DE | EN

*PI is an innovative software for re-engineering, monitoring and data management of technical processes*

At the moment you store your measurement data in separate files together with analysis results and related charts? But you would rather work with a professional solution to avoid laborious routine jobs while reducing errors and getting more control over your data?

With PI we created a process data plattform especially for engineers for fast and convenient data organisation, to automate analysis procedures and generating trends with only one mouse click. Thus saving you time for the important things.

You need access to your data everytime and everywhere?

PI is a cloud based database solution for definitely that: access everytime from everywhere with the browser you can work with best. Via Direct Access straight data access becomes real, e.g. to further process data in your own visualisation or analysis software solutions.

You have to carry out always the same analysis procedures, performance indicator calculations, interval conversions, measurement frequency determinations? Or to find sometimes completely new data relations and dependencies?

Our PI analysis modules are made for exactly these purposes: converting raw values in steady intervals, calculate correlations between an unlimited number of data channels in parallel (and display them in heat maps), train and export artificial intelligence / neuronal networks, integrate user defined formulas...

Engineers working with data all the time know this problem: Process models as flow charts are often to be read from left to right but data files are designed from top to bottom. Thus connecting process models with real measurement values is difficult.

With PI you can link your data via Direct Access to your own software applications, visualise, manipulate and convert them there and create individual alarms. This enables powerful but still flexibel process monitoring.

PI comprises four main components: Data storage, Analysis, Visualisation and Alarms & Messages:

Depending on the boudary conditions and clients demand these components are tailor-made: which Visualisation is required, which Analysis modules are relevant, how does Alarms & Messages should be sent etc.

What makes us special compared to others? Many things!

**Reliable**

We are service provider, not data collectors: You will keep full control of your data. No matter if they stay on your server or get stored on our servers.

**Independent**

No equipment manufacturers, no maintenance companies, no energy provider. Only your interests are relevant to us.

**Competent**

Don't understand each other, explaining things several times, friction losses between departments? Doesn't exist for us. As PI was developed by engineers for engineers we know how to handle process data.

**Innovative**

Nothing is as good as it cannot be improved. With our "Functional Updates" we teach PI new ideas.

Allocate data to datasources (plants, processes, devices etc.)

Organise data in channels

Setup date intervals, units and texts, minimum and maximum values and individual chart properties

Add 'result channels' for calculations results

Statistics are updated automatically

Import csv, txt, xls and xlsm files up to 20 MB within a few minutes or even seconds

No special table layouts are necessary

Already existing channels will be updated automatically

New channels including units and texts are created automatically during import

All trends are generated dynamical

Add channels to chart with only one mouse click

Move forward and backward or to certain dates easily

Show correlations between two channels as well as value distribution charts

Generate PNG files for your reports

Change channel date intervals either from raw and single values to regular date intervals or from given date intervals into other intervals

Calculate average values and frequencies as number of values within the given target time period

Use formulas to make simple calculations like average values, differences, ratios, sums, "KPIs" etc.

Run formulas on existing channels into 'result channels' and create new insights

Calculate the linear correlation ('R' factor) for all your channels at once

Apply time shifts to take residence periods into account

Results are shown 'Heat maps'

Train artificial neuronal networks to overcome unknown relationships or uneven measurements at certain processes

Create spreadsheet models out of these networks and use it in comprehensive calculations

Connect your spreadsheet application with PI datasource value database

Create complexe models and feed them with database values by remote control

Write back the results for further processing in PI

Benefit from generating "line-by-line" calculation scripts to make it faster by a ca. factor up to 100, machine code stable and independend from any spreadsheet application

Required for comprehensive process monitoring

Translated scripts run automatically

Engineering companies especially profit from PIs convenient data transfer between storage and calculation tool.

- Centralised measurement/data storage
- Simplified data import/export
- Convenient data transfer between data storage and office/engineering software
- Precise forecasts due to application of artificial intelligence/neuronal networks

Service companies needs to stay "up-to-date" what is going on in the plant/the process. Therefore response times can be kept at minimum; PI allows to set up many alarms & messages according to the given boundary conditions.

- Centralised measurement/data storage
- Data available from everywhere with all mobile devices
- Various alarms & messages, based on self defined limits and "limit" trends

Operators get able to enhance their monitoring by additional values calculated from measurements, individual visualisations and when it makes sense further alarms & messages.

