See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/311127950

Smart Fault Detection and Diagnosis (SFDD) for heat pump systems

Conference Paper · August 2015

CITATIONS READS

2 658

1 author:

Hatef Madani
KTH Royal Institute of Technology
45 PUBLICATIONS   947 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Smart Control strategies for heat pump systems View project

Photovoltaics for Swedish Prosumers View project

All content following this page was uploaded by Hatef Madani on 30 November 2016.

The user has requested enhancement of the downloaded file.

 SMART FAULT DETECTION AND DIAGNOSIS FOR HEAT PUMP
SYSTEMS

Hatef MADANI(*)
(*)
KTH Royal Institute of Technology, Brinellvägen 68, 100 44 Stockholm, Sweden
[email protected]

ABSTRACT

This paper introduces a holistic approach to develop a Smart Fault Detection and Diagnosis (SFDD) system
which can be integrated and widely used in both old and new heat pump systems worldwide. A SFDD acts as
the heart of smart heat pump system. It is able to minimize the installation and control errors, couple or
decouple the renewable energy sources, make load-shifting possible, detect the performance degradation
during operation, avoid unnecessary visual inspections and components replacement, and reduce the
maintenance cost and down-time of the system. In order to determine the most important faults in the heat
pump system, a comprehensive study is done on the faults reported to both OEMs and a main insurance
company in Sweden and the results are briefly presented. The paper also discusses how a knowledge-based
Smart Fault Detection and Diagnosis unit should be designed to support the heat pump systems at
commissioning, operation, and maintenance phases.

Keywords: Fault; smart; heat pump; fault detection; diagnosis

1. Introduction

Heat pump technology provides an efficient and sustainable solution for both heating and cooling of millions
of buildings around the world. Replacing the oil or gas boiler or pure electrical heater by the heat pump
system has saved a large amount of energy and cut the global CO2 emissions considerably. Furthermore, the
heat pumps play a key role in future energy scenarios, both at EU and global level. Bettgenhäuser et al.
(2013) predicts that the CO2-eq emissions will be reduced by 40% if 50% of all new buildings and 30% of
retrofitted buildings are equipped with heat pumps in 2030.

In spite of the huge contribution of heat pump systems in saving energy and cutting global emissions, there is
still a potential for improvement of heat pump efficiency and reliability. Madani and Roccatello (2014)
stated that some small modifications in heat pump system can reduce the number and cost of the faults in the
system considerably. Furthermore, they suggested a Smart Fault Detection and Diagnosis (SFDD) system for
the heat pumps to detect and diagnose the possible faults in installation and operation phases and to reduce
the maintenance cost and down time of the system.

Several methods to detect and diagnose the faults in heat pumps and air-conditioners are already discussed in
the scientific literatures. Payne et al. (2008; 2010) investigated fault detection in residential heat pumps and
air conditioners. Venkatasubramanian et al (2002) made three papers in series describing the methods for
fault detection and diagnosis, grouped into three main categories: quantitative model-based methods,
qualitative model-based methods, and process history based methods (Ibid). Likewise, in the work presented
by Katipamula and Bramble (2005), the authors addressed the FDD research and applications specific to the
fields of HVAC & R and provided a brief discussion on the current state of diagnostics in buildings,
discussing the future of automated diagnostics in buildings.

Rossi and Braun (1997) developed a method for fault detection and diagnosis for vapor compression air
conditioners. The differences between measured and predicted states obtained from models for normal
performance, called residuals, are used as performance indices for fault detection and diagnosis. A diagnosis
is performed by comparing the directional change of each residual with a generic set of rules unique to each
fault. Li and Braun (2006) proposed a new method for FDD of vapor compression air-conditioners dealing
with multiple simultaneous faults. The method is an improvement of the statistical rule-based technique
developed previously. If the current operation points are not inside the normal operation, it will be classified
as faulty. For the diagnosis, a mathematical methodology for decoupling is presented. Firstly air conditioner
most common faults and their causes were listed; then a decoupling feature was detected for each of them.
This feature is a parameter that can be either measured or derived from measurements.

