Predictive maintenance sensors and implementation

Posted on 10 Jan 2017 by Tim Brown

A solution overview from Dell.

To provide a blueprint for you to build your PdM deployment around, Dell has developed a flexible architecture centered on the Edge Gateway 5000® with partners including SAP and ifm for a complete solution.

This article follows on from the first in two articles in this series on Predictive Maintenance: 

  1. Achieving predictive maintenance (PdM)
  2. Six steps to using the IoT to deliver maintenance efficiency

1. Solution Overview

Device instrumentation is provided by ifm with a variety of ideal sensors to gather machine operation data. The sensor data is translated to Ethernet/IP traffic by the ifm IO-Link master block and then passed on to the Dell Edge Gateway 5000 which enables you to collect, analyze, relay, and act on the real-time data. Dell Edge Device Manager (EDM) allows you to deploy, group, and securely manage connected gateways regardless of their physical location.

The SAP edge stack including SAP Intelligent Edge, SAP Dynamic Edge Processing and SAP Device Management for IoT (by Telit) manages device connectivity, local storage and analytics, data stream analytics and synchronization with the backend system. This ensures that perishable data is acted on immediately by generating alerts, and also makes certain that only meaningful data is sent to the cloud to minimize consumption of expensive network bandwidth.

The ideal backend for your PdM deployment is the Dell PowerEdge R930 HANA appliance running on premise or in your cloud with the SAP Predictive Maintenance and Service (PdMS) application stack. This enables you to integrate structured and unstructured IT system data into your models, and run big data analytics to identify even more granular patterns to predict failures earlier. SAP HANA also provides integration into your Enterprise Asset Management system for reporting, generating alerts, and automating maintenance dispatching.

2. Edge Description

After choosing your PdM use case applications, identify which data sources are needed for implementation. Some existing data will only need to be integrated (data from existing MES and ERP systems and machines that simply need to be connected to the Internet), whereas other systems will need to be instrumented to capture data. In these cases, it may be necessary to add new devices, such as sensors that clamp over the wiring in an electrical panel or mount on a machine to measure vibration from a specific mechanism.

The key to successful implementation of a predictive maintenance program is utilizing the proper sensors to determine machine condition. Sensor selection starts with an understanding of a machine’s potential failure modes and the warning signs associated with these modes. Typical warning signs on equipment with rotating parts include unbalance, bearing damage, cavitation (pumps), increased machine vibration levels, increased temperature of machine components, loss or reduction of lubrication flow, and loss or reduction of cooling water flow.

Each warning sign can be monitored with the appropriate sensor technology. When new instrumentation is required here are a few sensors to consider including in your plan.

Vibration Sensors

Vibration sensors monitor the acceleration present during machine operation, and are the best starting point when developing a predictive maintenance strategy. Sensors should be mounted to a rigid portion of the machine (frame for example) and as close to the object being measured (bearing, shaft, etc.) to insure proper vibration signal transfer. Basic vibration sensors detect an overall increase in machine vibration indicating a potential machine issue. More sophisticated sensors use FFT (Fast Fourier Transform) signal processing to look at sensor signals in the frequency domain in order to detect issues with specific machine components like rolling element bearings or fan blades. Vibration sensors provide the best leading indicator information of machine condition.

ifm’s multiplex vibration monitor provides specific vibration analysis on components inside a machine such as rolling element bearings, rotational unbalances and gear drives. The system consists of a cabinet-mount diagnostic module and multiple compact accelerometers that are mounted permanently on the equipment. The system’s internal microprocessor tracks up to 20 different machine components. Programming of the monitor is straight forward using the “wizard” guided configuration software.

ifm’s UW0001 Vibration Starter Kit has everything you need to get started - image courtesy of Dell EMC.
ifm’s UW0001 Vibration Starter Kit has everything you need to get started – image courtesy of Dell EMC.


ifm’s compact TN series temperature sensor - image courtesy of Dell EMC
ifm’s compact TN series temperature sensor – image courtesy of Dell EMC

Temperature sensors monitor critical machine components or auxiliary systems to detect changes in machine condition. For direct measurement applications, RTD’s (Resistive Temperature Detectors) and thermocouples are used. For indirect measurement applications, non-contact infrared sensors are used. Typical applications are bearing temperature, lubrication temperature, and cooling water temperature.

ifm’s compact TN series temperature sensor combines evaluation electronics, digital display, sensing probe and process fitting, for a completely sealed and calibrated system ready to install in your application. A simple push button menu allows the sensor to be programmed to fit your application setpoints. These sensors are designed with a two color (red/green) integrated digital display indicating if process values are in an acceptable range or if limits have been exceeded.

