 Hello everyone, this is Ernesto Manuel Cantone with ST Micro-Electronics, part of the Mems and Sensors Marketing Team here in the US. One of ST's strategic objectives is to increase the footprint of our sensors in the industrial market and in this presentation we'll discuss ST products and tools to achieve it. Today we are going to talk about smart industry, or as you can find it often referred, industry4.0. The number four refers to a fourth industrial revolution after the first introduction of mechanization of manual work in the 18th century through the use of water and steam power. The second, the emergence of mass production and assembly line using electricity in the 19th century. The third, in the late 90s, early 2000, refers to the adoption of electronics, computers and automation, and the opportunity of using IT to further automate production lines. Industry4.0 refers to the opportunity of using IoT and cloud to automate complex tasks and building smart and autonomous systems fueled by data and machine learning. The characteristic of this smart system is the capability to sense data, exchange data between the system parts via some means of communication and process data even in a distributed manner. What really enables Industry4.0 today is the miniaturization of sensors and the widespread availability of microcontrollers and connectivity integrated circuits. You can already find references to Industry5.0 as the cooperation between men and machines, but let's consider that an application enabled by Industry4.0 for now. The smart industry effort unlocks multiple business opportunities, producing more efficiently and in more environmentally friendly manner, responding to demand with more flexibility and with more customization, enabling better use of manufacturing and supply chain data, and overall providing a better and safer human experience. An immediate opportunity we want to address today is to optimize the maintenance of equipment, taking advantage of the data available from the equipment itself. When we talk about equipment, we refer to various assets, rotating equipment and related parts, HVAC systems, elevators and escalators, just to name a few. Maintenance is a set of actions performed by a dedicated professional to keep a machine working properly. One common approach is to schedule a maintenance quarterly, pretty much what we do with the oil change in our cars. It is simple to plan, but maintenance may happen too late or too early and may not have been necessary. A second option is to condition the maintenance to an inspection and monitor of certain parameters. Here also we are tied to a schedule and a professional inspects the asset using bulky and likely expensive tools and needs to analyze the data before taking for directions. Think about our HVAC system check. Sometimes a repair is needed, sometimes not, and there is no guarantee that something could break between two checkpoints. With the availability of affordable sensor nodes systems, the industry for the zero concept comes to life. We can now instrument every asset with the sensors and communications to monitor it continuously. Sensors are there to detect radiation from normal behavior and the system will generate warnings and alarms to promptly schedule an intervention. This approach requires a complex overall system, but offers many benefits. Maximize running time of the equipment and target the maintenance only when needed. Detecting failures at the early stages often results in a smaller and cheaper repairs and lead to longer lifetime expectancy of the asset. Repairs can be more efficient. For example, thanks to the analysis of the data collected, the service technician will arrive on site with the right spare part already preordered. What does a complete architecture for a predictive maintenance application look like? We mentioned the asset in the factory floor and the sensor nodes, and those are connected directly or via gateway to an enterprise cloud, local or remote. This implements a system with distributed computing capabilities, with the opportunity of processing data from the sensor nodes themselves to the local gateway to the cloud itself. In this slide, we define as edge both the smart sensor nodes and the gateways, as both can process data locally addressing latency and privacy concerns. The data can be then stored and analyzed, and the integration with enterprise resource planning, manufacturing execution systems and data warehouse applications allows to close the loop with the physical world and the cyber-physical systems we mentioned earlier, and materialize the business proposition of the smart industry. In our example, this closed loop is represented with scheduling a maintenance call and sending a technician to the factory floor. This is the installation point of a failure curve, plotting a machine condition over time. For a motor-installed factory, when conditions start to change, one of the first symptoms is usually a change in the ultrasound signature. Being able to monitor in that spectrum allows us to spot a change very early, months from the actual failure. The next symptom is vibration, and the availability of dedicated eye bandwidth sensors also leads to predicting a possible failure way before it's time. Power refers to being able to monitor the voltage and current parameters of the motors, and they also can signal abnormal behavior, but only weeks from the failure itself. After this, one can start to hear audible noise and eventually sense increased heat and