Chen and Braun (2000) presented two easy to implement FDD method in rooftop air conditioner units:
Sensitivity Ratio Method (consisting of a ratio of residuals, one being sensitive to a fault and the other one
being insensitive) and the Simple Rule Base Method. The two methods were tested by laboratory experiment
for different fault types and fault levels and showed good performance and low computational effort.
Yoon et al. (2011) evaluated an air to air heat pump equipped with a thermostatic expansion valve (TXV)
tested in climate chambers during steady-state no-fault and imposed-fault operation. Different faults were
imposed and observations were made during heating and cooling mode. They identified sensitive and
insensitive features related to each fault under the heating or cooling mode. Furthermore, Kim et al. (2008)
presented a methodology for developing a steady-state detector for a vapor compression system based on a
moving window and using using the recursive version of the variance and standard deviation.

Moreover, Zogg et al. (2006) developed a tool for fault diagnosis particularly for heat pumps. A method
based on black box model and clustering technique was proposed. Clustering method is applied for fault
classification. Clusters representing all different faults are built. The properties of each one, such as centers
and standard deviations are obtained from parameters identified during the training phase. During the
operation, a membership grade of the parameters to the fault cluster is calculated, thus this fault occurring in
the system can be identified.

The current paper introduces a holistic approach to develop a Smart Fault Detection and Diagnosis (SFDD)
for heat pump systems. A SFDD system as the heart of a “smart heat pump system”, which will be defined in
the next section, is able to minimize the installation and control errors, decrease the performance degradation
during operation, avoid unnecessary visual inspections and components replacement, and reduce the
maintenance cost and down-time of the system.

2. Smart Fault Detection and Diagnosis as the heart of a smart heat pump system

Before going through Smart Fault Detection and Diagnosis (SFDD) as a vital part of smart heat pump
system, it is worth to describe a smart heat pump system.
As shown in Figure 1, the heat pump system can be viewed from four different boundary levels depending
on where the system boundary is drawn (Lundqvist 2010):
1. The “heat pump unit” level, mostly comprising of an evaporator, a condenser, a compressor, an
expansion valve and a working fluid,
2. The “heat pump system” level, including the heat source, liquid pumps or the auxiliary heater (if there
is any),
3. The “building system” level, including the building characteristics and inhabitants, building thermal
inertia, solar collector, etc.,
4. The “energy system” level, which not only includes the building system but also the primary energy
supplied to the system to provide electricity, electric power transmission and distribution losses, etc.

The first boundary level only covers the “heat pump unit” which is defined as the system facilitating the
thermodynamic cycle, i.e. the “heat pumping”, comprising of an evaporator, a condenser, a compressor, an
expansion valve and a working fluid. For example, a smart heat pump on “heat pump unit level” can use a
highly efficient compressor or it can detect and diagnose any fault on the unit level such as faulty compressor
or frozen evaporator.

The second boundary level which is called “Heat Pump system” level includes several sub-systems such as
the heat source (a liquid-filled borehole for example), the liquid pumps, the heat distribution system, the
auxiliary heater, or storage tank. For example, a smart heat pump on the “system” level can change the pump
or fan speed in the source or sink side to meet the heat demand; it can also use thermal storage mechanism to
shift the heat production from day to night, etc.

3 4

PV cell

2
Storage
DHW system Heat Distribution
Tank
System

Heat Pump Power plant

Unit

Heat Source

Figure 1. Smart heat pump system viewed from four different levels depending on where the system
boundary is drawn.

The third boundary level, the “building system” level, covers the building system including the building
inhabitants and building characteristics. In a smart heat pump on the building level, the control unit can be
trained to take the inhabitants’ behavior and the thermal inertia of the building into account and change the
control parameters continuously based on the dynamic behavior of the building and inhabitants. A smart heat
pump on the building level can predict the load profile based on the load history and weather forecast and
use this prediction to produce or store the required heat in an efficient way. It can also communicate with the
building inhabitants via smart phone or tablet apps in order to provide the comfort condition in an optimum
way.

The fourth boundary level, the “energy system” level, has a wider perspective on the heat pump and takes the
primary energy supplied to the system into account. The smart heat pump on the “energy system” level can
be regarded as a part of a smart grid. Some EU directives, such as the RES directives, promote increasing the
share of renewable energy sources in the electricity generation. This can lead to residual loads, caused by
large amounts of highly volatile renewable electricity generation such as wind turbines or solar PV cells.
Then, the smart heat pump can be utilized as a peak shaving mechanism or for storing the electric energy in
the form of heat.