Oil Particle

ifm’s LDP100 inline particle monitor - image courtesy of Dell EMC
ifm’s LDP100 inline particle monitor – image courtesy of Dell EMC

Oil particle sensors monitor the level of particle contamination present in lubrication systems. An increase in particle count can indicate potential wear issues in gear boxes. Oil particle sensors are inline devices that should be mounted after critical machine components in the lubrication loop.

ifm’s LDP100 inline particle monitor monitors the degree of cleanliness or the level of contamination in oils. The integrated data memory allows data recording over a longer period. The LCD display indicates the cleanliness level. The particle monitor operates according to the light extinction principle. The changes to light intensity of a laser beam caused by particles in the medium flowing through the measurement cell are detected by a photo detector.


ifm’s SV series vortex flow meter - image courtesy of Dell EMC
ifm’s SV series vortex flow meter – image courtesy of Dell EMC

Flow sensors monitor the flow rate of lubrication or cooling water flow. Reduction or loss of flow rate can lead to future machine issues. Flow sensors are mounted in the lubrication or cooling line, and the sensing principle used (thermal, magnetic inductive, ultrasonic, etc.) is dependent on the media type, flow rate, and line size required.

ifm’s SV series vortex flow meter is designed to simultaneously monitor flow rate and detect fluid temperature of circulating water in industrial applications. Encased in a compact robust housing for harsh environments, the meter utilizes the vortex flow principle. The vortex flow technology enables a simple sensor design that can be easily manufactured at a very low cost. The meter provides a high contrast TFT LCD digital display that can be customized to best suit the application.


Current sensors monitor the current draw of machine components. A typical application is monitoring the current draw of a motor. Increased current draw over time can indicate wear/issues with the motor. These sensors are clamped around the electrical cable feeding the motor.


Humidity or moisture sensors monitor the water content in hydraulic and lubrication oils. Excess moisture can lead to corrosion and other machine issues. These sensors are typically mounted in the lubrication or hydraulic tank.

ifm’s LDH100 sensor measures the relative moisture in the oil in the range of 0…100 % by means of a capacitive measuring element. Besides the relative moisture the sensor also provides the medium temperature as an analog signal.

Other Sensors

The sensor types discussed above are the most common devices used in machine predictive maintenance, but many other sensors are available for machine specific applications. A variety of position sensors (inductive, photoelectric, capacitive, and magnetic) are available for machine position or product detection. Other process sensors like pressure and level are also available.

2.1 Best Practices to consider

Start with Vibration Analysis: Vibration sensors are the best starting point for deploying a real time predictive maintenance program. Start with a critical, bottleneck machine, install vibration sensors to monitor the critical components of the machine, and begin collecting data to understand machine condition. Other sensors can be added later as the root cause of the critical failure modes are uncovered.

Unlock data sources already available through the machine: Machines already utilize a lot of sensors for control of the machine itself. Using the existing data from these sensors along with new sensor data (vibration sensors for example) will provide a lot of information regarding machine condition. Sensor data can be pulled directly from the PLC. If the sensors are IO-Link compatible, the sensor data can also be passed directly to an Edge Gateway.

IT OT convergence: Consider the relationships between organizations such as IT, operations, finance, and your project team. Is your IT and operational technology (OT) infrastructure converged, giving your operations equipment access to the core IT network and making it easier to link your systems together? It is always better to partner with both IT and OT organizations from the start.

Architect starting with what you have: Move from pilot to production quickly and cost-effectively by leveraging existing investments, without being forced to rip-and replace, and by integrating open, future-ready infrastructure solutions.

