Intelligent Condition Monitoring for Smart Factories
Linking condition parameters and process parameters on rotating systems and integrating them into the control process
The target and vision of the German federal government to achieve full networking of pre-sales, production and after-sales must, for complex products, necessarily start with the production process itself. At the operative level of the production plant, linking machine condition parameters and process parameters, monitoring these continuously and preparing meaningful conclusions are decisive for product quality and plant efficiency. One component of this controlmcircuit is an intelligent Condition Monitoring system that offers the flexibility required to accommodate a wide range of production plants and features the interfaces needed for smooth information exchange between the plant control system, process visualization unit and operator. The functioning principles of a monitoring system that meets these requirements are described below and illustrated in numerous practical examples.
1 Industry 4.0 and the monitoring strategy
The description by the German Federal Ministry of Education and Research (BMBF) of the Industry 4.0 project und intelligent monitoring states: “…..Along with increased automation in industry, the development of intelligent monitoring and autonomous decision-making processes is particularly important in order to be able to steer and optimize both companies and entire value-adding networks in almost real time.…..” “... On the basis of detailed production data recorded online combined with intelligent visualization, control systems should be able to support decision-makers in planning and controlling production to considerably improve corporate goals, such as delivery reliability….”1)
2 Practical requirements for an intelligent Condition Monitoring system
Implementing the conceptual requirement for intelligent Condition Monitoring in a practical system represents a certain challenge. Apart from purely technical demands, economic aspects must also be considered since the investment initially does not yield monetary benefits from the point of view of the business administrator. The following paper illustrates the theoretical requirements with descriptions of practical solutions.
2.1 Intelligent monitoring processes
To be able to collect reproducible and plausible monitoring data, control parameters must be synchronously correlated with the measurement data and flexibly classified into certain operating states. Examples are varying load conditions and speed changes controlled by frequency converters in highly dynamic, controlled motion sequences. Without information on these processes and classification of measurement data into different load or speed classes, for example, reliable interpretation of the monitoring data is only conditionally possible.
2.2 In almost real time
Especially in failure-critical plants, all process-related condition parameters should be recorded without delay and flow into the control system in almost real time. Classical monitoring systems with multiplexer scanning are not suitable here. Rather, synchronously measuring online systems with a time-based correlation of all incoming parameters and a field bus interface for rapid data exchange are called for.
2.3 Detailed data recorded online
Certain condition parameters that are subject to multiple influences should be recorded online with a high degree of detail. Frequency-selective monitoring of vibration-related condition variables provides a markedly higher information content. Depending on the rotational kinematics, high scanning rates may be required. This, paired with the required synchronicity and the phase information it gives rise to, results in high demands on the monitoring system in terms of computing power and data processing capabilities.
The following figures 1 and 2 illustrate how information is lost when discrete time windows are recorded (Fig. 1) compared to a continuous time window (Fig. 2) with overlapping and continuous data logging. Permanent monitoring, which necessarily generates vast amounts of data, is used on turbo rotors for machine protection monitoring, but is also useful for the highly dynamic motion sequences mentioned in section 2.1, such as on transmission test benches.
To keep redundant information to a minimum, the data are subjected to a testing procedure in which they must meet certain criteria before being stored. These include having been collected during certain operating states (e.g. at a defined speed and load) or exceeding the previously collected value by a certain defined differential. Fig. 3 shows a trend plot with a minimum storage frequency of one measurement/second.
2.4 Intelligent visualization
The analytical evaluation of the monitoring data with variable reporting tools requires a server-capable diagnosis software. In parallel to this offline platform, a real-time visualization software with trending and interactive functions (e.g. alarm acknowledgement) is required that is supported by mobile terminals.
Fig. 4 shows multiple frequency spectra (FFT) in a waterfall diagram. This offline display is generally used to study special frequency-selective phenomena (e.g. structural resonance) and is only called up when required. Real-time monitoring usually relies on trend monitoring of broadband characteristic values, as shown in Fig. 5.
3 Challenges in the field
3.1 Intermittent movements with variable load conditions
Conveyors are an important element of a production chain and malfunctions can lead to costly production downtime.
Automotive/body construction: assembly lines, conveyor belt lifts, lift tables
Short movement cycles with a variety of positions and load conditions are a challenge to obtaining reproducible and classified monitoring data. The high complexity of such monitoring tasks is illustrated here using the cable winch drive of a container crane as an example.
Fig. 6 shows the size of container crane drive units. The cable winch drives shown at the bottom right are located in the control room.