Finally, A SFDD system should act as the heart of smart heat pump system. it should be able to minimize
the installation and control errors, couple or decouple the renewable energy sources, make load-shifting
possible, detect the performance degradation during operation, avoid unnecessary visual inspections and
components replacement, and reduce the maintenance cost and down-time of the system.
 3. Methodology

The present study suggests a holistic approach to develop a Smart Fault Detection and Diagnosis (SFDD)
mechanism as the heart of a smart heat pump system. To develop a SFDD mechanism, the first essential step
is the development of a comprehensive database of the most common faults during the installation and
operation processes of heat pumps. Heat pump manufacturers (OEMs) and Insurance companies are the best
sources to find out the most common and costliest faults already occurred in the heat pump systems. Under
warranty period, the customer service departments of OEMs receive a lot of feedback from the end-users in
the form of phone calls, emails, etc. when the customer has any complaint or any fault or failure occurs in the
system.

Approximately, 68000 of the fault reports, including any type of complaints, which were sent to several
OEMs during the warranty period are collected and processed. During the data processing, the unclear and
uncompleted fault reports are excluded. Obviously, it is very important to know the faults in the most recent
models of heat pumps. The old fault reports usually refer to the problems which are already known and
solved during the recent years; so the faults which were reported before 2010 are also excluded from this
study. Therefore, about 37000 faults reported to the OEMs, from the beginning of 2010 to the end of 2012
are analyzed and the results are presented in the next section.

When the warranty period is passed, the insurance company is the one who usually receives the fault reports
from the end-users. Therefore, in the present study, about 14000 faults reported to one of the largest
insurance companies in Sweden, called Folksam, are collected and processed. In order to make the study
more relevant, the faults from the old systems are omitted and only the faults from the heat pumps which
were sold after 2001 are considered (2001-2011). The file regarding the fault reports from 2007 was
corrupted and could not be included in the statistic. After omitting the uncompleted or vague reports at the
final stage of data processing, the number of fault reported to the insurance company is reduced to 8659.

After the data reduction, all the fault reports from OEMS and the insurance company are split into four
different categories: Air/Air, Air/Water, Brine/Water (mostly Ground Source Heat Pumps, GSHPs), and
Exhaust air heat pump systems. The faults occurred in these types of heat pump systems are scrutinized
separately. Furthermore, several interviews are made in order to interpret the faults reported to OEMs. The
final results of this comprehensive investigation are presented in the following section.

The next step is development of a Smart Fault Detection system at different phases: commissioning phase
and operation and maintenance phase. Smart Fault Detection system in the heat pumps should follow a rule
that I call as “TRRR rule”: Transferring the Right information to Right person(s) at Right time. The data
which is transferred to the receiver should be concise and informative. Sending too much numbers or graphs
from too many sensors in the system can only confuse the data receivers or reduce their sensitivities and
abilities to digest the data that they continuously receive. So the smart system uses the minimum number of
sensors and delivers as concise and informative data as possible. Furthermore, it is very critical to send the
data from the smart heat pump to the “Right person(s)”. For instance, reporting some of the faults to the end-
users who may know nothing about the heat pump can make them nervous. The end-users may also think
that the smart system which gives them the warning of performance degradation is worse than a normal heat
pump or a heat pump with poor performance which does not give any information. So the data should be
transferred to the right person(s) who can use the data efficiently. The time for transferring the information is
also very important. There is a risk when the system warns about a fault too early or too late. If the fault
detection system is too sensitive, some sudden changes in the controlled parameter, which can be for
example due to starting up the system, are treated as faulty conditions and early warning are sent; whereas,
the system could come back to a normal condition after a while and there was no need for that early warning.
More importantly, too late response from the fault detection system can lead to a serious damage of the
components or the whole system.

When the performance degradation or fault is detected, the next step is to help the servicemen or operator to
diagnose the fault. The SFDD system should be able to at least give some hints to the servicemen where the
faults come from and how to proceed in order to fix the problem, similar to the techniques used in the car
industry.
 4. The most important faults in heat pump systems

4.1. Most common and most expensive faults during warranty period
Table 1 presents a summary of the most common faults and the faults with the highest total costs which are
reported to OEMs from 2010 to the end of 2012 (three years). The results presented in this study show that
faults in Control and Electronics are among the most common and costliest faults in all types of heat pumps.
Faults in Control and Electronics include any fault related to control unit, electrical faults (such as short
circuit, etc.), PCB, display, soft starter, overcurrent and motor protection relay, etc. To conclude, in order to
reduce the number and cost of the faults in the heat pump systems, it is essential to pay a special attention to
the control and electronics in heat pump system. For further information regarding the important faults in the
heat pump systems during warranty period, refer to Madani and Roccatello (2014).

