, ventilation and air conditioning) systems [39] [40]; aircraft equipment [41]; mobile telecommunications networks [42]; power plant machinery [43]; and more. XAI will potentially enable AI technologies of all kinds to better express their own repair and maintenance needs, with PdM and XAI almost guaranteed to merge in some fashion or another as XAI continues to mature. ...
Inside AI research and engineering communities, explainable artificial intelligence (XAI) is one of the most provocative and promising lines of AI research and development today. XAI has the potential to make expressible the context and domain-specific benefits of particular AI applications to a diverse and inclusive array of stakeholders and audiences. In addition, XAI has the potential to make AI benefit claims more deeply evidenced. Outside AI research and engineering communities, one of the most provocative and promising lines of research happening today is the work on "humanoid capital" at the edges of the social, behavioral, and economic sciences. Humanoid capital theorists renovate older discussions of "human capital" as part of trying to make calculable and provable the domain-specific capital value, value-adding potential, or relative worth (i.e., advantages and benefits) of different humanoid models over time. Bringing these two exciting streams of research into direct conversation for the first time is the larger goal of this landmark paper. The primary research contribution of the paper is to detail some of the key requirements for making humanoid robots explainable in capital terms using XAI approaches. In this regard, the paper not only brings two streams of provocative research into much-needed conversation but also advances both streams.
In the real application of chiller fault diagnosis models, the various working conditions may cause the different data distribution between test data and training data, which may lead to the lower diagnosis accuracy of models. In this paper, a novel fault diagnosis method for chillers across working conditions based on the domain knowledge-assisted Deep Extreme Learning Machine (DELM) is proposed. Firstly, the actual normal and performance degradation operation data of the chiller is collected through the Internet of Things Intelligent Agent (IOTIA). The Random Forest (RF) is used to determine the importance of features selected by domain knowledge analysis, and different feature subsets are selected as the model input. Then, the model of the DELM is used for fault diagnosis of chillers across varying temperature and load rate working conditions. Finally, experiments show that the proposed model achieves outstanding effect under the designed seven experimental across working conditions, indicating that the model has good generalization ability and is suitable for fault diagnosis across working conditions of chillers.
The automated fault detection and diagnostics (AFDD) of heating, ventilation, and air conditioning (HVAC) using data mining and machine learning models have recently received substantial attention from researchers and practitioners. Various models have been developed over the years for AFDD of complete HVAC or its sub-systems. However, HVAC complexities, which partly have roots in its close coupling nature and interrelated dependencies, mean that understanding the relationship between faults and the suitability of the techniques remains an unanswered question. The literature analysis and interactive visualization of the data collected from the past implementation of AFDD models can provide useful insight to further explore this question by applying artificial intelligence (AI). Association rule mining (ARM) is deployed by this paper, using the frequent pattern (FP) growth algorithm to generate frequent fault sets for most common HVAC faults from the body of AFDD models developed in the literature to represent the status quo. A new model is developed for common HVAC faults and the techniques most frequently used to detect and diagnose them. A recommender system is developed using the ARM model to extract knowledge from the body of knowledge of HVAC data-driven AFDD in the form of rule-sets that reflect the associations. Findings of this review paper can significantly help civil and building engineers, as well as facility managers, in better management of building HVAC systems.
Continuous Glucose Monitoring (CGM) sensors are key components in an artificial pancreas, an emerging tool for type 1 diabetes treatment. Malfunctioning of this component might reduce the efficacy of glucose control achieved by the system and even pose the safety of the patient at risk. Therefore, accurate and prompt detection of these anomalies is an important problem. This paper investigates a model-based method to detect CGM failures. Based on an individualized linear model of the subject, identified on hystorical data, the method predicts future glucose concentration through a one-step ahead Kalman predictor. The correct functioning of the system is then monitored using two different criteria: the first checks the magnitude of prediction residuals. The second checks the whiteness of the residuals through a correlogram test. The effectiveness of the two criteria is investigated and compared by performing tests on an in-silico dataset obtained by means of UVA/Padova Type 1 Diabetes simulator, accepted by the US Food and Drug Administration as a substitute of animal testing prior to artificial pancreas clinical trials on humans.
