TOP
Ho-Ho-Ho!
AI-based sensor technology for digitalization

Our soft sensors support you in your digitalization tasks. They capture an input data image of machines, systems or processes, evaluate it based on AI and provide the appropriate target values for your low-code applications, for example.

A soft sensor, also known as a virtual sensor or sensor fusion, is not a classic, real, physical hardware sensor, but a model-based linking of representative measured variables to a target variable.

Soft sensors calculate the desired target variable as an output value from various input variables that correlate with it.

The input data image can consist of measurement data from individual hardware sensors (such as our MLS/160A), but also measured values or variables from other sources. It is important that there is a mathematically describable correlation between the individual input variables and the target variable.

With the help of machine learning (ML), a sub-area of AI, very powerful soft sensors can be created for various areas of application.

Intelligent Data for Low-code/SCADA Applications

The results of the data evaluation of a soft sensor can then be used by a low-code application in the production area of innovative factory systems, e.g. for predictive maintenance, machine vision, control of storage systems, for the operation

of autonomous vehicles and for a wide range of condition monitoring applications, such as fill level monitoring using ultrasonic sensors or condition monitoring of a specific radio spectrum, and much more.

Soft sensor in practice

Cybersecurity for IoT Endpoints

We have developed an AI-based embedded intrusion detection system (IDS) based on a soft sensor that enables resource-limited embedded systems to use AI to monitor network traffic and identify and report anomalies.

This allows cyberattacks to be detected at an early stage and defensive measures to be initiated in good time.

Product description

Embedded IDS

Three Examples of a Soft Sensor

Gateway as a Soft Sensor

Here, the input data image is recorded by (classic) external sensors and/or interface connections to internal machine and system controls.

The necessary software components, such as the so-called inference engine for data evaluation via ML, are installed and configured accordingly on a gateway.

Fig. 1: Gateway as a soft sensor

Condition Monitoring of a Production Machine

Various MEMS sensors with 3D acceleration data form the input data image of the soft sensor. The (virtual) event and operating hours counters are supplied as target variables.

This solution can be used, for example, to optimize the maintenance schedules of a machine.

Embedded Soft Sensor

Here, the runtime environment of the soft sensor is not an external gateway but an embedded system integrated directly into the machine or system.

The input data image is created by the data from the internal sensors and controllers.

Fig. 2: Embedded Soft Sensor

Automatic Energy Management

In this scenario, a target variable is formed from machine states and environmental data, which is linked in a low-code application, e.g. with a variable electricity price, in order to generate a feedback signal for a power modulation input.

Wireless Sensors as a Source for the Soft Sensor

Of course, the data for the soft sensor can also be recorded wirelessly.

In this example, the input data image is generated by wireless sensors or a Wireless Sensor Network (WSN).

Fig. 3: Wireless sensors as a source for soft sensor

Climate Control for Greenhouses

The input data image is formed by measured values for temperatures, humidity, light, wind, rain, etc.

A data vector for controlling lighting, heating, pumps and other actuators is supplied as the target variable. The soft sensor also generates an alarm signal.

Seven Steps to an AI-based Soft Sensor

For the development of AI-based soft sensors, we have designed a tried-and-tested process consisting of seven steps that can be used to create soft sensors tailored precisely to the respective application scenario.

The core component of this development process is a tool developed by us, the SSV Data Exploration Tool, or SSV/DET for short. Among other things, it is used to train, generate and test ML models.

Process Step Description

1.

Application-related data acquisition Acquire real status or operating data in the target environment and store it in a data pool. A sufficient quantity of high-quality data is required for the subsequent steps, from which an ML model learns the decisive correlations.

2.

Application-related feature selection Analysis of the recorded data and selection of suitable features for the input data image of the soft sensor. Determining the correlations to the desired target variable at the soft sensor output.

3.

ML model development Collect and evaluate sufficient process knowledge. Determine the final input data image and select the required data pre-processing steps. Design ML model and select suitable algorithms to obtain the desired target value from the input data image. Prepare the adaptation of the SSV/DET.

4.

ML model training Prepare suitable training data from the data pool and create the ML model in a training phase. Integration of the ML training code into the respective SSV/DET to enable training repetitions.

5.

ML model testing and integration Create inference code. Prepare suitable test data from the data pool in order to validate the ML model. If the test result is insufficient, repeat the previous steps in part or in full. If the result is good, integrate the test into the SSV/DET so that new model versions can be generated and tested via SSV/DET in future.