Table 1. A summary of the most common and costliest faults in different types of heat pump system-
according to the reports to HP Manufacturers during 2010-2012
Type of Heat Pump Air/Air HP Air/Water HP Brine/Water HP Exhaust air HP

Fan (26%) Pressure switch Control and Control and

(44%) Electronics (31%) Electronics (32%)
The Most common
faults Control and Control and Shuttle valve (19%) Shunt valve/motor
Electronics (25%) Electronics (25%) (19%)
Temperature Temperature Liquid pumps Temperature
sensors (16%) sensors (10%) (17%) sensors (11%)
Control and Pressure switch Control and Control and
Electronics (23%) (25%) Electronics (28%) Electronics (24%)
The costliest faults Refrigerant leakage Control and Liquid pumps Refrigerant leakage
(17%) Electronics (21%) (18%) (17%)
Fan (15%) Compressor (19%) Shuttle valve (12%) Domestic Hot
Water tank (13%)

4.2. Most common and most expensive faults after the warranty period
When the warranty period is passed, the insurance company is the one who usually receives the fault reports
from the end-users. Table 2 presents a summary of the most common faults and the faults with the highest
total costs which are reported to one of the largest insurance companies in Sweden, Folksam, from 2001 to
2011, excluding 2007. As presented in Table 2, the compressor is the costliest fault reported for all types of
heat pump systems. Furthermore, compressor is the most common fault reported to the insurance company
for all types of heat pumps except for Ground Source Heat Pumps (GSHPs). For GSHPs, the shuttle valve is
the most common fault reported.

Table 2. A summary of the most common and costliest faults in different types of heat pump system-
according to the reports to one of the main Swedish insurance companies during 2001-2011
Type of Heat Air/Air HP Air/Water HP Brine/Water HP Exhaust air HP
Pump
Compressor (30%) Compressor (24%) Shuttle valve Compressor (40%)
(22%)
The Most common Fan (19%) Control and Compressor (19%) Control and
faults Electronics (8%) Electronics (15%)
Control and Fan (8%) Control and Evaporator (11%)
Electronics (13%) Electronics (14%)
Compressor (46%) Compressor (52%) Compressor (49%) Compressor (56%)
Control and Evaporator (6%) Shuttle valve (9%) Evaporator (15%)
The costliest faults Electronics (12%)
Fan (8%) Refrigerant Control and Control and
leakage (5%) Electronics (9%) Electronics (9%)
It is worth to mention that the quality and clarity of the reports from the servicemen after the warranty period
is not satisfactory enough to draw a solid conclusion regarding the most common faults in heat pump
systems. The process of handling a fault occurred in heat pump system is not structured and leaves a lot of
space for the servicemen to manoeuvre. Furthermore, a considerable number of faults reported by
servicemen look incomplete and non-informative; so it is suggested that the whole process of fault reporting
should be restructured to improve the quality and clarity of the faults.

5. Designing the Smart Fault Detection and Diagnosis system

Smart Fault Detection and Diagnosis unit should support the heat pump systems at different phases:
commissioning, operation, and maintenance phases. These phases are treated as following:

5.1. SFDD at Commissioning phase

During the commissioning phase, the Smart Fault Detection and Diagnosis (SFDD) system does the first
system check a short time right after installation to ensure the manufacturer that the system is installed
correctly and operates properly. At this phase, the SFDD system finds the common installation errors,
incorrectly sized equipment, control errors, undercharge and overcharge conditions, and faulty components
at both heat pump unit and heat pump system levels.

In order to detect the faults at this phase, the Key Performance Indicators (KPIs) for commissioning phase
must be defined and collected in a commissioning management database of the SFDD toolbox. The installer
should go through the commissioning management box and check that all the KPIs are within a reasonable
range. The SFDD system for a specific heat pump unit is already trained in the laboratory and it knows what
the reasonable range is for each KPI at each specific operating condition.