Operational risk evaluation for chillers is beneficial for reducing building energy wastage. An accurate reference model can significantly improve the evaluation performance. This paper presents a novel framework for developing and applying the reference model by integrating density clustering and deep learning. Unsupervised density-based spatial clustering of applications with noise (DBSCAN) is adopted to construct the library of operating conditions and recognize operating pattern of chillers. Deep learning approach, deep belief network (DBN), is presented to learn all process data in each operating pattern. Subsequently, multiple DBN models are developed for matching various operating patterns in the condition library. A simple strategy for tuning the hyperparameters of DBN is further presented to obtain a better performance. The prediction and generalization abilities of the proposed approach are validated and compared with multivariate linear regression (MLR), support vector regression (SVR) and radial basis function (RBF) models based on the experimental data obtained from a real screw chiller. Results reveal that the proposed method yields a significant performance advantage than MLR, SVR and RBF models, especially for the extended conditions in actual applications, the mean relative errors of MLR, SVR, RBF and the proposed method are 5.8%, 11.41%, 13.73%, and 2.11%, respectively.
In this paper, the calibration of the parameters of the Gordon-Ng Universal (GNU) chiller model is investigated. In its standard formulation, the GNU model is written as a linear-in-parameter model that can be calibrated by Ordinary Least Squares. It has been already observed elsewhere that, since the regressors are subject to measurement inaccuracies, the OLS approach is prone to yield biased estimates of the parameters. As a remedy, Andersen and Reddy proposed the adoption of an Errors in Variable (EIV) framework, showing that bias could be reduced or even eliminated by means of a corrected least squares algorithm. However, some questions remained open. Given that the EIV approach achieves bias reduction at the cost of increasing the variance, is it really preferable to OLS? If the final goal is not parameter estimation, but the prediction of the Coefficient of Performance (COP), how does OLS compare with EIV? And what is the most appropriate calibration method, under a statistical viewpoint? Finally, is the added complexity of a statistically rigorous approach employing Nonlinear Least Squares (NLS) really worth the potential improvements in COP prediction? In order to answer these questions, three estimation methods, OLS, EIV and NLS, are tested on two benchmarks: a public precise chiller performance dataset and an ASHRAE dataset. The results suggest, that OLS estimation, in spite of its suboptimality, may prove largely satisfactory both for parameter estimation and COP prediction, although it may be worth analyzing other more challenging COP prediction problem before the final word is said.
In this paper, an advanced evolving clustering strategy is used to design a fuzzy model-based sensor faults detection mechanism for a pilot parallel-type heat exchanger. The change in the process operating mode is detected by an incremental unsupervised clustering procedure based on participatory learning. Real experimental data is used to construct signals for fuzzy residual generation. The resulting residuals are then processed by the evolving classifier to supervise the heat exchanger operation. The obtained results clearly show the ability of the evolving fuzzy classifier to early detect the considered temperature sensor faults.
This paper presents an analysis conducted upon the sensor data gathered from the distribution control system of a central heating and cooling plant in Ottawa, Canada. After observing that the performance of four boilers and five chillers of this plant vary substantially in time under steady-state conditions, data-driven models were developed to explain this variability from the archived sensor data. By employing a forward stepwise regression and a repeated random sub-sampling cross-validation approach, two-layer feed-forward artificial neural network models with seven to fifteen hidden-nodes were selected for each boiler and chiller. The selected boiler models could explain 84 to 95% of the variability in a boiler's efficiency, and the selected chiller models could explain 65 to 94% of the variability in a chiller's coefficient of performance. Among studied nine variables, the most informative ones to predict a boiler's efficiency were identified as follows: flue gas O2 concentration, pressure, part-load ratio, forced draft fan state, and return water flow rate. Unlike boilers, all four studied variables were found useful in predicting a chiller's coefficient of performance. These four variables were the return water flow rate, part-load ratio, outdoor temperature, and return water temperature. A residual analysis was conducted to verify the appropriateness of the selected models to the datasets. In addition, potential use cases for the selected models were discussed with illustrative examples.
