Predictive Maintenance and Pulp & Paper Industry

Pulp and paper production is a vital industry that produces a wide range of products, from paper and cardboard to tissue and pulp. However, the production process is complex, energy-intensive, and requires a large number of machines and equipment to operate. As a result, ensuring that these machines and equipment are running smoothly and efficiently is crucial to the success of the industry. This is where predictive maintenance comes in. In this blog post, the main focus will be on the relationship between predictive maintenance and the pulp and paper industry.

Predictive maintenance is a proactive approach that involves collecting data and analyzing it to predict when equipment is likely to fail. This allows maintenance teams to take corrective action before a failure occurs, reducing downtime and increasing the overall efficiency of the production process. In the pulp and paper industry, predictive maintenance can be used to monitor and maintain a wide range of equipment. The machinery used in the pulp and paper industry includes pulping equipment, such as kraft pulping systems and mechanical pulping systems, as well as chemical recovery systems and bleaching systems. Other important machinery includes papermaking machines, such as fourdrinier machines and cylinder machines, and finishing equipment, such as calenders, rewinders, and slitters. Additionally, various types of pumps, valves, and conveyors are used to transport and process the materials throughout the mill. The industry also has a wide range of supporting equipment like boilers, turbines, generators, and air compressors. Depending on the layout, production strategy and focused products of a facility, all of the machinery mentioned above has the possibility to be crucial. In order to maintain product quality and efficiency of production predictive maintenance applications are a must for the pulp and paper industry.

One of the most effective predictive maintenance techniques for the pulp and paper industry is vibration analysis. Vibration analysis involves measuring the vibration of equipment and analyzing the data to detect abnormal patterns that indicate a potential failure. This can be done using specialized sensors and software, which can detect issues such as unbalance, bearing failure, and gear wear. By detecting these issues early, maintenance teams can take corrective action before a failure occurs, preventing costly downtime and ensuring that production lines are running smoothly. Main focus of vibration analysis is rotational machinery. Vibration data gathered from critical components of an equipment such as bearings of electric motors and gearboxes can shed light on many possible malfunctions of the machinery long before large-scale damage occurs. Predictive maintenance can also be done with the help of IoT-based data collection and analysis. This allows the monitoring of a large number of data points on the equipment and can provide real-time data to the maintenance team. This can help reliability agents to analyze data trends and operational tendencies of a machinery and address potential issues before they escalate.

Sustainability is one of the key concepts the pulp and paper industry is focusing on. Predictive maintenance and sustainability are closely related since predictive maintenance applications aim to increase overall efficiency of a system. Sustainability is the practice of ensuring that resources are used responsibly and in a way that does not harm the environment or future generations. In the pulp and paper industry, this can include measures such as reducing energy consumption, minimizing waste, and using sustainable raw materials. By implementing predictive maintenance, the pulp and paper mills can reduce energy consumption and waste by identifying and correcting inefficiencies before they lead to breakdowns. Additionally, regular maintenance and repair can prolong the life of equipment and reduce the need for replacements, which can help to reduce the environmental impact of the mill. Furthermore, by detecting equipment failures early, the mills can avoid unexpected downtime, which can result in wasted raw materials and energy. Predictive maintenance also allows mills to plan maintenance activities in a more effective way, which can reduce the environmental impact of maintenance activities by minimizing the amount of chemicals used and reducing the amount of waste generated during maintenance.

In conclusion, the pulp and paper industry relies heavily on the performance and reliability of its equipment. Predictive maintenance is a crucial tool for ensuring that this equipment is running smoothly and efficiently, reducing downtime and increasing productivity. Techniques such as vibration analysis combined with IoT-based data collection and analysis, can provide valuable insights into the health of equipment, allowing maintenance teams to take proactive, corrective actions before a failure occurs. Implementing predictive maintenance can help the industry to become more sustainable and cost-effective.

Predictive Maintenance with Electrical Signature Analysis

As a result of targeting more reliable and more sustainable factories by reducing the need for human resource, the increasing needs of the industry, and the costs caused by these needs; machine health and predictive maintenance topics maintain their place among the rising trends of the agenda. The maintenance costs and loss of profit that an unplanned downtime on the production line will cause to the manufacturer can be prevented by monitoring the health of the machines on this line. In this context, investments in predictive maintenance return with the early detection of possible malfunctions.

There are some analysis techniques for monitoring machine health. The most popular among them are vibration analysis, oil analysis, ultrasound analysis, and electrical signature analysis. The diversity of these techniques allows higher accuracy in the diagnosis of the possible malfunctions. While some types of faults can be detected better by vibration analysis, some types of faults are better detected by electrical signature analysis. For instance, electrical signature analysis is used in the detection of broken bar failure, which is the most common rotor failure. Electrical signature analysis can be used as an alternative to vibration analysis in misalignment faults as well.

With electrical signature analysis, faults such as stator faults, rotor faults, bearing faults, load-induced faults can be detected in advance.

Fig. 1 Faults Detected by Different Methods

Electrical Signature Analysis (ESA)

Electrical signature analysis is an umbrella term covering some machine condition monitoring methods that are accomplished by analyzing electrical signals such as current and voltage. Some of these techniques are: Current signature analysis, Voltage signature analysis, Park’s Vector approach, Instantaneous power signature analysis. As a result of these analyzes, reduction in downtime, increase in machine availability, reduction in maintenance costs, better management and planning of maintenance are expected.

With the remote monitoring provided in the facilities where electrical signature analysis is applied, the hazards that the personnel may encounter physically in the monitoring of equipment health are prevented. With electrical signature analysis, the health of any asynchronous motor can be monitored and possible failures can be detected beforehand, without power restrictions. It offers sensitivity in the detection of mechanical faults in motor and load, electrical faults in stator and network problems etc.