detect smoke, but probably too late to avoid a serious damage. How can ST technology help? ST Microelectronics is a market leader in MEMS sensors technologies, with more than 17 billion spars shipped as of today. We are expanding our leadership in consumer MEMS to the industrial segment, leveraging our proven manufacturing capabilities with the introduction of the industrial sensors portfolio, all tested to higher standards, extended temperature ranges, and with a sheltering gravity of 10 years guaranteed. You can see here the industrial version of our leading success is IMU, ISM's 3200DHCX, accelerometer gyroscope combo, best in class in regards to power consumption and noise density. This device supports the traditional ST embedded capabilities with smart FIFO up to 9 kilobyte, and the X in the part number denotes the availability of a programmable finite state machine and machine learning core. Unique features that enables local data processing in the sensor ASIC for best possible speed of operation and power consumption on the overall sensor node. We have then the industrial versions of our magnetometer IAS-2MDC and E-Compass ISMs-330DHC and the standalone accelerometers IAS-2DH LPC, allowing lower power consumption while still monitoring an event. Let's focus then on the inertial sensors designed with a specific task in mind. IAS-3DHHC, our first high-precision n-clinometer with low noise and high stability over temperature and time, and IAS-3DWB, the ultra-wide band thermometer that we later discuss in detail. We mentioned the sound and ultrasound, and here we present our MEMS microphones. First, the digital ones, the industrial-grade top port IMP-34DT05 for sound acquisition up to 24 kHz and best-in-class total harmonic distortion. And our latest portfolio addition, the bottom port MP-23DB01HP, best-in-class signal-to-noise ratio at 64 dB and acoustic overload point at 134 dB SPL. For ultrasound detection, we offer MP-23 ABS-1, our analog bottom port microphone capable of acquisition up to 80 kHz. Environmental sensors can also play a role in predictive maintenance, and in this category we have pressure with LPS-22HH, which is dust-resistant fully molded package and LPS-27HHW with a metal lead and gel ceiling to allow water-resistant grading and resiliency versus harsh environment. Both are high-performance and accurate and robust to thermal and external stress. Our temperature sensor offering then is composed of both analog and digital output devices with industry standard packages and features, and of high-precision miniaturized devices like STTS-751 and STTS-22H, the latter with the same 10-year longevity commitments of our industrial grade sensors. If in need of temperature and relative humidity combo, we offer the low power consumption as both sides, it's yes 221. Let's try to match the curve we discussed previously with our sensor technologies. On the X-axis we have the bandwidth and we can position our inertial and acoustic sensors compared to various use cases. Different failure modes have different frequencies signatures. For example, unbalances and misalignments can be identified using a standard accelerometer, the most limited in bandwidth here, but the best in thermal power consumption. Bearings and cavitation failure modes would require the capability of sensing vibration at higher frequencies, and here we cover the case with our IMU or even better with our latest vibration sensor. Issues in gear mesh and bearings are in general creating the highest frequency failure signatures, hence the use of standard and ultrasound microphones. All these time domain signals can be converted to peaks in frequency domains running a fast Fourier transforms on the host microcontroller, and the results can then be transferred with digital wireless communication methods with limited bandwidth. Let's take a mental note on these as we touch back later on the topic. Let's now focus on the newly introduced IISDWB vibration sensor. How does a MEMS vibration sensor compare versus an industry standard piezo solution? Let's try to list the advantages of a MEMS approach in designing a vibration sensor. MEMS will be smaller in size, weight, and in power consumption. The sensor will then have a direct digital output simplifying the interface with digital microcontroller host. Robustness to shock and fast recovery after shock is another pro, and the opportunity of having embedded functionalities is a big plus and everything at a more affordable price. A piezoelectric vibration sensor will still offer an higher bandwidth and lower noise, but at a dramatically higher price point. A thousand and something versus sub ten dollars to give you an idea. And the MEMS vibration sensor approach target is not to replace a piezo one-to-one in their existing use cases, but rather to enable higher volume, lower cost monitoring of industrial machines with high quality sensor data, and possibly using wireless interface. So let's introduce our IIS 3DWB ultra wide bandwidth low noise three-axis digital vibration sensor. The full scale of the sensor is selectable up to an optimal 16g, while most of the motor application will use 4g to 8g. The output data rate is fixed at 26.6 kilohertz, and this enables a wide measurement bandwidth of five kilohertz minimum. SPI interface to the MC host is the recommended communication bus allowing to fully utilize the vibration sensor measurement bandwidth for all three measurement axes. The communication to the host includes two fully programmable interrupt pins and