"Predictive Maintenance is based on taking measures before an actual breakdown occurs thanks to a number of analyses. That maintenance method enables 25-30% decrease on maintenance efforts, 35-45% on breakdowns and 20- 25% increase on investment returns. Thanks to predictive maintenance, part replacements can be planned for downtime. This decreases total costs."     Asst. Prof. Selçuk MISTIKOĞLU İskenderun Technical University, Faculty of Engineering and Natural Sciences, Department of Mechanical Engineering 1. INTRODUCTION Maintenance is defined as the set of tasks that involve activities like changing, lubricating, cleaning etc. of components that are worn or have to be changed periodically or have a life time that expired and making adjustments according to user guides. Aim of maintenance is to decrease costs, increase production efficiency and product quality, ensure production continuity and personnel safety, and increase useful life. When performing maintenance and repair works; ensuring machines, tools, equipment and systems to run efficiently for longest possible period of time is essential. [National Occupational Standards, 2009]. Historical development of maintenance philosophy is given below (Figure 1.) [Köksal and Uzun, 2016]. 2. MAINTENANCE TYPES The aim of maintenance activities is to keep the whole production system or specific equipment in operation. If those activities are performed as immediate defect handling, this is called REACTIVE maintenance. If scheduled maintenance is performed in a controlled way this is called PROACTIVE maintenance. In order to ensure manufacturing machines to be active all the time, engineers have developed various maintenance methods from past to present. Those maintenance methods are as follows: Maintenance and Repair, Run to Failure, Preventive Maintenance (Periodic Maintenance) Predictive Maintenance (Maintenance Based on Condition Monitoring), Proactive Maintenance and Reliability Based Maintenance. See Figure 2. Maintenance and Repair: Maintenance and repair are a whole set of activities for ensuring all equipment, systems and machines in plants to function properly with maximum performance Run-to-failure: This method is performed when a breakdown occurs and this method requires minimum personnel. Although it is used commonly all over the world, large companies that keep pace with advanced technology left this method behind. Preventive (Periodical) Maintenance: This method which is also called time based maintenance or scheduled maintenance is especially applied on critical equipment that will halt the production or will have a high repair costs. In order to maintain production and avoid high costs, those workbenches and equipment are subjected to periodical cleaning, part replacement, lubrication and oil change applications. Predictive (Based on Condition Monitoring) Maintenance: Predictive maintenance method is based on principles of comparing actual measurement results, performance information and lubricating oil samples from active machine parts with manufacturer set standards and deciding on actions according to statistical information and experience of many years. Most important advantages of predictive maintenance are: Avoiding breakdowns and high repair costs and avoiding production loss during repairing period. Thanks to predictive maintenance, plants can determine worn out parts, order spare parts and replace them during first scheduled stop. As a result, repairing time period decreases, customer dissatisfaction and extra costs are eliminated. [Köse, 2003]. Predictive maintenance is now ab increasingly accepted method in the world. It is discussed in detail below: Proactive Maintenance: Proactive maintenance uses data from maintenance methods in order to define the problem and isolate root causes. As a result, equipment life increases and most of random failures are prevented. And repairing similar equipment over and over is avoided. Reliability Based Maintenance: Reliability based maintenance is a combination of proactive, predictive and periodic maintenance methods. Root causes and history of the breakdown are defined with help of a number of analysis and measurements. Failures that can be tracked with measurement methods are monitored in the scope of predictive maintenance. Failures that cannot be tracked are tried to be monitored in the scope of preventive maintenance. Quality of the maintenance is checked by proactive maintenance. [Köse, 2003]. 2.1. Choosing Maintenance Method Maintenance method has to be chosen according to conditions of the plant and production and montage floors, all kinds of machines taking role in manufacturing at any level. Cost-benefit analysis should be applied before choosing a method. The following factors should be taken into account when choosing a method: • Criticality of the machine in terms of production, • Operating principles of the machine, • Type of the machine, • Non-stop or intermittent operation, • Frequency of breakdown or defects, • Operating and ambient conditions of the machine, • Information about project, • Engine load • Specifications of the machine. Predictive maintenance, which is prominent than other maintenance methods, enables to perform maintenance accurately and adequately before actual breakdown occurs according to data coming from the equipment that are being tracked. However, only 12% of maintenance projects apply that method. Yet, predictive maintenance offers a range of opportunities for plants. In short, the Predictive Maintenance is based on taking measures before an actual breakdown occurs thanks to a number of analyses. Predictive Maintenance enables 25-30% decrease on maintenance efforts, 35-45% on breakdowns and 20- 25% increase on investment returns. Thanks to predictive maintenance, part replacements can be planned