Obtaining the Data

Electrical signature analysis does not require sensors mounted on the rotating driveline of the machines. Since the data to be monitored are current and voltage, these measurements can be taken from the phase cables at the motor input. In the motor control center (panel), the phase cables from the driver output to the motor input related to the equipment to be monitored can be detected and current and voltage signals can be obtained with the analog output sensors to be positioned on these cables. Then, by transferring these data to an analyzer or a cloud analysis software via IoT data acquisition devices, the data can be analyzed and possible malfunctions can be detected.

The demands of measurement are precision, accuracy/linearity, high bandwidth, and low power consumption. In order to display the current data, sensors that will reduce the motor current to the input range value of the data collection devices should be used. Sensors such as current transformers, Rogowski coils, Shunt and Trace resistors, Hall-Effect sensors and fluxgate sensors can be used. For voltage measurements, the voltage level is reduced to the appropriate range that analog data acquisition devices can detect with sensors such as voltage/potential transformers, and voltage dividers.

The techniques that make up the electrical signature analysis are used to interpret the obtained data.

Fig. 2 Current Transformer

Motor Current Signature Analysis (MCSA)

MCSA is a method for analyzing and tracking the health of the electrical machine and for detecting the possible malfunctions in the near future with their root causes. It aids in the identification of stator winding issues, rotor issues, coupling issues, load issues, efficiency issues, and bearing issues.

This method employs an asynchronous motor as a transducer, allowing the user to assess the electrical and mechanical conditions. Current signals are taken from the motor and then they are collected by a data acquisition device. After that, the signals are conditioned and spectrum of the signal is obtained. This spectrums are analyzed to detect if there is any matches with fault patterns.

Fig. 3 Monitoring System for Motor Current Signature Analysis

Spectrum of the motor current interprets the frequency components of the current signal with their respective magnitude. In other words, it is the expression of the time waveform signal in frequency domain. This allows patterns in the current signature to be identified, distinguishing healthy motors from unhealthy motors, and even determining which component of the machine may fail. The Fast Fourier Transform is the main instrument in this investigation (FFT).

Voltage Signature Analysis (VSA)

The Voltage Signature Analysis is quite similar with the motor current signature analysis; however, the signal is collected from the motor’s voltage source. It can be beneficial in the examination of stator electrical imbalance along with the analysis of the current signature and the issues caused by motor power in the case of motors. This method is most typically utilized in the analysis of production units.

Instantaneous Power Signature Analysis (IPSA)

Another failure analysis approach based on spectral analysis is instantaneous power analysis. It differentiates from MCSA and VSA with that it evaluates the information contained in the voltage and current signals of a motor phase together, and the demodulated fault component which is named Characteristic Frequency.

The spectrum of the instantaneous power has an extra component which is linked to the modulation induced by the fault, in addition to the fundamental component and sidebands to be formed. This component is known as the Characteristic Component, and it may be used for diagnosing the machine’s status.


Electrical signature analysis has an important role in detecting possible failures. In order to detect these failures, the data must be obtained correctly and in sufficient quantity, then must be analyzed. Sensemore offers its customers the Duck solution, an IoT data acquisition device, to collect and transfer the data to the analysis software. With its 8 channels, the Duck device can take manual, periodic and triggered measurements simultaneously from 8 different analog output sensors. Analog sensors such as current, voltage, temperature, vibration, humidity, pressure etc. can be connected to these channels and the measurements can be viewed in the Sensemore Cloud App.

Fig. 4 Analog IoT Data Acquisition Device: Duck

For electrical signature analysis, Sensemore offers its customers split-core current transformers, Rogowski coils for motors with high currents; and voltage divider sensors to measure voltage values. These are easily placed on the motor phase cables in the electrical control center (panel) and the current-voltage data is transferred to the Duck device. Duck transfers this data to the Sensemore Cloud Application. Thanks to the software in the Cloud Application, this data is analyzed and reports are created for the user about possible failures, also energy monitoring is provided.


  • Bonaldi, E. L., de Oliveira, L. E. D. L., da Silva, J. G. B., Lambert-Torresm, G., & da Silva, L. E. B. (2012). Predictive Maintenance by Electrical Signature Analysis to Induction Motors.

Understanding the Machine Data

Machine data is the set of physical quantities and attributes that the machine has or creates. This data is provided by various sensors, software, and machine manufacturers. Recently, machine data has started to attract more attention as the use of IoT and big data management technologies has increased. Machine data sometimes called machine-generated data; is digital information automatically generated by the activities and operations of networked devices, including computers, embedded systems, and connected sensors. In a broader context, machine data may also contain information generated by suppliers, end-user applications, manufacturers, etc.

We can mainly divide data types into two categories:

  • Quantitative data
  • Qualitative data

Quantitative Data

Quantitative data are numbers or the value of data in countable form, where each dataset has a unique numerical value. This data is any measurable information that researchers can use to make real-life decisions based on these mathematical derivatives for mathematical calculations and statistical analysis. Quantitative data makes the measurement of various parameters controllable due to the ease of mathematical derivatives they come with.

In predictive maintenance applications, evaluations and plans are usually made based on the quantitative data obtained from the sensors. The data collected from these sensors can also vary within itself. The sensors convert the analog signal into a processable digital data and while applying this conversion, it is subject to various rules according to the data to be collected. If the data to be collected is only time-based data such as position, pressure or flow, overall value measurement are usually made. If the data to be collected is also desired to examine frequency-dependent changes, it should be sampled at least 2 times the maximum frequency desired to be monitored.

Vibration, which is one of the most important data used in the condition analysis of rotating machines, can be collected in 2 different types. Vibration sensors used in PLC-SCADA systems are generally overall value vibration sensors. These sensors use certain features such as RMS, Crest, Kurtosis, which it extracts from the vibration signal collected with fixed sampling frequencies at a certain time. These features are actually another data extracted from the measured data. On the other hand, the sampled signals can be analyzed in time-wave form at certain interfaces and the processed signals in different spaces such as spectrum and cepstrum.

Different statistical data can be obtained from a collected signal data, as well as data which is also called “data about data” calculated from 2 different data. These data can be the extraction of power data from current and voltage values, as well as the phase difference from two different vibration data or the stiffness/damping properties of the structure from force and vibration data. At the same time, coherence, correlation and covariance data, which includes statistical comparison results of 2 different data, can also be examined under quantitative data.