in line with other ST MEMS sensors, the device supports a big three-kilobyte FIFO, programmable filters, temperature sensor, and self-test functions. The noise density of the component is a best in class 75 micro g over square hertz, and it can be further down reduced to 60 in single-axis mode. The current consumption of the sensor is 1.1 milliamp, and while not comparable with the standard accelerometer, it's still suitable for battery powered applications. This is an industrial grade sensor, so it operates at the extended minus 40 to 105 degrees Celsius range. Operating voltage is from 2.1 to 3.6 volt, and the package measures 2.5 by 3 millimeter, with a pinout compatible to our 6-axis IMUs. What are the KPIs for a sensor designed for vibration monitoring? First, wide and flat measurement bandwidth. Check on the flat frequency response with sharp autobend roll-off and noilizing. Low noise levels and finally stable thermal behavior. Let's see how IIS-3WB checks all these boxes. First and foremost, let's take a look at the frequency response of an analog sensor with its evident resident frequency peak. Filtering is required to obtain unusable response, and some of the available devices are single-axis, so when in need of multiple-axis sensing, the bomb needs to be instantiated for each. The direct digital output of the IIS-3WB with no external components is flat and ready to use, and already includes all the three axes. When we look at the frequency response graph of IIS-3WB, we find it flat up to the 6.3 kHz cutoff frequencies at minus 3 dB point. Thanks to this flat response, no calibration, nor additional filtering in the OSM-SU is needed in the final product. Another important part of the frequency response characteristic is how steep it falls after 6.3 kHz, with a slope greater than 90 dB per decade, resulting in an attenuation higher than 70 dB for frequencies above the 26.6 kHz ODR. Last and very important as well, there is an high attenuation of more than 50 dB achieved for high frequency signals folding potentially back inside the signal bandwidth. Let's take a look at the noise density. The IIS-3WB filtering chain consists of four blocks. We have the mechanical sensing element, the analog front end and high speed ADC, the first low pass filter, LTF1, and lastly the optional composite filter which allows additional filtering or digital functions. As a result of the sensor characteristic and the internal digital processing, we are reaching excellent values for the noise densities, 75 micro G per square hertz for X and Y axis in three axis mode of operation. In single axis mode, thanks to a reconfiguration about the internal sampling and filtering operates, the achievable sensor resolution of the active axis significantly improves with the noise density going down to 60 micro G per square hertz. Such noise densities are perfect for the resolutions required in the condition monitoring application the vibration sensor was designed for. The last important key performance indicator of the vibration sensor is its sensitivity drift over temperature. The drift stays within plus and minus 2% tolerance from the ideal linear curve over the full temperature range from minus 40 to plus one of five degrees C. This is a benefit for the condition monitoring application since no additional calibration or sensitivity compensation is required. 0G offset drift, a parameter important for traditional accelerometers application, is not important for a vibration sensor since DC operation is not of interest. Let's take a look at how we can position IIS 3WB versus some of its competitors in terms of performance and price. On the top right we have an high bandwidth low noise single axis analog thermometer with higher performances but at a much higher price point even without considering the more complicated BOM and signal processing required to end the signal. On the bottom left competitor B is offering both the three axis analog and digital devices with medium bandwidth slightly higher than ST IMUs and overall not great noise performances and with a price that reflects the performance. ST positions IIS 3WB at higher performance for its little price premium what we consider as withspot for the application of interest. Our customers can evaluate the IIS 3WB as a standalone sensor with our professional MEMS tool and the Unico GUI available for Windows, macOS and Linux. The evaluation kit available for IIS 3WB is composed by a standard D24 interposer connected to the board hosting the sensor with a flex cable allowing optimal placing of the sensor itself as close as possible to the vibration source. Alternatively the vibrometer is part of the recently launched STWIN sensor style wireless industrial node as part of a complete system. Let's dig into the STWIN characteristics and discuss its applications in the next slides. Here's a first snapshot of the kit. It includes a combination of industrial sensors for vibration analysis, sound emissions up to 80 kHz and environmental monitoring. The system MCU is an STM32 L4 Plus and the out-of-the-box connectivity is Bluetooth low energy. An optional Wi-Fi module is also available. The kit includes also an ST-Link V3 Mini to program interface with and debug the application. A plastic case with mounting holes and a rechargeable 480 mAh LiPo battery. Looking at the core system boards we can quickly identify the main MCU, the various sensors and the expansion sockets including STMod Plus for connectivity modules already available in ST ecosystem. On the