- Simplified data import/export
- Various alarms & messages, based on self defined limits and "limit" trends
- Precise forecasts due to application of artificial intelligence/neuronal networks
- Maximum transparency and flexibility also for very special investigations

Heat exchanger biased with solid loaden fluids get dirty over time. In order to avoid decreasing performance and high pressure drop regular cleaning while operation is necessary.

There are different possibilities to improve a prediction of fouling behaviour in boilers running with solid fuel - e.g. to clean more efficient, to better overhaul scheduling or to ensure making proper boiler modifications:

- Monitoring of attemperator injection
- Monitoring of flue gas temperature differences
- Calculation of heat exchanger efficiencies based on heat transfer coefficients
- Calculation of heat exchanger efficiencies based on simplified "Number of transfer units"
- Monitoring of pressure drops

As every plant is different and flue gas fluctuations hamper analysis there is no "best solution" for this. That's why it makes sense to follow different methods in parallel and validate them during operation.

In PI excactly this is the idea: set up of different models in parallel, minimize measurement fluctuations based on process engineering principles and/or by artifical neuronal networks and check these models continuously while operation. After this we choose an appropriate method together with the operator and integrate it into the DCS.

Flue gas pressure drop: measured (red/green) and "process corrected" in PI (blue)

How efficient is a power plant? Really simple: Benefit (= delivered energy) divided by effort (= fuel energy content) equals the efficiency. What is not a problem for power stations operated with natural gas or oil gets more complexe for solid fuels especially for inhomogenious solid fuels like municipal solid waste, residue derived fuels (RDF) or biomass. Heating value then is normally unknown and in municipal solid waste incineration plants and some biomass power stations the fuel is feeded batchwise.

If this is the case the fuel energy content must be calculated "backward" and all other efficiency related parameters must be taken into account:

- Live steam and feed water
- Primar air preheater and all other existing preheatings
- Slag burnout
- Flue gas volume flow and temperature
- Auxiliary energy for combustion (burner, steam injection)
- Auxiliary energy and operational materials for flue gas and residue treatment

etc.

It is obvious that doing so while operation is not only elaborate but requires also many calculation steps.
In PI we solving this by an overall balance in "flow chart design" enabling two important things:
1. calculation depth is not limited but 2. calculation steps stay transparent and comprehensible anytime.
And CO_{2}? If not measured as concentration at stack we simply calculate this species
from a material balance or by correlation with the fuel energy content...

Three chosen process modules designed with PI: Air preheating, 1. and 2. step of combustion

With PI we are able to create alternative models for process engineering applications based on artificial neuronal networks. In basic terms a neuronal network is a mathematical model to generate output signals from input signals where the output signals statistical are as close as possible to the output signals used for the network training. Important engineering areas are A) pattern recognition and B) development of alternative process models in case standard modelling (inner connection of input and output signals) is not possible. Reasons therefore are:

- mathematical relationship is not known (too many, parallel equations, unclear equilibrium states, unknown transient terms etc.) or
- mathematical relationship is known but available measurements do not fit the model.

Characteristic for alternative models is that inner physical parameters are not known but the model development is much faster and they are often more accurat compared to the "engineering-equation-derived" models. For these reasons alternative process models are important for validation and forecast resp. the conformity of theory and practice.

Process substitute model of a flue gas-steam-heat exchanger

Nowadays power stations are operated based on a large number of measurement devices. These measurements are used on the one hand for process control and on the other hand for balancing all the incoming and outgoing material and energy streams.

To make sure these measurements are "correct" anytime – meaning running within the given accuracy – they have to be checked from time to time. There are three ways doing so:

- Calibration due to a test measurement at the same place and/or under defined boundary conditions
- Lengthwise comparison: How do measurement deviations change over time?
- Crosswise comparison: How does a measurement behave compared to other measurements?

Calibration and lengthwise comparison are state-of-the-art and done by staff or in predetermined periods by accredited testing institutes.

Little known are crosswise comparison methods including balance equalising – if the measurement can be integrated in a overdetermined mass or energy balance – or pattern recognition methods. Both need more effort to set up, but do not cause additional costs during operation. Therefore the operator can see anytime how precise a measurement is and has available reliable results calculated from this measurement.