The collected signal data is also divided into different groups within itself. Most machine components give rise to certain vibration signals that characterize their separation from others, as well as distinguishing the faulty state from the healthy state. Distinctive features may be different repetition frequencies. For example; there may be a gear network frequency characterizing a particular gear pair, and different sideband gaps characterizing the modulating effects of two meshing gears on their common network frequencies. Vibration signals collected for predictive maintenance can be categorized in several ways. The type of signal obtained also affects the signal processing technique that should be applied. We can categorize vibration signals as in the image below.

Fig. 1 Signal Categorization

The most basic distinction in signals is stationary and non-stationary. Stationarity means that statistical properties are time-invariant. Deterministic signals basically mean that they are composed entirely of discrete frequencies of self-forming sinusoids. Knowing the frequency, amplitude and initial phase (ie at time zero) of these components, the value of the signal can be estimated at any time in the future or in the past; hence it is “deterministic”. Random signals are somewhat more complex as their values cannot be predicted at any time, but for stationary random signals their statistical properties do not change over time. “Non-stationary” means anything that does not satisfy the stationary conditions and can be divided into two main classes, “continuously changing” and “temporary”. There is no hard rule to distinguish these two types, but in general it can be said that transient signals only exist for a limited period of time and are typically analyzed as presence only during that time. To explain this; for example, a decreasing exponential function theoretically decreases to infinity, but in practical terms it has measurable value only for a finite time.

By definition, a stationary random signal has a constant power and thus infinite energy. Cylostationary signals, by definition, have power that varies periodically over time, and hence their total energy is infinite. Other non-stationary signals, such as vibration signals measured during operation or deceleration of a machine, also have a finite length, but are typically more likely to be considered as constantly changing non-stationary signals rather than transient events. Continuously changing non-stationary signals are usually analyzed by time/frequency analysis techniques.

Fig. 2 Sample Signal Types

According to Figure 2, the first two signals (periodic and semi-periodic) are deterministic and consist of completely separate sinusoidal components. For the periodic (saw-tooth) signal, these components are in integer multiples/harmonics of the fundamental periodic frequency. For a semi-periodic signal, the individual frequencies are not all multiples of a harmonic series. This means that in semi-periodic signals, the frequency ratio between at least two components must be an irrational number. A typical example is given by vibration signals from a gas turbine engine with several independent shafts. Each shaft will normally generate harmonic families, but the total signal will be semi-periodic. The third signal (stationary random) does not look very different from the quasi-periodic signal in the time domain, but its spectrum is completely different, there are no discrete frequencies, and its spectral power is constantly distributed with frequency. The example shown is ‘white noise’, which has a uniform spectrum in the frequency range considered.

The cyclo-stationary signal is an amplitude modulated white noise. Signals in which amplitude modulation of a signal (with a single frequency) results in pairs of sidebands in the spectrum spaced around each frequency component modulated by an amount equal to the modulation frequency.

A typical cyclically non-stationary signal is not shown in Figure 2, but may be similar to the cyclo-stationary signal there, except for example, that the period of the modulating signal is not constant but changes deterministically with time.

Qualitative Data

Another important category of machine data is qualitative (Meta) data. Metadata is data added to an event to describe the conditions under which the event occurred. For example, every time you take a photo with your phone’s camera, metadata about the photo is automatically generated, including the date the photo was taken, the aperture of the camera lens, the exposure time, the GPS location, and more. Similarly, vibration, temperature, speed, etc. collected in predictive maintenance in addition to information such as the collected bearing model, the connection of the pedestal to the ground, the information that characterizes the system, such as the coupling type, are analyzed by correlating them with quantitative data in fault root-cause diagnosis and maintenance planning.

Machine-generated data is raw and factual, often providing a simple record of the value of an event or a particular parameter at a particular time. Machine health software, predictive maintenance platforms, are used to monitor data at another time and correlate it with additional machine-generated data and data from other sources. Adding context to data answers questions such as:

“Where did this data come from?”

“What do these data represent?”

“When was this data collected?

Answering these questions contextualizes data and turns it into knowledge. At the knowledge level, we begin to analyze, understand and develop insights into the relationships that exist in the data and what it tells us about the overall state of the system. Whether we look at data from a service or security perspective, the goal is to use the data to make a concrete determination or prediction about something.

Machine Data Analytics tools follow the basic DIKW pyramid for processing machine data. First, data is collected from various sources in the network. Next, an AI application uses algorithms to sift through data, identify trends and monitor changes. The information is then extensively analyzed and correlated system-wide to generate new knowledge and insights. Finally, when insights are reported to users, someone can take action on the insights to improve the state of the system.


  • Randall, R. B. (2022). Vibration-based condition monitoring: Industrial, automotive and Aerospace Applications. Wiley.

Preventive vs Predictive Maintenance

As a part of the digitalization process of manufacturing, the industrial internet of things (IIoT) is the use of smart sensors and AI/ML algorithms to enhance production and industrial processes. Every machine speaks with several measures such as vibration, temperature, voltage, temperature, pressure, sound, etc. The only way to understand them is by converting the raw data into meaningful diagrams through data analysis. This is what the whole predictive maintenance (PdM) concept is trying to do.

Preventive Maintenance

First of all, what is the aim of these two different maintenance methods before going into detail? In parallel with the definition of maintenance, Predictive and Preventive Maintenance methods are designed for increasing asset reliability and minimizing the cost of failures through the process of monitoring the condition of the machines. However, their approach is different. When we look at the current attitudes towards maintenance in the industry, the most preferred method is preventive maintenance. Preventive maintenance includes regular and routine maintenance to prevent downtimes and reduce the possibility of unexpected machine failures. This method relies on past statistics and lifetime data, while predictive maintenance focuses on monitoring and analyzing data from the current condition of the machine in the field or operation.