Wired vs. wireless sensors considerations: Another consideration is whether to use wired or wireless sensors to collect data from your various energy sources, a decision that is typically influenced by a number of factors. For example, are you working with a large manufacturing plant, distributed field assets, or an oil rig? Are the various energy consumers close together, spread out, indoors, or outdoors? Can you easily pull cables to sensors at major energy consumption points? Hard-wired connections provide reliable connectivity, but pulling cables can be cost prohibitive in both retrofits and new installations, or even impossible in some facilities. In such cases, wireless sensors can be attractive alternatives. Can you get a wired power source to each sensor, or do they need to be self-powered (e.g., battery, solar)? What locations are in relatively close proximity to sensors, and therefore are more suitable for data aggregation? These are important factors when deciding between wired and wireless sensor solutions.

Cloud vs. on premise considerations:

The backend server in this reference architecture may be setup in a cloud or on premise. Benefits of an on premise setup include ownership and control of IT resources and faster performance for local users, this should be balanced by the cost for resources to support the infrastructure; cooling, electricity, and physical security costs and unused capacity when equipment is purchased for peak demand. Setup in a cloud environment, on the other hand, offers predictable monthly expenses, ease in scaling up or down based on demand, maintenance and overheads are the responsibility of the service provider. This should be balanced with less control and access to systems as compared to on premise environment, and performance is reliant on quality of internet connections in customer locations.

Standard vs. extended algorithm access:

In many cases the standard algorithms included in the SAP Predictive Maintenance and Service application are sufficient but for situations where they are not the application can be extended. Three generally applicable data science algorithms are available within the application out-of-the-box: Anomaly detection, Principal Component Analysis and Weibull Remaining Useful Life. These are purposely meant to cover cases where little or no additional information beyond the sensor data itself is known. They assume that the majority of the time a machine is running normally and deviations from that may indicate an issue. If additional information is available using customized data science algorithms can provide even better results.

Security considerations:

Data security is priority number one. At a minimum, protect businesscritical and personal information using data encryption and tunneling. Intelligent gateways offer hardware-level security with a Trusted Platform Module (TPM), Secure Boot, and BIOS-level password and port lockdown for minimizing unauthorized access to the system. For more discussion about securing your energy management solution check out this Dell security white paper.

2.2 Accessing Sensor Data through IO-Link

Industry 4.0 is based on making our factories smarter, so our machines can gather valuable data, then communicate and utilize that data in real time. This communication enables enhanced production capabilities while being able to immediately access and analyze data that was not previously available. The foundation of Industry 4.0 is utilizing real time data from the plant floor. Once collected, this data can be analyzed, formulated and shared to make every aspect of Industry 4.0 a reality. Sensors on the machines collect vast amounts of data but, until now, the data has been essentially trapped, with no pathway to control or data collection systems. IO-Link can access this trapped data through existing wiring.

IO-Link is a new, manufacturer-independent point-topoint connection for sensors and actuators. It can be used to transmit sensor data digitally. Using an IO-Link master block, predictive maintenance sensor data is easily transferred to a wide variety of SAP solutions via Ethernet.


ifm’s LINERECORDER Agent software allows for simple mapping of sensor data to SAP PCo or HANA via the SAP IoT adapter.


ifm’s LINERECORDER SMART OBSERVER software is used for condition monitoring of systems. Machine condition is displayed locally, and sensor data can be stored in the Edge Gateway for further pre-processing.

3.0 Fog Description

Dell Edge Gateways connect your field sensors, devices and manufacturing systems to the internet. Ruggedized, fan-less, with industrial-grade form factors and DIN-rail or wall mounts, our gateways are expandable and purpose-built for long life and 24/7 use, reliable at -30°C to 70°C. The Dell Edge Gateway 5000 enables you to collect, analyze, relay, and act on real-time data from machine sensors and generate accurate, dynamic predictions. Perform analytics on aggregate data at the edge, and then send only summarized results over the network. Decentralized data analytics also allows local systems to maintain normal operations when the connection to the cloud is temporarily lost.

  • Connect your systems, sensors and devices securely to the internet
  • Accept diverse I/O — wired or wireless—natively, and integrate industry protocols, like BACnet, Modbus, CANbus, ZigBee…
  • Aggregate, normalize, analyze and send only prime data to the back-end for storage and trend analysis
  • Supports Trusted Platform Module (TPM) v1.2 to securely store artifacts (including passwords, certificates, or encryption keys) used to authenticate the gateway.
  • Exchange data with the cloud, data center, other devices — providing analytics in the “fog”
  • React in real time, even when the network is down or latent
  • Offer hardware-level security and BIOS lockdown of I/O ports, with intrusion alerts on the IP65 enclosure
  • For environments with poor connectivity or limited network capacity, gateways can process, store, and forward filtered data, as well as provide offline capability to maintain operations.

