Machine Controller Using Partial-Derivative Neural Control

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing machine learning-based control systems for machines are not guaranteed to comply with predefined limits or technical constraints, especially when state values are missing or invalid, leading to suboptimal or undesired control actions, particularly in safety-critical applications.

Innovation Solution

A method that assigns vector elements of state vectors to input variables of an artificial neural network, trains the network to reproduce training target values through partial derivatives, and generates control signals using subsets of valid state values, implicitly accounting for hidden correlations and reducing errors by integrating over extreme values.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If learning-based machine control is used to optimize machine behavior, then machine performance is improved, but reliability deteriorates because control actions may not comply with predefined limits or technical constraints when state values are missing or invalid

Engineering Contradiction:
Improvemachine performanceVSAvoidcompliance with technical constraints
Core Design Contradiction:
ProductivityVSReliability

Solution Approach 1:

The system performs preliminary validation of state values before they are used for control decisions. By checking whether state values are present and valid in advance, the system prevents unreliable data from compromising control compliance, thus maintaining reliability while preserving the performance benefits of learning-based control

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The patent introduces an intermediary validation layer between the learning-based control system and the machine actuators. This intermediary component verifies control actions against predefined limits and technical constraints before execution, ensuring compliance even when the learning system produces suboptimal outputs due to missing or invalid state values

Inventive Principle:
Principle #24Intermediary (Mediator)

2Duration of action of moving object

If mean values, median values, or default values are used to compensate for missing state values, then the system maintains operation continuity, but manufacturing precision deteriorates because these measures are insufficient to adequately compensate for absent valid state values

Engineering Contradiction:
Improveoperation continuityVSAvoidcontrol accuracy
Core Design Contradiction:
Duration of action of moving objectVSManufacturing precision

Solution Approach 1:

The system employs a disposable fallback mechanism where pre-computed fallback control actions are used temporarily when state values are missing. These fallback actions are designed to maintain safe operation continuity but are explicitly intended as temporary measures until valid state data becomes available again, preventing long-term degradation of control precision

Inventive Principle:
Principle #27Cheap short-living objects (Disposable)

Solution Approach 2:

The system implements feedback monitoring to detect when state values become missing or invalid. Upon detection, the system switches to fallback control modes and continuously monitors for recovery of valid state data. When valid data returns, the system transitions back to normal precision control, thus maintaining both operation continuity and manufacturing precision

Inventive Principle:
Principle #23Feedback

Data Source

PatentEP4686984A1Method and machine controller for controlling a machine
Publication Date: 2026.02.04 SIEMENS AG
  • EP4686984A1 patent drawingFigure 1~2
  • EP4686984A1 patent drawingFigure 3
  • EP4686984A1 patent drawing

AI summary

Using a large number of training state vectors (TSVs), each assigned a training target value (TTV), a neural network (NN) is trained so that, when a training state vector (TSV) is input, a partial derivative (DN) of the neural network with respect to its input variables (X1,...,XK) reproduces the assigned training target value (TTV), at least on average. Furthermore, a temporal sequence of machine (M) state vectors is recorded. For each state vector, it is detected which vector elements (SV1,...,SVK) contain valid state values ​​of the machine (M), and subsets (S1,...,SM) of those input variables that are assigned valid state values ​​are generated. For each subset (S1,...,SM), a partial derivative of the trained neural network (NN) with respect to the input variables of the respective subset is provided as an evaluation network (DN1,...,DNM). The input variables of the respective subset (S1,...The state values ​​assigned to the subset (SM) are then fed into the evaluation network (DN1,...,DNM), whereby the remaining input variables are set to predefined extreme values ​​(L1,...,LK, U1,...,UK). Depending on this, the evaluation network (DN1,...,DNM) generates a subset-specific output signal (AS1L, AS1U). Finally, a weighted sum of the subset-specific output signals (AS1L, AS1U,...) is calculated and output as the overall output signal (ASG) to control the machine (M).