Predictive Maintenance

The starting point of predictive maintenance is the same problem as preventive maintenance, which includes efficiency loss due to downtimes and unexpected stops in the production facilities. On the other hand, predictive maintenance uses data analytics to analyze machine health via vibration, pressure, and temperature data in order to predict machine malfunctions before unexpected stops occur. When the AI algorithms detect the anomaly in the dataset, in other words, observe irregular behavior from the standard parameters, the system alerts technicians in the field to check the condition of the machine.

Main Benefit of Predictive Maintenance

Cost reduction seems to be the key benefit of condition-based predictive maintenance. Since it does not require skilled staff and maximizes the efficiency of the monitored machine, it directly saves money by creating planned maintenance optimization. However, this cost-saving is not the most significant benefit of predictive maintenance; the bathtub explains it. This theory says that if any equipment is newly assembled, it starts working with all the risks in the early failure zone, and the risk of failure is high until it moves into the safe operating zone. If you do not carry out your planned maintenance activities when necessary, you increase the risk of failure every time. The production facilities that have the technology to predict when equipment could fail with condition monitoring techniques do not only save money through efficient production capacity also predictive maintenance directly extends the lifetime of the machines used.

Obviously, the side benefit of predictive maintenance is the continual and steady production process. Since the system predicts the machine malfunctions before it happens, required maintenance and repair actions are taken, and the production line does not stop anytime. 


Regular and periodic operations with preventive maintenance have been rapidly abandoned, and it seems that it will not be able to take place in this market no longer again. In order to maintain their competitiveness level, manufacturing companies need to take action immediately regarding the digital transformation in their production facilities. In short, digitalization in manufacturing with predictive maintenance starts to cease as a choice and becomes indispensable.

First Step in AI Based Predictive Maintenance

With the increasing demand of IIoT in the age of digitalization, data-driven maintenance optimization of industrial equipment is one of the most trending topics among maintenance professionals. At the beginning, data gathered by various sensors placed on the machines or embedded in the machine itself are pre-processed. Later on, this data is used for Decision-support systems with Artificial Intelligence (AI) tools. In this article, we will talk about how artificial intelligence applications can be implemented in the predictive maintenance area.

At the beginning Predictive Maintenance was just a rule-based system. But it was not even close to solving the problem. In rule-based predictive maintenance; maintenance is performed based on hard-coded thresholds, and an alert is sent if a measurement exceeds the thresholds. E.g; It is rule-based predictive maintenance for a pump to generate an alarm when the vibration RMS value exceeds 7.1 mm/s. This predictive maintenance application is widely used with the help of ready-made ISO standards along with traditional methods. 

Predictive Maintenance has evolved over time from rule-based predictive maintenance to machine learning-based predictive maintenance. In predictive maintenance based on machine learning; It uses advanced analytics and machine learning techniques to predict when the next failure will occur and pre-maintain accordingly. But this issue is pretty huge. In this article, anomaly detection in rotating machinery, which is one of the first steps to be applied in predictive maintenance based on machine learning, will be emphasized.

Anomaly Detection in Rotating Machinery

A rotating machine such as a pump or fan is going to deteriorate from regular use and will begin to produce what we might call ‘anomalies’. This should not be considered a complete shutdown state, but rather a warning that the machine is not in optimal condition and needs to undergo maintenance.

The first tool for detecting these anomalies is sensors. The sensors and sensor outputs used in anomaly detection have a very important place in the anomaly detection performance. Consider sensors with the right capabilities and the features are used. But still there is a risk for the data. If the obtained data from these sensors are not transferred to the model correctly, it can cause unnecessary loss of time and effort with false-positive alarms, and very serious material and moral damages with missed alarms. If there is no historically defective signal and healthy signal labeled data from the machines, supervised learning is not in this equation. In this case, we can train the machine learning model on a training set with only “normal” samples and use a distance measure between the original signal and the predicted signal to trigger an anomaly alarm. But we need to define “normal” and “abnormal” at this stage.

Feature Extraction

Rotating equipment naturally generates vibrations and operates on these vibrations. The aim is to keep these vibrations at acceptable levels to ensure production reliability. Therefore, although various sensors can be used in predictive maintenance applications in rotating equipment, the most basic measurement unit is vibration. At this stage, what information the signals received from the accelerometers contain and the processability of this information into the machine learning model constitute the basic performance parameter in anomaly detection. Two main features can be extracted from the acceleration signals collected from the accelerometers -features extracted in the time domain and features extracted in the frequency domain-.

Fig. 1 Some Extracted Features from Acceleration Signal

While metrics such as RMS, Crest, Kurtosis, Peak can be obtained from the time domain, features such as total harmonic distortion, 1X,2X,…,10X harmonic indications, spectral centroid and sideband energy are extracted from the frequency domain. At this stage, it is highly important to transfer the features in the frequency domain to the model to create a reliable machine learning model. The frequency content is the signature of the normal or abnormal state of that machine. On the other hand, it is equally important to enter the speed information into the model for harmonic analysis independent of the vibration sensor. When we include each frequency amplitude in the frequency domain from a 3-axis vibration sensor to the machine learning model as a feature, it is seen that more than 20,000 features are transferred to the model in each measurement, depending on the sensor bandwidth. So here, the machine learning model can compare many features in multidimensional mathematical space better than the human eye can distinguish between data in 2D and 3D space. In this way, even the most sensitive anomalies can be detected according to the resolution of the model.

Machine Mode Analysis

There are hundreds of different machine types that vary according to the usage area in the industry. Many of these machines operate in cycles, power and processes arising from factors such as constantly changing production speed, raw materials and processes. E.g; A simple rolling mill will be operated at different speeds depending on the desired material quality, material thickness and production speed. Vibration data from the machine will also vary depending on the speed. In rule-based predictive maintenance applications, anomaly alarms will occur at every speed change, but this is not true. At this stage, each process parameter (speed, power, etc.) must be included in the model and first of all, the machine modes must be determined. Vibration data were collected at 4 different speeds from a test roller. When these data were inserted into the machine learning model, it was observed that there were 6 different modes. In these modes, 4 of them work at different speeds, and 1 of them is the mode in which the machine does not work. The last mode is the mode in which the anomaly occurs.