bottom of the board there is a slot for a micro SD card for local storage of sensor data in not connected application. Looking at the block diagram the STWIN is centered around STM32 L4 R9, ultra low power ARM Cortex-M4 clocked at 120 MHz, sporting 2 MB of flash and 640 KB of RAM. It's powered by STDC to DC battery charging and LDO devices and all the expansion sockets and connectors are protected by ST-ASD protections guaranteeing the design robustness. Soldered on the board as a standard wireless communication there is STBTLE-1S Bluetooth low energy for the tube module and NRS485 transceiver is available for wired industrial applications. There is a footprint also for the STSAVE A110 secure element to securely store credentials of the connected product and offload cryptographic services of the main MCU. From the sensor's perspective one can find your IIS3DWB vibration sensor, ISM330DHCX 6-axis IMU with machine learning core, the standalone lower power lower bandwidth accelerometer IIS2DH and the 3-axis magnetometer IIS2MDC. Several environmental sensors are also present and include temperature with STTS751, humidity with HTS221, atmospheric pressure with LPS22HH and the two MEMS microphones, MP23-ABS1 with analog output and ultrasound capability, IMP34DT05A with digital EDM interface output. From the firmware standpoint there are multiple options that we'll discuss in the next slides. Focusing on the sensors you can see that all the part numbers we discussed previously are all available here in a convenient sensor node systems. This enables the user to experiment with a combination of sensors and understand which ones are matching best the needs of his end application. In this diagram you can conveniently locate all the sensors and the connectivity module in the core system board. The schematics and layouts are available on ST.com so that ST-WIN can be used as a starting point in designing a custom device. Two expansion modules are also available on ST.com. First and foremost, a Wi-Fi module board to enable cloud-connected application supporting a pre-certified InVentec module supporting 802.11 BGN 2.4 GHz Wi-Fi connectivity. Second, a microphone array with four analog microphones to enable application with advanced audio algorithms. Let's discuss the available firmware packages. ST Software ST-WIN Kit 1 supports multiple examples exercising basic functionality on the ST-WIN. We'd like to bring your attention to the iSpeed Datalog application that enables simultaneous data collecting from all the sensors at IODR via USB or recording on SD card. This is key to interface ST-WIN with the standard sensors development tools and take full advantage of the machine learning core capability available in the ISM-300DHCX. You can find dedicated training material on machine learning core on ST6SASISIMUSE on ST.com. Today's focus is Predictive Maintenance and the Function Pack Industrial Predictive Maintenance 1 is the optimal starting point to prototype an end-to-end application. Like all ST's Function Packs, it's built on STM32Q that includes an hardware abstraction layer and middleware to support the various functions including dedicated algorithms for advanced time and frequency domain signal processing and analysis of IIS-WB up to five kilohertz. The package includes also pressure, relative humidity and temperature sensor monitoring and audio algorithms for acoustic emission up to 20 kilohertz and ultrasound emission analysis up to 80 kilohertz. The out-of-the-box experience of ST-WIN is this Function Pack in its BLE version while the full capability of an end-to-end application including a cloud part is enabled with the Wi-Fi version for which the Wi-Fi expansion kit is required. Let's take a look at the experience enabled by ST-WIN and the ST-BLE sensor app available on Google Play and iOS App Store. If not already programmed with this application, the ST-WIN can be easily programmed connecting the provided ST-Link with 3Mini and dragging the pre-compiled binary in the Function Pack package in the target flash exposes mass storage via USB. Alternatively, STM32Q programmer can be used and the firmware is available in source code as project for IAR, Kyle and STM32Q IDE. Once programmed and supplied via USB or with the provided battery, the ST-WIN will advertise and can be connected to the ST-BLE sensor app selecting it from the list. Once connected successfully, the first screen on the ST-BLE sensor app shows the real-time environmental sensor data values. In the second screen it is possible to configure and to display the FFT charts from the vibration measurements. A sub-menu is available to change the parameters of the FFT engine, in particular the vibration sensor configuration, number of FFT samples, etc. The next screen in the iPhone app is called plot data and it is designed to display the raw sensor data upon selection in a time graph. Finally, in the predicted maintenance screen, it's possible to check the real-time status of the RMS speed and peak amplitude acceleration of any of the three vibration sensor axes. This functionality is also available for the frequency domain through preconfigured thresholds at particular frequencies. Thresholds are predefined for good warning and alarm status via project source code adder file and modifiable by the user. The most interesting use of ST-WIN involves a cloud application hosted by ST and powered by AWS services. One, we need the Wi-Fi expansion board for direct connectivity rest between to the AWS IoT core service with