In this paper, a multiple classifier machine learning (ML) methodology for predictive maintenance (PdM) is presented. PdM is a prominent strategy for dealing with maintenance issues given the increasing need to minimize downtime and associated costs. One of the challenges with PdM is generating the so-called “health factors,” or quantitative indicators, of the status of a system associated with a given maintenance issue, and determining their relationship to operating costs and failure risk. The proposed PdM methodology allows dynamical decision rules to be adopted for maintenance management, and can be used with high-dimensional and censored data problems. This is achieved by training multiple classification modules with different prediction horizons to provide different performance tradeoffs in terms of frequency of unexpected breaks and unexploited lifetime, and then employing this information in an operating cost-based maintenance decision system to minimize expected costs. The effectiveness of the methodology is demonstrated using a simulated example and a benchmark semiconductor manufacturing maintenance problem.
This paper is the second of a two-part review of methods for automated fault detection and diagnostics (FDD) and prognostics whose intent is to increase awareness of the HVAC&R research and development community to the body of FDD and prognostics developments in other fields as well as advancements in the field of HVAC&R. The first part of the review focused on generic FDD and prognostics, provided a framework for categorizing methods, described them, and identified their primary strengths and weaknesses (Katipamula and Brambley 2005). In this paper we address research and applications specific to the fields of HVAC&R, provide a brief discussion on the current state of diagnostics in buildings, and discuss the future of automated diagnostics in buildings.
The term Fault Detection and Diagnosis (FDD) is a development of the term Fault Detection and Isolation (FDI). Generally speaking,
FDD goes slightly further than FDI by including the possibility of estimating the effect of the fault and/or diagnosing the
effect or severity of the fault. Hence, the term FDD also covers the capability of isolating or locating a fault. Both of
these topics have received considerable attention worldwide and have been theoretically and experimentally investigated with
different types of approaches, as can be seen from the general survey works [1, 2, 3, 4, 5, 6, 7].
A most critical and important issue surrounding the design of automatic control systems with the successively increasing complexity is guaranteeing a high system performance over a wide operating range and meeting the requirements on system reliability and dependability. As one of the key technologies for the problem solutions, advanced fault detection and identification (FDI) technology is receiving considerable attention.
The objective of this book is to introduce basic model-based FDI schemes, advanced analysis and design algorithms and the needed mathematical and control theory tools at a level for graduate students and researchers as well as for engineers.
Faulty operations of Heating, Ventilation and Air Conditioning (HVAC) chiller systems can lead to discomfort for the occupants, energy wastage, unreliability and shorter equipment life. Such faults need to be detected early to prevent further escalation and energy losses. Commonly, data regarding unforeseen phenomena and abnormalities are rare or are not available at the moment for HVAC installations: for this reason in this paper an unsupervised One-Class SVM classifier employed as a novelty detection system to identify unknown status and possible faults is presented. The approach, that exploits Principal Component Analysis to accent novelties w.r.t. normal operations variability, has been tested on a HVAC literature dataset.
In general there are two ways of arriving at models of physical processes:
Physical principles modelling. Physical knowledge of the process, in the form of first principles, is employed to arrive at a model that will generally consist of a multitude of differential / partial differential / algebraic relations between physical quantities. The construction of a model is based on presumed knowledge about the physics that governs the process. The first principles relations concern, e.g., the laws of conservation of energy and mass and Newton’s law of movement.
Experimental modelling, or system identification. Measurements of several variables of the process are taken and a model is constructed by identifying a model that matches the measured data as well as possible.
In this paper, an engine fault detection and classification technique using vibration data in the crank angle domain is presented. These data are used in conjunction with artificial neural networks (ANNs), which are applied to detect faults in a four-stroke gasoline engine built for experimentation. A comparative study is provided between the popular backpropagation (BP) method, the Levenberg-Marquardt (LM) method, the quasi-Newton (QN) method, the extended Kalman filter (EKF), and the smooth variable structure filter (SVSF). The SVSF is a relatively new estimation strategy, based on the sliding mode concept. It has been formulated to efficiently train ANNs and is consequently referred to as the SVSF-ANN. The accuracy of the proposed method is compared with the standard accuracy of the Kalman-based filters and the popular BP algorithms in an effort to validate the SVSF-ANN performance and application to engine fault detection and classification. The customizable fault diagnostic system is able to detect known engine faults with various degrees of severity, such as defective lash adjuster, piston chirp (PC), and chain tensioner (CT) problems. The technique can be used at any dealership or assembly plant to considerably reduce warranty costs for the company and manufacturer.