Fig. 2 Scatter Plot of Machine Modes

If the measurements are not colored in the chart above, we can observe that they are in 5 different groups. However, as can be seen in the graph, the model has 6 different modes. The reason for this is that the machine operating cycle in the cluster, which is divided into 2 groups, is the critical speed for that machine, creating a new abnormal operating mode by creating mechanical looseness in time in the machine assembly and coupling over time as you can see as purple measurements in Figure.2. Although the standard deviation of the data in the cluster, which is divided into 2 groups, is clearly evident compared to the other clusters, it is very difficult to distinguish that there are 2 different modes in that cluster.

Fig. 3 Machine Modes of Spectrum Signatures

Two measurements operating in the same mode are as above and can be clearly distinguished between blue and purple plot in the spectrum. However, it is very difficult to distinguish these two measurements by looking at the average Vrms values. Of course, another reason for this is that Vrms is insufficient for high frequency vibrations, although it varies from machine to machine, Vrms in some fault types and Grms static alarms in some fault types may be insufficient for anomaly detection. Therefore, transferring all spectral information to the machine learning model ensures optimum results in anomaly detection.

Fig. 4 Waterfall Plot of 6 Machine Modes


In AI-based predictive maintenance applications, in the absence of historically labeled data, supervised learning is not possible, so anomaly detection using unsupervised learning algorithms will be the best start for the first step. In order to implement this, meaningful features from the data received from the sensors should be included in the models. After the anomaly detections are approved by the maintenance personnel inspection, the semi-supervised learning model can be started by labeling the detected anomalies with their root causes. With the labeling of the data, the root-causes of the anomalies of the equipment can be estimated after the anomaly. Therefore, a generalized AI model based on data is very difficult due to variations such as transmission types and working conditions that change according to industry. For this reason, starting the applications of generalized anomaly detection algorithms using machine-specific transfer learning methods will be the first step in the process. Then, the detected anomalies are labeled by the field workers on the system and root-cause analysis is made. By all these efforts, the first step of the AI-Based Predictive Maintenance journey has been taken.

Fig. 5 Flowchart of Algorithm

Importance of Integration of Condition Monitoring into Existing Systems

We talked about condition monitoring and its importance in our previous blog post. The basic principle in condition monitoring is to detect failures in machinery and equipment to predict using equipment-specific condition parameters, such as temperature, vibration, pressure, current, etc. and to prevent destructive downtimes. One of the most important factors affecting the performance of condition monitoring applications is the ability to integrate into existing systems.

When we talk about an industrial telemetry system, the first terms that come to mind are; there will be SCADA (Supervisory Control and Data Acquisition), DCS (Distributed Control System), PLC (Programmable Logical Controller), and RTU (Remote Terminal Unit). At this point, the fact that condition monitoring systems talk to such traditional data collection and control systems increases the success rate of the condition monitoring system significantly. The communication here should be in the form of both data sharing from IIoT platform condition monitoring systems to traditional PLC-SCADA systems and data sharing from traditional PLC-SCADA systems to condition monitoring systems.

Fig. 1 PLC-SCADA Systems to Condition Monitoring Systems

The first of these integrations’ value propositions in terms of predictive maintenance; is to optimize measurement strategies. In other words, prevention of data pollution. Trying to collect data continuously, especially in equipment that works intermittently, working at variable speeds and loads, creates meaningless data pollution. The data collected in systems that cannot be correlated with each other or that require very different parameters to be correlated becomes meaningless and dirty. There are many different methods to overcome such problems, but first of all, it is necessary to understand the nature of the system correctly and to get the maximum benefit from the integration with the right methods.

Let’s explain these cases one by one by going through them;

Variable Speed Machines

In order to create the desired process parameters in many sectors from FMCG to Automotive, from Energy to Iron-Steel and Metal Industry, rotary equipment is constantly working at different cycles according to these process parameters. In this case, the most basic method used in predictive maintenance creates problems in vibration analysis. False-positive alarms may occur as vibration values will change according to the equipment cycle. On the other hand, it may give wrong outputs in the calculations of the root cause of the failure and the remaining useful life. In this context, it is insufficient to set upper/lower limits or observe deviations, which are the most basic methods to be followed.

Fig. 2 Variable Speed Machine Vibration Signal and Spectrum

On the other hand, it may give wrong outputs in the calculations of the root cause of the failure and the remaining useful life. These are parameters that can change very easily with dozens of factors, from the fact that the environment in which the equipment is located is different from the environment in which it is tested, to the way the operator uses the machine. At this point, the integration of online condition monitoring systems with traditional data collection and control systems constitutes the key solution.

Performing the analysis of the data by taking measurements under constant environmental and process operating conditions will allow you to reach optimum results. For instance; the characteristics of vibration signals received from an electric motor operating at 1500 RPM and the characteristics of vibration signals of the same electric motor operating at 2500 RPM are very different. In determining the root cause of the malfunction, the vibration signal alone will be insufficient. The technique used in vibration analysis is harmonic analysis, that is, the root causes of failure are estimated by examining the equipment excitation frequency and harmonics. At this point, since the harmonics at 1500 RPM and 2500 RPM will be at different frequencies, it will not be possible to predict the root cause of the fault. With the integration of online condition monitoring systems with traditional data collection and control systems, it is possible to take measurements in certain cycles by giving a speed, etc. parameters can also create the trigger mechanism on the condition monitoring system. Similarly, predictive maintenance outputs can print both measurement metrics and machine learning outputs to data acquisition systems such as SCADA.

Non-Continuous Machines

Although condition monitoring systems are generally applied to machines that operate continuously with high downtime costs, when criticality analysis is made, condition monitoring systems can be applied to machines working in shifts or machines that make stop-starts if they are more critical. In this type of machine, the measurements taken at certain periods are insufficient. While it causes the measurement to be missed, on the one hand, it also negatively affects battery power consumption in wireless sensors, as mentioned above. With the integration of PLC-SCADA systems into the condition monitoring systems, optimum results can be obtained with the SCADA system triggering the status monitoring system while the machine is running.