the function pack industrial predictive maintenance one. ST allows any user logged with these myST.com credentials to connect up to five sensor nodes seamlessly and use them for up to six months. The data collected by the sensors are accessible to the user only and available to download for further analysis. This is a sandbox for our customers to play, free of charge and rapidly build a proof of concept with the ST development kits. Today we'll discuss how to connect ST-WIN but the dashboard is also compatible with IO-Link sensor nodes connected via an AWS IoT Greengrass Gateway built with the STM32 MP1 as explained in details in the solution for the team maintenance edge to cloud on ST.com. The data streamed from the sensor node to the cloud are humidity and temperature from HTS221, pressure from LPS22HH, vibration from the IIS3DWB, ultrasound emission from MP23 ABS1. Both eye bandwidth signals are pre-processed on the STM32 with a local FST before being sent to the cloud. Rather than explaining all the functions let me walk you through a live demo of the dashboard for which I have my ST-WIN programmed with the Wi-Fi application of a function pack industrial predictive maintenance one. The ST-WIN is then directly connected via USB to an USB port on my PC. Okay first thing let's see how to program the correct firmware in the sensor tile.box. I'm on the function pack predictive maintenance one folder let's go through projects ST-WIN demonstration Wi-Fi application AWS on the binary and let's drag and drop the binary into the ST-Link B3 that will program the flash of the target the ST-WIN STM32L4+. Okay we're done here I'll disconnect from the laptop the ST-Link and verify that I can connect to the board via terminal. Let's select the serial COM port 12 and wait for the board to reboot. There you go so once on the dashboard let's configure a new device. We already have one device created here in my instance of the dashboard and as you see it will expire in six months from now. So let's configure a new device and you can augment the device characteristic also with the information like a location portland in my case and with the latitude and longitude of the location. I'll show you why is that helpful and I create my object. Once I create the object the dashboard will provide a zip file containing certificate material that uniquely identified the ST-WIN node in these AWS systems. I will download this file and I can download this file only once upon creation and I will also copy the IoT endpoint which is the URL that the device is going to connect to. So let's close this. I have a text file in which I'm going to paste the URL I just noted and you can notice here the name of the device and I created a folder in which I'm going to copy all these certificate material. Okay so let's go back to the to the terminal. I'll insert here the wi-fi parameter. The device is connecting. The first thing that the device is going to ask is the URL for the endpoint. I'm going to cut and paste the endpoint and the device name. Second thing the program is going to ask me for the identity of the device. The identity is formed by three components the root certificate. So a certificate common to all the device and issued by a certificate authority. I'm going to paste it here. Then it's going to ask me and press enter. It's going to ask me for to the device certificate. A device that is unique to the to this particular node. I cut and paste and paste the device certificate and the private key that is used to sign during a TLS handshake and should never leave the sensor node in discussion. There you go. Once given all these parameters the device has an identity and will start to connect to the AWS cloud. Here's the connection happening and the first handshake with AWS and then I'm starting to send data and as you see the board is now online. Now if I go on the dashboard I can start monitoring immediately the data coming out of the board and I can also add the other reference board that I already had connected previously. You can see here the data coming live from the two boards one in Portland just created the other one as we would see in Boston. Data from the environmental sensor pressure humidity temperature peak data from the vibration analysis the rms speed the board is stationary now so that's why you see all zeros and the FFT transform of both the vibration analysis and of the ultrasound emission picked up from the analog microphone. If I go on the sensor map you will see that the previous board was provisioned in Boston in one of my colleagues house and now the sensor in Portland that we just provisioned is also online. You will notice that there is a status here warning I have no status yet for the Portland node I can and this is where the status come from it's basically a conditioned monitor over a certain parameter to set that on my new device I will go on the properties and I'll select for example temperature and select a normal range let's say 0 to 25 a warning range let's say 0 to 35 and an alert range 0 to 40 I'm saving and now if I go on the same asset conditioning monitor you will see that my ST wind just newly connected is also expressing a conditioned monitor value so I'm monitoring the temperature temperature here is 26 so I'm on warning state and if I go on the map now both devices in both locations I can see are in warning states. One function that is really powerful and that's the reason why I created an extra device earlier is the the data lake function I'm able to download