Proper dimensioning of the heat pump unit and also system components can be checked by SFDD. The
amount of refrigerant charge is an important parameter which should be traced at the initial phases in order
to detect under- or over-charged systems. Sub-cooling temperature is a well-known indicator of the
refrigerant charge in the system which can be used as a KPI at the commissioning phase. The faults related to
Control and Electronics are one of the most common and costliest faults occurred in Swedish heat pump
systems (see section 4); therefore special consideration should be granted to control and electronics when
SFDD is designed. In the traditional way to control the on/off controlled heat pumps, installers set the control
parameters based on what was mentioned before in the control unit of the heat pump and then leave the
system until a serious fault occurs. However, SFDD at the commissioning phase should learn from the
heating system and the building what the most appropriate control parameters are and what the best heating
curve is to be set in the control unit. In later stages, the control parameters should change dynamically by
SFDD based on different inputs such as time, building system characteristics and ambient temperature.

Finally, there is a significant challenge to be faced when SFDD at commissioning phase is designed: how
long the commissioning phase should take? For how long the inspection should go on to check if a proper
installation was made and the system operates at an optimum condition? Of course, the answer depends on
the type, size and conditions of heat pump, installation, etc. However, it would be appropriate to choose a
smart commissioning strategy: SFDD in general records the data continuously and keeps the data for a
certain time, for example the last 24 hours. The installers should be assigned to do at least two obligatory
inspections via SFDD, one right after the installation and one within 24 hours after the installation; it is
recommended to have minimum 12 hours between the first and second inspection; then based on the
outcomes from these two inspections via SFDD, it can be decided if there is any need to make corrections or
modification in the system and if the SFDD should record the data for a longer time for further investigation
or not. In the case of need for further investigation, the installer should log into the commissioning
management database in the SFDD toolbox within 24 hours after the last inspection to make further
investigation or take the essential steps to fix the possible problem. The process should go on until the SFDD
shows a green light as an indication of a healthy condition.
5.2. SFDD at Operation and maintenance phase
After reviewing the different methods for fault detection and diagnosis, the authors concluded that a
knowledge-based method is one of the best methods to detect and diagnose the faults in heat pumps
particularly when several models had been already developed. The SFDD development process is as
following:
1. The models of heat pump unit components such as compressor, evaporator, condenser and expansion
valve are developed and validated against the experimental results.
2. The Key Performance Indicators (KPIs) at both unit and system levels are determined. The models
developed at the first stages are used to calculate some of the KPIs based on the online
measurements in heat pumps.
3. The models developed at the previous stages are used to emulate the possible faults and analyze the
effect of different faults on the measured parameters and consequently on KPIs and: how the
different KPIs are affected by the fault (the trend), how intense is the effect, etc.
4. The parametric studies in the previous stage lead to categorization of different faults; there are some
common symptoms between different faults but also some distinguished symptoms for every fault
which can be used as the key input in order to detect and diagnose the fault.
5. All the findings so far are used in order to develop a computer program which is able to detect and
diagnose the faults based on the trends and behavior of different KPIs and their categories. The
program is used in SFDD operation and maintenance management system which helps the
servicemen to detect the real cause of the fault and avoid unnecessary component replacement.

The Key Performance Indicators (KPIs) are the key parameters which are calculated based on the
measurements continuously being done on the heat pump systems. The heat pumps built nowadays are
already equipped with sensors which measure several temperatures and pressures at different points in heat
pumps. Therefore, there are a lot of measured data already available for many HP installations. However,
there is a little understanding about the meaning of these measured data and the massive measurement points
make it very hard to evaluate what is really happening in the system. KPIs in SFDD system act as the
interpreter of the measured data. SFDD receives all the measured temperatures, pressures, electrical powers
etc. and converts these massive data to only a few Key Performance Indicators related to different parts of
the systems. These KPIs will be used at the later stages to detect and diagnose the faults.

Figure 2. Warning message given by the computer program, fir further info, refer to Vecchio (2014).

Consequently, a computer program can detect and diagnose the possible faults in the system by following the
trends of KPIs. The program can act as a support to the servicemen by giving hints about what to check and
what to do as the next step, similar to the FDD mechanisms already used in the car industry. Figure 2 shows
an example of a warning message given by the program developed in the present study. For further
information, refer to Vecchio (2014).