Fig. 3 Non-Continuous Machine Vibration Trend

Integration of condition monitoring systems into PLC-SCADA systems is not only at the point of triggering and optimizing measurement periods, but also at the point of sharing all collected data with condition monitoring systems, along with current process parameters and predictive analytics applications, general equipment efficiency, and production and maintenance outputs such as key performance indicators. It is also possible to monitor and control online.

Machines with Impulse and Transient Responses

Packaging machines, presses, cam mechanism machines are widely used in all kinds of industries from automotive to pharmaceutical industry, from fast moving consumer goods to white goods. Although this type of machines can work continuously and at constant speeds, they are equipment that make predictive maintenance applications difficult due to both the fact that they have a complex mechanism and the mechanical effect they create. For example, the data received from the accelerometer placed in the electric motor of a pressing machine may generate high vibration values ​​due to the mechanical effect of the press impact, even if there is no fault in the electric motor.

On the other hand, special packaging machines, which are used in different ways but for the same purposes in every industry, can create meaningless and distorted signals by creating temporary reactions during packaging. However, the basic parameters that can be used to detect the malfunctions that may occur in the rotating equipment of these types of machines at an early stage are also vibration, temperature, force, etc. mechanical metrics. At this stage, in order to benefit from the predictive analytical power of condition monitoring systems, such parameters that IIoT sensors cannot provide but exist in PLC-SCADA systems should be integrated with condition monitoring systems. In this way, unplanned equipment stops are prevented by increasing the predictive analytical performance, and a more efficient decision-support mechanism is created.

Integration Options

In a business that already has a data collection system, you need to integrate IIoT sensors with new capabilities into your existing system. On the other hand, it is also possible to integrate existing metrics and records into the condition monitoring system. To exemplify, in order to benefit from the high analytical capabilities of online condition monitoring systems, it is necessary to share the existing system parameters with the status monitoring system, while the alarms created as a result of the analysis and estimations should be seen in the PLC-SCADA system and trigger the machine when necessary. There are many methods for integrating online condition monitoring systems into existing systems. These methods are explained in detail in our previous blog post.


With the integration of traditional methods into online condition monitoring systems, both periodic measurement strategies are optimized and fault root cause estimations are made more accurate. At this stage, both a more sustainable machine and higher profitability are obtained by making detailed predictive analytics applications, especially with the integration of process parameters into online condition monitoring systems.

It will become inevitable to establish communication between all interconnected processes. Until that day comes, it would be a right step to prefer integrated solutions and ensure that the investment made is long-term. If you do not have a digital transformation team or your technical skill set does not cover these developments, working with a provider who can support you at this stage plays a critical role in bringing the project to life.

Condition Monitoring in CNC Machines and Case Study

For many years, machine tools have been the backbone of many manufacturing industries. Machine tools also play an important role in the global economy nowadays. These machines, which transform complex designs into products by processing materials of different hardnesses, need to be maintained for performance, quality, and ultimately profit.

The spindle unit of CNC lathes affects the reliability, overall production efficiency, and stability of the machine. During machining, the spindle is subjected to extreme loads and thermal conditions. Cutting forces transmit, in form of static and dynamic, through the tool system to the spindle bearings. Issues in the spindle are one of the main sources of downtime for CNC lathes in the manufacturing industry. According to a study conducted in Germany where maintenance information was collected from 250 machine tools in the automotive industry, there are four main reasons that cause machine tools stoppages. Among these, spindle and tool changer equipment malfunctions cause 26% of downtime. And as seen in the graph, this is the second biggest issue after drive axle failures. Due to the high cost of spindle downtime, it is common practice in the manufacturing industry to have a spare spindle for critical machine tools in the production system. By this means, a defective spindle can be quickly replaced and sent to a spindle repair shop. This minimizes repair time and therefore non-production cost.

Fig. 1 Machine Fault Types

Considering that machine tools are the most productive equipment of manufacture, it is expected to make them operate at their maximum capacity and avoid production losses caused by unplanned repair costs, downtime, or quality problems. Thus, keeping them in optimum condition is a priority in highly efficient production systems. This situation can be achieved through a preventive maintenance strategy for critical components of the machine tool. Thanks to this strategy, the maintenance task can be pre-planned according to the actual maintenance need of the equipment. For this reason, a decent maintenance plan which includes condition monitoring for spindles becomes important to reduce downtime and total cycle cost, and to increase the service life of the machine tool.

Condition Monitoring on Spindles

In condition-based monitoring, there are several monitoring techniques to monitor possible failure conditions of components. Among these, oil analysis, thermography, ultrasonics, and vibration are the most common. Each of these techniques has different abilities to predict startup failure. As we mentioned in the P-F curve in our previous blog on condition monitoring, malfunctioning machines can signal failure with increased vibration levels before contaminant particles on the oil are detected. Thanks to its versatility and reliability, vibration analysis as machine tools has become one of the most accepted and used for rotary equipment. Vibration monitoring detects deteriorations such as wear, imbalance, misalignment, and fatigue in parts that are under rotational or reciprocating motion.

Fig. 2 CNC Lathe Spindle

Machine tool spindles are complex delicate assemblies with a considerable number of components. Each of them plays a vital role in the functionality of the spindle. The individual performance of the components contributes to a certain extent to the overall performance of the spindle. Bearings in machine tool spindles are the most sensitive component in practice. Consequently, any major issues are usually reflected and detected through the bearings. Most of the time, other issues affect the performance and working life of bearings. Therefore, bearing issues are the most common malfunctions encountered in the spindle.

Roller bearing life (L10) is a nominal calculation at the design stage that does not take all factors, which may affect the bearing life, into account. Due to various static and dynamic loads such as working at variable speeds, lubrication conditions, vibrations generated during machining, unbalance and assembly tolerances, roller bearing life ends before the calculated cycle. Therefore, periodic maintenance of the spindles is of utmost significance in order to prevent crucial stoppages and high damage to the machine.