all the data that historically has been sent to the dashboard I can select all the six months and I can select all the data and the data is available the day after that was streamed and that's why the device that I just created data is not available for download but I'm downloading the data from the device that was already created previously in Boston it will be outputted as a zip file and as you see there's no new board there's only the the older device with 0.5 21 I can check for example the environmental I can check today's date so this week and the output is a json file so it's basically is a text file indexed that can be used in an imported for example in excel and can be used for data analysis and to get inside out of the data themselves so in conclusion we showed how to connect devices to this sandbox environment how to stream data out of these devices store them in the data lake and download them to your computer for later analysis the st dashboard is a sandbox environment as mentioned previously and while the user owns the data she or he can't connect the stwin to its own aws account if a developer would like to build an aws application from code using free artist and its middleware I'd like to highlight that stwin is a free artist certified and listed in the aws device catalog and supports wi-fi integration and aws over the air update scheme the code is available upon being white listed and the request is available from the aws device catalog itself if the user chooses to use azure as cloud provider we support that with the stwin through the function pack cloud azure one the firmware supports microsoft azure iot central in a nutshell a set of pre-defined cloud applications that developers can use to jump start their development a user with a microsoft azure account can create an instance of a given application in his own account and start the cloud development from there as he created an iot plug and play capability model for stwin to be imported in a custom application here's a snapshot of the application that can be created in your own microsoft azure account and the dashboard visualization part you'll appreciate that this is not a sandbox environment nor an application built from scratch from services but rather an application running on a private account in a matter of minutes also in this case let me walk you through a live demo of the functionalities of the stwin running function pack cloud azure one connected to microsoft azure iot central in this case the stwin needs to be connected via the provided stlink to the ost pc via usb let's start with a clean browser i will paste the the iot central application and as you'll notice that i am logged in here as myself on the microsoft azure cloud and i'm about to instantiate the application on my own account i can choose between three placing plan i will go with the free one and the only thing i'm asked for is a phone number to be able to just and to basically confirm my identity so i'll confirm and create and the application is provisioned once provisioned i'm presented with this uh dashboard i can go to devices and create a new stwin device so new it's a not simulated device an actual device so let's create it and now it's created and i'm going to its property next thing we need to do we need to connect it and get an identification to uniquely identify the device in the azure cloud and here i have the id scope device id and the primary key that we use to provision i am connected via stlink to the usb on my pc so let's reset the board like we did earlier so we want to change the parameters so let's enter the ssd the password and the security type now the scope id i will not automatically configure manually and roll the device i will copy the device id and the primary key and that's all i need to identify the device in azure so the device is initializing and start to sending data up to the azure cloud and one when i'm going here i can check on the board status you'll see that the board is coming online it's coming online with the device model the firmware setting manufacturer etc on very powerful thing that i can do from this dashboard is to configure the sensors let's take a look for example here at the accelerometer on the six axis i'm showing here a 2g full scale if i go on commands i can change the full scale of the accelerometer for example to 8g let's run the command and going back to the board status you end the refreshing you will see that now the device is configured for an 8g full scale i have another tab here on telemetry where i can see the data streamed from the device i have the magnetometer gyroscope accelerometer and environmental sensor very powerful tool thanks to the function pack azure cloud i instantiated the cloud application on my own account and i'm now streaming data to an account so this can be a nice starting point for today's presentation we'd like to wrap up i like how ST is offering a complete ecosystem for developers to jump start their design ST has all the building blocks to build an IoT device like IAS 3DW for vibration monitoring in today's discussion and to lower the barrier to get started we provide development kits like the sensor tile wireless industrial node STWIN that thanks to pre-integrated software allow the user to target a specific vertical application smart industry in our case with function pack industrial predictive maintenance one thank you for your time today and for further information please visit www.st.com and look for solutions for smart industry or STWIN