for downtime. This decreases total costs. Workforce, tools and spare parts needed for maintenance are made available for planned stop. Actual data reflecting actual mechanical conditions from equipment can be used. Maintenance schedules are performed and updated in the light of actual data. Predictive maintenance prevents most unplanned breakdowns and their negative influence against other systems. It also offers post-repair actual data. Thanks to this method, significant breakdowns can be minimized. Maintenance based downtime of the machine can be minimized. Unnecessary stops for the machines that are in good condition are also prevented. So, this is a time and cost effective method. Today, machine components need to be replaced rather than repairs. To determine machine components that need to be replaces, we have to perform periodical measurements with appropriate tools and we have to have experiences maintenance workers who can analyze the results of those measurements. Predictive Maintenance enables the plants to decrease failure based unplanned stops, wastage in production, repair costs and work related accidents and to decrease operational efficiency, work safety, quality of products and customer satisfaction. [Göçülü, 2015]. Most plants perform in-house tests like vibration analysis, thermographic analysis to detect flaws by their own maintenance personnel and tools. But for oil analysis, oil samples are sent to oil vendors since this method requires more sophisticated equipment, experience and investment. [Göçülü, 2015]. There are 3 main methods for performing predictive maintenance approach. Please see Figure 3. Vibration Measurement and Analysis: Vibration analysis methods is the most used predictive maintenance method that gives immediate result. Vibrations are converted to electrical signals by a receiver and directed to a device that processed electrical signals. Those data are sent to a computer and analyzed by special software. Results give extensive information about machines. Vibration analysis is a widely used predictive maintenance method since it offers a comprehensive problem solving structure. (Please see Table 1). Failure detection by vibration analysis can be used on all kinds of machines that have rotating parts. But detecting a failure with this method requires a comprehensive knowledge based on long-continued analysis. Detecting root causes may be hard if there are chronic problems stemming from manufacture, design or montage of the machine. Therefore, analysis should be done by comparing different graphics based on results of measurements from different measurements that are performed with specified intervals. [Yücel, 2009]. Today, vibration measurements and analysis can perform many functions thanks to advance technology and support from various devices, computers and software. Those devices are measuring vibration amplitude, vibration frequency and phase difference in order to perform vibration analysis. Vibration frequency is one of the most important data when we evaluate vibration problems of a machine. Vibration frequency is measured by devices that are placed at specific parts of a machine (like clutch, bearing, shaft etc.) and they measure in terms of Hz (Hertz) and revolutions per minute. Each component of machines creates vibrations that are special for them. When there are defects in machines, a special vibration frequency is created. By evaluating those frequencies, nature and location of defects are detected. This frequency analysis is also called spectral analysis. Spectral analysis evaluates amplitudes according to vibration frequencies of machines and finds root causes and how the breakdown has evolved from its beginning and they are eliminated at the right time. Latest technology enables us to use remote access and access to remote areas. Developments in data transfer technologies like data flows are also used in this area. Bluetooth communication is now being used for especially vibration measurements. Vibration devices that can continuously send data to control center (online situation monitoring) by Bluetooth communication are placed on machines and they include simple vibration meters and communication systems. Data gathered from those devices are transferred to machine control systems or mobile phones or computers of maintenance workers if they are within Bluetooth range. Measurements results can be displayed as spectrum graphics or wave form graphs. The aim is to view all measurements for a specific measurement point for a long period of time on the same graph and immediately realize differences in order to evaluate vibrations measurements. This can be done for both spectrum and wave form graphs. Another evaluation method is tendency graphs which enable vibrations level changes of failure frequencies according to time to be viewed. Tendency graphs can show seriousness of developing breakdowns. Increase in vibration amplitude gives a clear sign of the seriousness of failures. [Hancı, 2009]. Vibration analysis are used to detect problems like electric problems, gear and bearing damages, axial maladjustment, mechanical looseness Figure 5 shows examples of measurement points for vibration analysis [Kalyoncu, 2004]. Machine parts like machine frame screws get loose in time and cause vulnerabilities. And on spectrum vibration graph, frequencies on multiple stratifications of shaft turning cycles can be viewed. When this type of a problem is viewed, it may be a symptom of looseness and necessary actions are taken in order to prevent looseness based problems [Karadayı, 2011]. Vibration Analysis is a culture that requires continuous technical analysis. No signs simply mean a specific failure. Interactions between data should be analyzed in a cause-effect relationship. [Köse, 2003] Comparative analysis