Case Study

As we have just explained above, monitoring the condition of the spindles has great importance for CNC machines. In the industry, such studies are usually carried out every three or four months, or the equipment health is monitored with the built-in temperature sensors in the grinding wheel spindle. Nevertheless, such maintenance strategies can have costly consequences for the long haul. Therefore, condition-based maintenance is the most effective method for achieving maximum performance. At this point, applications in the industry monitor systems by taking measurements at certain time intervals. This technique does not constitute a very applicable process as it is not known whether the CNC machine is working during those time intervals or not. Besides, parameters that vary according to the application such as fly cutters, machining surface, and rotation speeds also make the situation even more chaotic. Sensemore solves this problem with the Trigger device. Trigger sends the measurement order to the sensor by transmitting the 5V signals to the receiver with a code to be returned at the end of each process. Thus, it can monitor the vibrations of a CNC lathe in the grinding wheel spindle at the end of each machining and in a way that the parameters remain the same in every measurement.

As we stated above, various factors such as working at variable speeds, different types of machining tools, and materials change the vibration characteristics that may occur in the grinding wheel spindle. It would be inaccurate to make inferences about the machines’ health from these measurements. For this reason, operating the grinding wheel spindle at a constant speed for a short time at the end of each machining and taking vibration measurements during this process will create a correct trend as it is within certain boundary conditions. At this point, thanks to the code added to the CNC lathe, it sends a 5V signal to the Sensemore Trigger device after each part is machined. Sensemore Trigger gives the measurement order by triggering the Wired with the signal it receives. The whole process is completed in as little as five seconds and analyzes are performed on the Sensemore cloud application.  

Nowadays, the human factor in the industry continues to decrease gradually due to advancing technology, and all manufacturing, maintenance, and quality processes are carried out in an automated manner. Vibration measurements, especially in maintenance and quality control processes, are starting to keep up with this technology. Otherwise, vibration measurements create a critical loss of labor and time. Sensemore, thanks to the portable accelerometers, application-specific hardware, and cloud-based analysis program they developed, offer easier and more reliable maintenance and quality control processes.

Hidden Costs in Maintenance Organizations

Most of manufacturing companies conduct several projects and studies on reducing unit costs and gain sustainable advantage in competition in order to exceed its existing situation to an upper level. But it is not enough to just make a machine more efficient. You also need to reduce the cost of it. To decrease the total cost and increase the savings, you should also concentrate on maintenance costs, which is between 15% to 40% of total costs of a manufacturing plant. On the other hand, the reactive approach (fix-when-broken) is also a widespread option. This states that these companies embraced a short-term approach on savings and efficiency. In short-term, this approach is far cheaper than other methods. But this approach concentrates on visible issues and completely ignores the costs related to the system, itself. In this entry, I will concentrate these hidden costs.

Improvement Processes

In the maintenance operations, downtime and spare part costs are the most widely known issues. But the other costs are hidden in the processes. If you concentrate on downtime and spare costs, you probably miss the costs of processes and operations. But this is a trap so many companies fall. When the costs of operations increase, most companies reduce training cost, the number of auditors, maintenance frequency and so on. In short term, they save money and survive, but in long-term this is not sustainable. The short-term plans provide unsustainable improvement policies which cause a loss of money in long-term.  In order to gain sustainable advantage in long-term, these improvement processes must be long-term. This issue not only requires decisions, but corporate policy changes.

In order to improve operational reliability in long-term, modern predictive maintenance (PdM) technologies are really important. PdM is actually a very cost-effective approach, which saves 8% to %12 more than Preventive maintenance and 40% than fix-when-broken approach. At this point, we need to mention about Condition-Based PdM and Traditional PdM. The biggest difference between them is time. Both monitor the health and condition of machines such as pumps, motors, reductors, fans etc.  However, the traditional one consists of periodical measurements performed once in 60 or 90 days. On the other hand, certain problems can occur between the measurements due changing working conditions, temperature or load and cause catastrophic problems. Condition-based provides continuous monitoring and provides opportunity to maintenance teams react faster and plan their schedule.

Energy Consumption and Waste

As I mentioned above, the visible costs are widely known. But hidden costs can cost more than those. The vibration, misalignment, assembly and other issues issues can cause defects on bearings, couplings and gaskets. Due to this simple issue, machine consumes more energy. Let me give you an example. Consider there is an imbalance of 0.11 Watt/ If there is 50 g imbalance in a 500 mm shaft, there would be 2.75 kW energy waste. Moreover, with the help of resonance and increased damping capacity of oiled interfaces, the imbalance will be bigger, so the waste. Generally this simple imbalance issues cost 5% more energy consumption. In rotating machinery working in continuous manufacturing environments, even 1% can mean millions of dollars. In addition to this, also carbon emission will be larger.

Fig. 1 Maintenance Types


Downtime was very underestimated in the past. These days people get familiar with the issue. But downtime is not just the time the broken machine does not work. It is also the other machines cannot work, the personnel cannot work and the personnel work for overtime to fix the issue and produce ordered amount. A skipped maintenance work can pave the way for other elements to stop and unless the root cause is well analyzed, the frequency of downtime will increase. In addition to these, a manufacturing line after a fix lose 10% of its efficiency in re-start period.

Human Resources & Training

The other hidden cost is human resources and training. Let us start with an example. Human Resources Department of a factory hires a technician for PdM and instead of Condition-Based systems, they give this technician a hand-held vibration analyzer. They train the technician, and the technician starts getting measurements point by point, one by one. After 2 years of work, the technician quits the work. The factory needs a new technician and information the technician collected these 2 years is lost. They need find a new technician and collect the data and face the same problems in maintenance as happens before him/her. Or COVID-19 breaks out and due to the limitations, the technician cannot enter the plant as commonly occurs in these days. Since there is no condition monitoring system is established or any one else in the factory can fill up the technician’s position, the conditions of all machinery go dark and malfunctions happen.