Edge Device Manager (EDM) for Dell Edge Gateways allows managers to deploy, group, and securely manage connected gateways regardless of their physical location. EDM provides the quickest path to productivity with no management software to install on premise or maintain and automatic access to the latest functionality from the cloud. The powerful push notification framework coupled with OS and application imaging functionality allows EDM to deploy and manage OS images, updates, applications to devices seamlessly.

With EDM you get:

  • GUI-driven configurations and exceptions handling
  • Simplified device registration and grouping
  • Advanced insights into your Edge Gateway and embedded PCs
  • All from a browser

Key considerations:

  • When thinking about data sources, consider the nature of your organization’s environment. In the case of equipment located outdoors or in non-temperature controlled locations, the sensors and gateways you install must be able to natively withstand extreme temperatures in the -30°C to 70°C (-22°F to 158°F) range.
  • Consider selecting gateways with more computing power than you initially need to support new applications you’ll likely find in the future. Investing in computing headroom now will be much less expensive than paying for a truck roll to upgrade a gateway’s capacity at a later date.

SAP IoT Intelligent Edge provides a set of software components that can be installed on the Dell Edge Gateway 5000 to manage device connectivity and management, local storage and analytics, data stream analytics and synchronization with the backend system.

The following bundle of products are recommended for collecting, storing and processing data at the edge including:

SAP Plant Connectivity (PCo) – connects to many databases and sensor protocols on the shop floor such as:

  • OLEDB and ODBC
  • Proficy, OSIsoft PI and IP21 Historians
  • Citect SCADA systems
  • File monitors and web sockets

Streaming Lite – provides the ability to implement complex event processing on the edge including:

  • Consolidating input from multiple sensors or sources
  • Aggregating, enriching and filtering raw data
  • Generating actions or alerts on the edge without requiring a round trip to a central server
  • Streaming value added event data to a central Smart Data Streaming server for additional event processing and recording

SQL Anywhere – provides a small footprint embedded database gathers and stores sensor data from devices.

SAP Partner Edge Products

  • Dynamic Edge Processing – brings critical business data from ERP and other business suite applications from the core to the edge for real-time decisions based on a fusion of business and sensor data even if connectivity to the core is not currently available.
  • HANA IoT Connector by OSIsoft – provides a direct connection between OSIsoft Pi historians and SAP backend Predictive Maintenance and Service applications
    • Accesses the rich data sets generated by control and automation systems and smart devices
    • Merges operations data from OSIsoft with business data from SAP and third-parties o Combines time-series data with transactional business data
    • Enables enhanced business analytics and delivers applicable business insights o Supports most national and industry best-practice doctrines for security
  • SAP Device Management for IoT by Telit –designed to easily capture, process and forward data from production machines to the device management platform or another SAP system. Telit deviceWISE supports all popular PLCs from Siemens, Mitsubishi, Rockwell, Omron and most widely used production equipment, and is compatible with virtually any database, message queuing and application server system available, including IBM, SAP, ORACLE and Microsoft.

SAP HANA and Predictive Maintenance and Service Application (PdMS) combines both operational and IT data to perform analysis on the combined data set. Although some data processing can be accomplished at the edge, for more advanced analytics data analysis needs to be done at the core where more storage and compute power are available. SAP Predictive Maintenance and Service is an application that runs on top of SAP HANA and the following components are required from SAP:

  • SAP Predictive Maintenance and Service, On-Premise Edition
  • SAP HANA Platform or Enterprise Edition

These optional components may be useful in many situations to get more value from your Predictive Maintenance installation:

  • SAP IQ or OSIsoft Pi – either of these databases can serve as warm data stores to persist sensor data not currently needed within SAP HANA. This can help reduce the total cost of ownership as well as increase performance for large amounts of data
  • R in SAP HANA studio – is a method for developing customized data science algorithms for individualized analysis.