The goal of the operator is to identify anomalies in the drive components early on and monitor these components in terms of load and activation. The monitoring tasks in detail are:
Torque load of the drive shaft of the cable spool
Force applied by the service brake
Optimization of the actuation of the motor/service brake
Condition parameters of the gearbox/roller bearings (specific frequencies/housing temperature)
Cloud-based remote monitoring
This application places requirements on the monitoring system that conventional systems can hardly meet:
Actuation (start, load, direction of rotation):
Flexible trigger processing
Motion dynamics (startup/stop)
Extremely low-speed machine
High scanning rate
Large data volume
Fig. 7 shows a plot of the drive speed of the cable winch motor and the torque load of the drive shaft against time. Ideally, the main brake should release immediately when the motor is started up to prevent undue loading of the drive shaft. However, a zoomed analysis of the trend shown above reveals that there is a small time offset between starting up of the motor and releasing of the brake, which requires adjustment of the actuation
3.2 Highly dynamic motion sequences
Rapid and highly dynamic motion sequences require continuous, gap-free monitoring and a protection function provided by the monitoring system that prevents catastrophic breakdowns. The following application describes the monitoring of drive components of vehicle test benches. The test pieces are random samples taken from the production line of a large automaker. Any interruptions would therefore have an impact on the entire engine manufacturing process.
The performance specifications in detail:
Impermissibly high vibrations that occur during a test run, e.g. due to structural resonances, must be reported immediately to ensure that the system can be switched off promptly.
The bearing condition of the drive machines should be monitored and any damage reported immediately
All measured values should be recorded in parallel and processes synchronously (phase-related)
Management and selection of different testing regimens with highly dynamic test runs
The special demands that such a monitoring system must meet far exceed the performance capabilities of classical online systems. These are:
Highly dynamic test runs, which generally emulate certain road profiles, result in continuously changing load and speed conditions
The structures consist of different drive machines, depending on the test piece. Thus the system must be capable of handling different measurement location configurations.
A protection mechanism that switches off the system if vibrations become excessively high requires a permanent comparison of limit values and status signals, which should occur without delay at the field bus level.
To be able to accurately identify the causes of impermissibly high vibration behavior and the condition of the drive components, sharp frequency-selective measurement data is required. Considering that the operating behavior comes from a variable speed machine, this is only possible using special algorithms (order tracking).
The monitoring plan in Fig. 8 shows the positioning of the vibration sensors on the roller bearings and drives.
4 VIBGUARD –Online CMD of the next generation
The newly developed VIBGUARD system delivers a real-time continuous and detailed condition overview of the relevant machine components with a focus on rotating parts. By synchronously and continuously gathering data from all analog channels, VIBGUARD delivers time-correlated relationships between the machine components and process parameters. For example, it can provide information on how the smoothly running characteristics of a shaft change with temperature.
Overview of the most important VIBGUARD features and their applications:
20 synchronous analog channels with continuous data recording
Frequency-selective vibration monitoring of rotating components
Roller bearing/journal bearing monitoring
x/y/z monitoring of static and moving machine components
Process parameters (e.g. pressure, temperature, speed)
Bidirectional field bus communication
Exchange of conditions and control variables via the Modbus protocol
Classification into 6 operating states
Flexible limit value adjustment to different load states/operating conditions
Selective data analysis according to operating states
Data recorder with “event recording“
Recording of a high-resolution time signal before an event (e.g. damage occurrence)
Root cause analysis
Data analysis/reporting/real-time visualization
Powerful multi-user software for data analysis and report creation
Trend/limit value display in real-time via Modbus on PC/mobile terminals
It appears that, the VIBGUARD system is an important component in integrating a modern monitoring strategy into the Industry 4.0 environment.
Fig. 9 shows two variants of the VIBGUARD online system: a board for stationary online applications and an industrial case with an integrated PC for mobile trouble shooting.
5 Big data and pertinent information
Total networking to create a smart factory throws up the topic of “big data”. Without intelligent data reduction, a monitoring system of the performance class described above alone generates information in the gigabyte range in a matter of a few weeks. So-called expert systems try to deliver precise information from an abundance of data using condition patterns and a statistical approach. These expert systems are not universally applicable to the condition monitoring of production plants and therefore only partially suitable. The likelihood of reliably predicting the failure of a certain component is still a matter of chance. A significant obstacle to creating a smart factory is networking, but handling, correlating, and correctly and automatically interpreting the data volume are of concern as well. Maintenance specialists can only make the best use of the wear reserves of machine components if they obtain precise and compact information on the plant. System manufacturers and research institutes have been working on a universal and practicable expert system for some time. However, a solution is not yet within sight, and control circuits will continue to rely on the human factor for some time to come.
1) Bundesministerium für Bildung und Forschung (2014), Industrie 4.0-Innovationen für die Produktion von morgen (German Federal Ministry of Education and Research (2014), Industry 4.0 Innovations
Published in: „Der Instandhaltungs-Berater“, TÜV Media GmbH, Cologne (2015).