To sum up, without ignoring the visible costs, you should also consider hidden costs, a manufacturing plant has and build up maintenance operations. Since the short-term plans will always cause long-term loses, the long-term plans should be made, and corporate culture should be reshaped. Condition-Based is a strong tool to increase efficiency and reduce cost in long-term. Beyond the advantages of the traditional PdM, it saves both money, workforce and energy. With the increased reliability, maintenance teams can plan their schedule, define priorities and save workforce.


  • Mukesh A. Bulsara, Anil D. Hingu and Pratik S. Vaghasiya(June 2016), Energy loss due to unbalance in rotor–shaft system
  • E. Estupinan, D. Espinoza, A. Fuentes (2008)Energy losses caused by misalignment in rotating machinery: A theoretical, experimental and industrial approach

What is Condition Monitoring?

Condition Monitoring is the process to detect a fault in a machine or equipment by observing parameters such as vibration, temperature and current. Condition Monitoring systems give the opportunity to spot faults that can lead to catastrophic failures and allows maintenance teams to program their schedule. This provide a unique value proposition for taking care of life-shortening faults & malfunctions before they make equipment unserviceable. Condition Monitoring Techniques are generally used for rotating equipment and auxiliary systems such as compressors, pumps and gear boxes. There two types of it; Online Monitoring and Offline Monitoring.

Offline Monitoring is used for low and semi-critical equipment and generally vibration is recorded periodically and analyzed. Offline Monitoring is performed with a handheld measurement tools and the data on the defined equipment is measurement. This process is both exhausting and requires qualified workforce. This method is used when periodical measurements are enough and generally vibration is measured. On the other hand, Offline oil analyses are commonly performed. For these analyses, viscosity, water levels and quality of the oil defined by taking samples.

Online Monitoring is performed with the sensors deployed on the certain points of certain equipment. Although it is more costly than offline, it provides great benefits where there is no qualified personnel available for offline monitoring; continuous production lines such as automotive, concrete, cement plants or plants containing critical equipment such as Natural Gas Compressor Stations, Wind turbines etc. With the help of online monitoring, you can stop or program schedule for maintenance before a fault causes a catastrophic failure. For the scope of HSE, you can get safer measurements by not using your personnel in toxic and explosive environment. Typically, this method is used for rotating equipment, secondary and auxiliary systems such as pumps, compressors and internal combustion engines.

There are various methods applied in industry for this. The methods vary from machine to machine but generally vibration, temperature and current are monitored. At this point, choosing the most suitable method makes the job easier. For example, a current sensor can be applied for an electric motor to detect failures. On the other hand, a current sensor cannot provide you a meaningful data for pumps, compressors and fans. For these kinds of machines, it is better to use a vibration sensor. Due to this reason, the equipment on the manufacturing line should be classified and determined about how critical they are. This will lead to choose the right sensor and method. If a critical equipment will be measured and continuous data should be kept, online monitoring must be used.

In Condition Monitoring systems, the time between failure and detection of a fault is a critical parameter for choosing the right sensor. On the graph above, we can check P-F curve. This curve represents the relation between potential failure (P) and functional failure (F). Potential failures are the point where the equipment starting to show defects. For example, from the moment when GRMS value of a bearing exceeds 3 G, till the moment when this bearing is unserviceable; there are 15 days. On P-F curve, you can measure time from horizontal axis and the state of the equipment can be seen on vertical axis. The bearing in this example should be kept under monitoring for potential failures.

Condition Monitoring is a critical tool to detect the time interval on P-F curve. By using these methods, you can maximize this time interval. As you can see on the curve, oil analyses, acoustic emission, thermography and other methods exist. But the important point is maximizing the lifetime of an equipment instead of changing it regularly. This increases the efficiency and decreases the costs.

Rotating equipment is a term defining motors, reducers, pistons and centrifugal machinery. In industrial applications, condition monitoring is generally applied to rotating equipment. The most common method for these items is vibration analysis. Vibration analysis is the technique to measure vibration severity and spectrum. The data gathered are used to define the condition of machine and detect potential faults. By this method, problems such as imbalance, bearing faults, looseness, crooked shaft and even cavitation can be observed. Vibration allows you to detect potential faults up to 3 months before faults become failures. If you are interest in vibration analysis, you can reach our blog post by clicking the link above.

There are various representations of data in condition monitoring systems. The most common method is following up trends of parameters. Following up trends is performed by collecting data for a certain time interval, comparing data day by day and analyzing it regularly. Trends are analyzed to determine potential failures when certain limits are exceeded. In Online Monitoring systems, the sensors are set to measure periodically and transmit telemetric data and set alarms with certain thresholds for potential failures. By this way, the critical equipment in your production line can be monitored and the monitoring process can be automized. On the other hand, defining overall limits instead of interpreting data/spectrum cannot always spot potential failures. The certain decisions should be made after interpreting spectrum. Each piece of equipment provides different spectrums. Some frequencies peak up when a defect occurs. But trends show overall values and cannot define these faults. To solve this issue, Sensemore allows you to set alarms for different frequencies.

These systems provide unique value propositions such as minimum maintenance cost, energy-saving, minimum downtime and less spare part cost. Consider that VRMS values are measured in a system. By the time, VRMS value in a component increases and you spot a potential failure or set an alarm for a limit and system informs you. You send redirect your maintenance personnel or schedule their time for fixing. By this way, you can save energy both from that machine and the energy of your workforce. Also, by avoiding a catastrophic failure you minimize downtime and spare part cost.

The current state of technology allows online tools to be available in the market. The sensors and the software connected to internet can provide you measurements all day. These measurements are kept in integrated platforms to monitor and analyze. These tools give you the opportunity to solve a problem when it occurs or to schedule your time to fix it. To perform these tasks, all these tools are connected in the system such as sensors, receivers, software and mobile.

Instead of a reactive approach, these systems provides you a proactive/predictive approach to maintenance and by the integration to your facility, you can monitor the health of your equipment and show you a clear picture for all your equipment.