of graphs from periodical measurements enables to recognize increasing frequencies as a result of a developing breakdown. This approach can meet the needs of the Maintenance Engineer. So vibration analysis is facilitated by tendency tracking methods in the scope of predictive maintenance. [Köse, 2003] The equipment used for vibration analysis consist of a sensor converting to electrical signals, a device that can receive and process this signal, and a data storage unit. In order to perform analysis, the device should have "Fast Fourier Transformation" feature on it. Failure in a machine creates periodical signs. These signs are split into its harmonics within itself to detail. Harmonic signal can be simplified as a sinusoid. This curve gives the period and amplitude of the movement. [Köse, 2003] The cause of the failure reveals itself on harmonics of machine speed frequency. This physical information is the basis of vibration analysis and detection of failures. [Köse, 2003] The value of vibration is determined according to displacement, speed and momentum of the vibration. Defects cannot be fixed with no reason so it reiterates the pattern in every turning of the shaft. So each pattern in every period should resemble the other one. If there is no iteration; the source of vibration may be the process or other machines nearby rather than engine revolutions [Köse, 2003]. Vibration Analysis is a culture that has to be applied at every plant. But it is not yet popular in Turkey since industrial applications are not covered practically in engineering courses. [Köse, 2003] Oil Analysis: Oil film decreases inside machine parts when oil quality decreases. As a result, various wears are seen in machine parts. Oil analysis measures oil quality and determines worn parts by inspection metal chips inside oil sample. Particle count can determine wear, friction and filter defects inside the machine. Amount of opaque particles and water within oil are determined in order to evaluate its chemical value. If it has started to lose its quality and a source for failure is determined (filter failure, torn seal, mixture of cooling agent into oil); oil is changes or a maintenance schedule is prepared for that machine. [Köse, 2005]. Defects are detected by analyzing viscosity, total base number, conditions of the oil and wearing rate in laboratory environment. Thanks to this method; problems like wearing, friction, filter failure, torn seal, mixture of cooling agent into oil may be detected easily. In addition to this, during regular inspections, a number of problems are prevented like using incorrect oil, water and dirt inflow to system (if seals and wipers are not working properly or worn), not purified or contaminated oil in sumps of machines that are under maintenance or new or under repair. [Denli, 2007]. Oil analysis may be the only solution to detect both wears and seriousness of special breakdowns and vibration analysis is not successful to recognize wears in sliding bearings that are lubricated by oil. When all analysis methods show the same problem; diagnosis and suggestions may rarely be incorrect. [Orhan, 2009]. Thanks to the results of engine oil analysis appropriate oil is used and expected life of the engine increases. Engine problems are detected before they grow into serious levels. Viscosity, base number and specifications of engine oil are analyzed. Oil samples are taken periodically from the engine, in order to detect possible flaws. When engine oil is being analyzed; oil viscosity, total base number and amount of oil-carbon deposit, oxidation, nitration, sulphur, antifreeze, water and fuel inside oil are measured. Hydraulic Oil Analysis enables early detection of possible flaws by taking samples from hydraulic systems. Oil viscosity, total acid value, amount of water and particles are measured during hydraulic oil analysis. Metal Wear Analysis: Atomic absorption and emission spectrometer are two different methods for metal wear analysis. Metal Wear Analysis (Atomic absorption method): Oil samples are burned at high temperatures in order to find energy absorption of metallic particles. Sample of all metals are processed by a calibrated tool. This method requires a long period of time but accuracy rate is very high. Metal Wear Analysis (Emission Spectrometer Method): Oil samples are burned at high temperatures in order to measure radiation levels of the device. It is possible to make measurements for 18 different wear particles. The time required for analysis is short. [Hancı, 2009]. Infrared thermography: Thermography may be defined as infrared thermal measurement method. It is used to monitor heated areas in various systems. Thermography is preferably used in various heating systems like power plants, bakeries and boilers. Changes in temperatures are detected with measuring devices. Electrical charge/resistance, harmonic effects, friction, inspection of electric motors, induction heating, coupling maladjustment, inadequate lubrication, valves and lines are easily checked with infrared thermography. Tendency analysis of those changes is performed. A threshold level of temperature which may prevent system's operability or harm the system may be detected at early stages. When temperature is close to the threshold level a maintenance schedule is prepared and incident is responded accordingly.