6. Conclusions

The current paper introduces a holistic approach to develop a Smart Fault Detection and Diagnosis (SFDD)
for heat pump systems. A SFDD system should act as the heart of smart heat pump system, which is
described at different boundary levels. SFDD should be able to minimize the installation and control errors,
 couple or decouple the renewable energy sources, make load-shifting possible, detect the performance
degradation during operation, avoid unnecessary visual inspections and components replacement, and reduce
the maintenance cost and down-time of the system. In order to develop a smart system which detects all the
common and important faults in the heat pump system, it is essential to first determine what the most
important faults are. A comprehensive study is done on the faults reported to both OEMs and a main
insurance company in Sweden. The results show that faults in Control and Electronics are among the most
common and costliest faults in all types of heat pumps. Faults in Control and Electronics include any fault
related to control unit, electrical faults (such as short circuit, etc.), PCB, display, soft starter, overcurrent and
motor protection relay, etc. To conclude, in order to reduce the number and cost of the faults in the heat
pump systems, it is essential to pay a special attention to the control and electronics in heat pump system.
Furthermore, Smart Fault Detection and Diagnosis should be designed to support the heat pump systems at
commissioning, operation, and maintenance phases. During the commissioning phase, SFDD does the first
system check a short time right after installation to ensure the installer that the system is installed correctly
and operates properly. At this phase, the SFDD system finds the common installation errors, incorrectly
sized equipment, control errors, undercharge and overcharge conditions, and faulty components at both heat
pump unit and heat pump system levels. During the operational phase, a knowledge-based SFDD analyses
the measured data with the assistance of the models and converts them to a few Key Performance Indicators
(KPIs). The trends and behaviour of KPIs helps the SFDD mechanism to detect and diagnose the faults and
performance degradation in the whole system. The SFDD system can be widely integrated and used in both
old and new heat pump systems worldwide.

7. REFERENCES

Bettgenhauser, K., Offermann, M., Boermans, T., Bosquet, M.,et al., 2013. Heat Pump Implementation
Scenarios Until 2030. Ecofys.
Chen B., Braun J.E. 2000 Simple fault detection and diagnosis methods for packaged air conditioners,
International Refrigeration and Air conditioning Conference, Purdue, US.
Katipamula S., Brambley M.R. 2005, Methods for Fault Detection, Diagnostics, and Prognostics for
Building Systems—-A review, Part II. HVAC & R Research, 11:1:169–186.
Kim M. Yoon S., Domanski P., Payne V. 2008, Design of a steady-state detector for fault detection and
diagnosis of a residential air conditioner, Intrenational Journal of Refrigeration 31: 790-799.
Lundqvist P. 2010. System thinking for efficient use of heat pumps part 1: Experiences and perspectives
from Sweden. REHVA journal, no. 17: 69-73.
Madani H., Roccatello E. 2014, A comprehensive study on the important faults in heat pump system during
the warranty period, International Journal of Refrigeration, 48: 19-25.
Payne W.V. et al. 2008. Design of a steady-state detector for fault detection and diagnosis of a residential air
conditioner. International Journal of refrigeration, 3I:790–799.
Payne, Yoon, Domanski. 2010., Residential heat pump heating performance with single faults imposed,
International Refrigeration and Air Conditioning Conference, Purdue, US.
Rossi T.M, Braun J.E. 1997 A statistical, rule-based fault detection and diagnosis method for vapor
compression air conditioners. HVAC & R Research, 3:1:19–37.
Vecchio Daniel, 2014, Fault Detection and Diagnosis for Brine to Water Heat Pump Systems, Master of
Science thesis, Royal Institute of Technology, Stockholm, Sweden.
Venkatasubramanian V. et al. 2002., A review of process fault detection and diagnosis Part I: Quantitative
model-based methods. Computers and Chemical Engineering, 27:293–311.
Venkatasubramanian V. et al. 2002., A review of process fault detection and diagnosis Part II: Qualitative
models and search strategies. Computers and Chemical Engineering, 27:313–326.
Venkatasubramanian V. et al. 2002., A review of process fault detection and diagnosis Part III: Process
history based methods. Computers and Chemical Engineering, 27:327–346.
Yoon S., Payne V., Domanski P. 2011, Residential heat pump heating performance with single faults
imposed, Applied Thermal Engineering 31: 765-771.
Zogg, Shafai, Geering, 2006., Fault diagnosis for heat pumps with parameter identification and clustering.,
Control Engineering practice, Vol 14: 1435-1444.
.

View publication stats