A most critical and important issue surrounding the design of automatic control systems with the successively increasing complexity is guaranteeing a high system performance over a wide operating range and meeting the requirements on system reliability and dependability. As one of the key technologies for the problem solutions, advanced fault detection and identification (FDI) technology is receiving considerable attention. The objective of this book is to introduce basic model-based FDI schemes, advanced analysis and design algorithms and the needed mathematical and control theory tools at a level for graduate students and researchers as well as for engineers. © 2008 Springer-Verlag Berlin Heidelberg. All rights are reserved.
Some classical schemes in algebraic system identification are first recalled and compared. It is shown that, in most cases, the solution is obtained thanks to additional assumptions which are not deducible from the available data. The identification problem for linear dynamic systems is then solved on the basis of the Frisch scheme, in order to obtain the whole set of models compatible with noisy input-output sequences. The main result here proposed concerns the unicity of the solution when the data are affected by additive white noise.
The appearance of this book is quite timely as it provides a much needed state-of-the-art exposition on fault detection and diagnosis, a topic of much interest to industrialists. The material included is well organized with logical and clearly identified parts; the list of references is quite comprehensive and will be of interest to readers who wish to explore a particular subject in depth.
The presentation of the subject material is clear and concise, and the contents are appropriate to postgraduate engineering students, researchers and industrialists alike. The end-of-chapter homework problems are a welcome feature as they provide opportunities for learners to reinforce what they learn by applying theory to problems, many of which are taken from realistic situations.
However, it is felt that the book would be more useful, especially to practitioners of fault detection and diagnosis, if a short chapter on background statistical techniques were provided.
Joe Au
This paper presents a method for automated detection and diagnosis of faults in vapor compression air conditioners that only requires temperature measurements, and one humidity measurement. The differences between measured thermodynamic states and predicted states obtained from models for normal performance (residuals) are used as performance indices for both fault detection and diagnosis. For fault detection, statistical properties of the residuals for current and normal operation are used to classify the current operation as faulty or normal. A diagnosis is performed by comparing the directional change of each residual with a generic set of rules unique to each fault. This diagnostic technique does not require equipment-specific learning, is capable of detecting about a 5% loss of refrigerant, and can distinguish between refrigerant leaks, condenser fouling, evaporator fouling, liquid line restrictions, and compressor valve leakage.
In this final part, we discuss fault diagnosis methods that are based on historic process knowledge. We also compare and evaluate the various methodologies reviewed in this series in terms of the set of desirable characteristics we proposed in Part I. This comparative study reveals the relative strengths and weaknesses of the different approaches. One realizes that no single method has all the desirable features one would like a diagnostic system to possess. It is our view that some of these methods can complement one another resulting in better diagnostic systems. Integrating these complementary features is one way to develop hybrid systems that could overcome the limitations of individual solution strategies. The important role of fault diagnosis in the broader context of process operations is also outlined. We also discuss the technical challenges in research and development that need to be addressed for the successful design and implementation of practical intelligent supervisory control systems for the process industries.
In this part of the paper, we review qualitative model representations and search strategies used in fault diagnostic systems. Qualitative models are usually developed based on some fundamental understanding of the physics and chemistry of the process. Various forms of qualitative models such as causal models and abstraction hierarchies are discussed. The relative advantages and disadvantages of these representations are highlighted. In terms of search strategies, we broadly classify them as topographic and symptomatic search techniques. Topographic searches perform malfunction analysis using a template of normal operation, whereas, symptomatic searches look for symptoms to direct the search to the fault location. Various forms of topographic and symptomatic search strategies are discussed.
2011 Buildings Energy Data Book
- J D Kelso
Identificazione dei Modelli e Sistemi Adattativi
- S Bittanti
Development of analysis tools for the evaluation of fault detection and diagnostics in chillers
- M C Comstock
- J E Braun
Fault detection and diagnosis in chillers-part 1: Model development and application
- I B D Mcintosh
- J W Mitchell
- W A Beckman