[Feyzullahoğlu, 2001]. Infrared rays help us for early detection of most electrical and mechanical flaws. Their wavelengths are between 0,75 and 1000 microns. They are invisible but they can be felt as temperature. Today, these rays are displayed by the help of thermal cameras. They enable us to see conditions of machines, wires, insulating materials etc. Thermal cameras display coldest areas in black. When temperature goes up, those areas are displayed as close to white. Thermal cameras can only detect the temperature on the surface of objects. They cannot detect internal temperatures. If an object is at the same temperature with its periphery, it would not be displayed by thermal cameras. Thermal cameras are used to detect flaws in electrical and electronic circuit components or generally in iron steel industry since many years. But it is a costly method. Hand-held thermal cameras have become widespread in recent years. They detect temperature increase and display 2 dimensional graphs of failing components that may cause breakdowns. So an accurate maintenance schedule can be prepared for necessary replacements. Advantages of thermography method are as follows [Hancı, 2009]: • No need to contact to the surface. • Does not trigger any activity during control. • Temperature increases are displayed with graphs which make easy to interpret the situation. • Not affected by electromagnetic waves. • Can be used easily during uptime. • Gives precise information. • Does not require expertise to interpret its results (unlike vibration analysis). This figure shows looseness at one of insulator connectors inside an electric transformer. Many high frequency signals occur inside electronic systems and this depends on operating principles of devices. The signals may cause harmonics and therefore local temperature increases on cables and their connections. Flawed eccentric connection, damaged bearing or inadequate lubrication causes increase in friction and therefore in temperature. This figure shows the temperature increase in turbocharger stemming from excessive friction. Thermographic method is preferred over vibration analysis method by plants because it is performed from a safe distance and without contact. Electric motors are the backbone of industry. Electric motors are used in many areas from spindle drive to conveyor bell drum rotation and functioning of pumps to axial movement. Motors are used in all environments like water, sand, with chemicals at hot or cold environments. They are specially designed for their intended use and all have different surface temperatures. So, we can easily determine which temperatures are normal or abnormal when using thermograph thanks to past experiences and with comparing similar motors. Thermography enables to detect incidents like inadequate air flow, gear box failure, shaft problems and isolation flaw at rotor or stator part of a machine. Induction heating is used during some steel processes in a controlled way. Yet, it may be formed in an uncontrolled way if high alternative currents are very close to uninsulated metals like steel consoles. High AC transmitting cables may cause induction heating in steel support objects and channels. Radiation causes cable insulation to become hard and fragile. It may break off unexpectedly and cause short circuit. Ultrasonic Tests: Ultrasonic monitoring is a method of predictive maintenance approach which offers significant gains for the plant. By using this method, pressured air leaks can be found, functioning tests for steam traps can be performed, lubrication and damage controls for bearings can be done, thickness measurements of equipment and corrosion tests can be done.. Also accuracy of lubrication can also be measured. So inadequate or excessive lubrication problems can be spotted. [Karadayı and Yaman, 2014]. Another important factor that facilitates ultrasonic systems is that it can be used at high nosise levels, although it is measuring the voice. This is because of the fact that, voices are mostly within threshold of hearing and not ultrasonic. Ultrasonic measurements coronas of motors have crackling noise, bearings normally have slight scratch voice and boom voices when they have flaw. [Denli, 2007]. Air can be easily ionized at humid and foggy environments. If that air is ionized on surface of phase conductors of power lines, purple circles are seen and this is called "Corona" [Anonim-2017] Bearings are critical elements. If they break down; detecting, ordering the part and replacement may cause a downtime of hours or even days. Studies show that 30% of bearings fail because of incorrect installment or maladjustment of the shaft and as a result of this their lifetime decreases 2 to 5 times. 10% of bearings fail in 6 months after installment. [Hancı, 2009]. Machines make ultrasonic voices since races contact with ball bearings and rollers. Well-functioning and adequately lubricated bearings make a low level of rustle sound. If bearings are inadequately lubricated they make strong cracking or squeak voices. Over lubricated bearings make no noise. 3. CONCLUSION To avoid stoppages during production, reliability checks of mechanical systems are essential. Unplanned downtime causes important production losses. Today, as Industry 4.0 philosophy has penetrated into plants and need to take all kinds of actions to avoid unplanned stops is well-understood. Industry 4.0 or the 4th Industrial Revolution is an umbrella term that involves modern automation systems, data exchanges and production technologies. This revolution is a set of values consist of internet of things, web based services and cyber-physical systems. Maintenance involves all activities which enable mechanical systems to show desired and anticipated performance at all times. Performing maintenance according to today's technology is only possible with adopting systematic and methodologic understanding and Engineering perspective. Maintenance is performed in order to enable all elements and parts of the system to function at anticipated levels. Successful maintenance means to help all elements of the system to fully complete their economic and nominal lifetime. Predictive maintenance approach proactively ensures all elements and equipment of the system to be under control.