Waste mass flow and heating value: in PI calculated without and with balance check

Heat exchanger biased with solid loaden fluids get dirty over time. In order to avoid decreasing performance and high pressure drop regular cleaning while operation is necessary.

There are different possibilities to improve a prediction of fouling behaviour in boilers running with solid fuel - e.g. to clean more efficient, to better overhaul scheduling or to ensure making proper boiler modifications:

- Monitoring of attemperator injection
- Monitoring of flue gas temperature differences
- Calculation of heat exchanger efficiencies based on heat transfer coefficients
- Calculation of heat exchanger efficiencies based on simplified "Number of transfer units"
- Monitoring of pressure drops

As every plant is different and flue gas fluctuations hamper analysis there is no "best solution" for this. That's why it makes sense to follow different methods in parallel and validate them during operation.

In PI excactly this is the idea: set up of different models in parallel, minimize measurement fluctuations based on process engineering principles and/or by artifical neuronal networks and check these models continuously while operation. After this we choose an appropriate method together with the operator and integrate it into the DCS.

Flue gas pressure drop: measured (red/green) and "process corrected" in PI (blue)

How efficient is a power plant? Really simple: Benefit (= delivered energy) divided by effort (= fuel energy content) equals the efficiency. What is not a problem for power stations operated with natural gas or oil gets more complexe for solid fuels especially for inhomogenious solid fuels like municipal solid waste, residue derived fuels (RDF) or biomass. Heating value then is normally unknown and in municipal solid waste incineration plants and some biomass power stations the fuel is feeded batchwise.

If this is the case the fuel energy content must be calculated "backward" and all other efficiency related parameters must be taken into account:

- Live steam and feed water
- Primar air preheater and all other existing preheatings
- Slag burnout
- Flue gas volume flow and temperature
- Auxiliary energy for combustion (burner, steam injection)
- Auxiliary energy and operational materials for flue gas and residue treatment

etc.

It is obvious that doing so while operation is not only elaborate but requires also many calculation steps.
In PI we solving this by an overall balance in "flow chart design" enabling two important things:
1. calculation depth is not limited but 2. calculation steps stay transparent and comprehensible anytime.
And CO_{2}? If not measured as concentration at stack we simply calculate this species
from a material balance or by correlation with the fuel energy content...

Three chosen process modules designed with PI: Air preheating, 1. and 2. step of combustion

With PI we are able to create alternative models for process engineering applications based on artificial neuronal networks. In basic terms a neuronal network is a mathematical model to generate output signals from input signals where the output signals statistical are as close as possible to the output signals used for the network training. Important engineering areas are A) pattern recognition and B) development of alternative process models in case standard modelling (inner connection of input and output signals) is not possible. Reasons therefore are:

- mathematical relationship is not known (too many, parallel equations, unclear equilibrium states, unknown transient terms etc.) or
- mathematical relationship is known but available measurements do not fit the model.

Characteristic for alternative models is that inner physical parameters are not known but the model development is much faster and they are often more accurat compared to the "engineering-equation-derived" models. For these reasons alternative process models are important for validation and forecast resp. the conformity of theory and practice.

Process substitute model of a flue gas-steam-heat exchanger

Nowadays power stations are operated based on a large number of measurement devices. These measurements are used on the one hand for process control and on the other hand for balancing all the incoming and outgoing material and energy streams.

To make sure these measurements are "correct" anytime – meaning running within the given accuracy – they have to be checked from time to time. There are three ways doing so:

- Calibration due to a test measurement at the same place and/or under defined boundary conditions
- Lengthwise comparison: How do measurement deviations change over time?
- Crosswise comparison: How does a measurement behave compared to other measurements?

Calibration and lengthwise comparison are state-of-the-art and done by staff or in predetermined periods by accredited testing institutes.

Little known are crosswise comparison methods including balance equalising – if the measurement can be integrated in a overdetermined mass or energy balance – or pattern recognition methods. Both need more effort to set up, but do not cause additional costs during operation. Therefore the operator can see anytime how precise a measurement is and has available reliable results calculated from this measurement.

Waste mass flow and heating value: in PI calculated without and with balance check

Got interested?