6.

Practical use of the soft sensor Integrate the inference code and ML model into the target environment and test under end-to-end conditions if possible. A suitable test and debug environment or tool selection should be used for this. Furthermore, the inference code and model should be exchangeable with the help of the SSV/DET via an OTA update option. The respective cybersecurity requirements must be observed, e.g. through model signatures.

7.

Softsensor CI/CD pipeline Setting up a CI/CD pipeline. Continuously monitor the operation of the AI-based soft sensor and, if necessary, create, test and deploy the required updates via SSV/DET.

On-Demand-Webinar

Low-Code-Anwendungen mit KI-basierten Softsensoren

Wenn Sie mehr zum Thema Softsensoren erfahren möchten, dann nutzen Sie gerne unser kostenloses On-Demand-Webinar, in dem wir u. a. folgende Fragen beantworten:

  1. Was ist ein Softsensor und wie ist er aufgebaut?
  2. Wofür benötigt ein Softsensor die KI?
  3. Wie sieht der Entwicklungsprozess eines Softsensors aus?
  4. Wie sehen typische Anwendungsbeispiele für Softsensoren aus?

Selbstverständlich gehen wir während des Webinars auch gerne auf Ihre individuellen Fragen ein!

Die geplante Dauer des Webinars beträgt ca. 60 Minuten.

Jetzt Termin vereinbaren!

Additional Services through Docker Containers

In many cases, soft sensor applications require various support functions, for which we offer so-called Service Docker Containers. The SDU Docker Container (SDU = Secure Device Update) enables secure OTA updates (OTA = Over-the-Air), for example to update the ML model of a soft sensor.

With our WRD Docker Container (WRD = Wireless Remote Debugging), remote debug connections can be established.

This allows soft sensors to be tested remotely in the real application environment directly from your desk.

Docker container for soft sensor

SSV/ITB: Decentralized Testbed for IoT Data Evaluations

A sufficient amount of high-quality data is required to develop a suitable machine learning model. Although this data can be generated synthetically in a development or laboratory environment, the practical benefit of such simulation data sets in industrial IoT applications is very limited. In order to collect, prepare and process the required data for the respective application directly in an OT and evaluate it directly in an OT

environment, we offer a decentralized testbed as a service. This IoT Data Evaluation Testbed (SSV/ITB) was specially developed for AI-based soft sensor solutions.

It is particularly suitable for all IoT applications in which sensor data from machines, systems and processes is required for use in a backend system.

Interrelationships in the Decentralized Testbed

SSV/ITB: Decentralized testbed for IoT data evaluations

Fig. 5: Interrelationships of the IoT Data Evaluation Testbed (SSV/ITB)

In this example, the goal is to develop a low-code application (1) to monitor selective processes within an OT environment (2), coordinate a maintenance appointment for a service technician in certain situations and enter it in an electronic appointment calendar. The entire application digitizes a manual monitoring and scheduling process.

The data aspects:The respective actual state in the OT environment (i.e. the testbed target environment for data evaluation) is to be measured using a suitable sensor system (e.g. with hardware sensors or by accessing control data). The resulting output data, i.e. the target variables of the soft sensor, must be easily processed by the low-code application. This requirement can be met with an AI-based soft sensor (3).

Preparation: A data evaluation with a suitable (distributed) testbed is planned for the design of the soft sensor. A Mobile Testbed Unit (MTU) is used on site at the data source.

It is connected to the OT environment via interfaces and sensors (4) in order to capture an input data image suitable for the task. The MTU is also linked to a cloud service (5) via a mobile phone interface.

The collaboration phase: With the help of the cloud service, a multidisciplinary team of SSV data experts and the user can iteratively define the input data image, perform data analyses and design a suitable model or AI algorithm to generate the target value for the low-code application.

Want to Know More?

If you have any questions about soft sensors, our sales team will be happy to help you!

Phone: +49(0)511 · 40 000-34

Email: sales@ssv-embedded.de

SSV SOFTWARE SYSTEMS

Dünenweg 5
30419 Hannover

Phone: +49(0)511 · 40 000-0
Fax: +49(0)511 · 40 000-40

sales@ssv-embedded.de


Imprint    ·    Privacy Policy    ·    Terms & Conditions

© 2025 SSV SOFTWARE SYSTEMS GmbH. All rights reserved.

ISO 9001:2015