GB2512087A - Apparatus for controlling a manufacturing plant - Google Patents

Abstract

Controlling a process plant using a coarse tuning assembly and a fine tuning assembly, the coarse tuning assembly having an adaptive fuzzy logic controller 3, the fine tuning assembly 4 having a non linear dynamic linearized regression controller feeding into an APACC or adaptive cognised control.

Description

APPARATUS FOR CONTROLLING A MANUFACTURING PLANT

This invention relates to control apparatus and, more especially, this invention relates to apparatus for controlling a manufacturing plant. The apparatus is able to use artifical precognition (AP) using adaptive (model) cognized control (APACC) as a technique in control engineering that improves upon model predictive control (MPC) for the manufacturing plant.

The control of multivariable multi-input multi-output (MIMO) systems is a common problem in practical control scenarios including manufacturing plants. However, in the last two decades, of the known advanced control schemes, only linear model predictive control (MPC) has been widely used in industrial process control. The fundamental idea behind all MPC techniques is to rely on predictions of a plant model to compute the optimal future control* sequence by minimization of an objective function. MPC models include controlled variables, manipulated variables and disturbance (perturbation) variables. In the predictive control domain, generalized predictive control (GPC) and its derivatives have received special attention. Especially, the ability of GPC to be applied to unstable or time-delayed MIMO systems in a straight forward manner and the low computational demands for static models make the GPC useful for many different kinds of tasks. Systems behaving unexpectedly, human factors, failures and the environment are all factors that contribute to non-linear plant dynamics. The drawback with MPC is that it is limited to linear models. If nonlinear dynamics are present in the plant, a linear model might not yield sufficient predictions for MPC techniques to function adequately.

The two most general approaches to closed loop identification are direct approach and indirect approach. The direct approach ignores the presence of feedback, and directly identities the plant by plant input and output data. This has the advantage that rip knowledge about the type of control feedback or even linearity of the controller is required The indirect approach identifies the closed loop, and obtains the open loop model by deconvolution if possible. Obtaining the open loop model is only possible if the controller is known and both the closed loop, plant model and the controller are linear. - * It is aim of the present invention tp reduce the above mentioned problems.

Accordingly, in one non-limiting embodiment of the present invention there is provided apparatus for controlling a manufacturing plant which apparatus comprises a coarse tuning assembly and a fine tuning assembly, the coarse tuning assembly being such that it comprises: a. sensor interfaces measuring large-but-low-frequency perturbations, including predictable equipment failures and emergency maintenance, b. fuzzy performance descriptions for human dynamics, machine processes controllers, resource schedulers product release generators, c. an adaptive fuzzy logic con&oller for nonlinear MIMO systems with subsystems which comprise fuzzification, inference, and output processing, which comprise both type reduction and defuzzification and which provide stability of a resulting*closed loop system, the adaptive fuzzy logic controller including: (I) inference engine identifying relationships using a rule base and outputs as fuzzy sets' to a type reducer, and (ii) output control demands including commands which are for human resources and actuators including mechanical $ and electronic to the fuzzyfier and which are for fuzzyfying' the signal, and the fIne tuning assembly being uch that it compris: a. inputs from the coarse tuning assembly, b. precognition horizons determining how many future samples the objective function considers for minimization and the length of the control sequence computed, c. a linearized MIMO regression model extracted from the adaptive fuzzy logic controller at each time step providing the fine' tuning parameters, and d. a non-linear dynamic linearized regression controller providing: - (i) a crisp output signal feeding into APACC synthesis computing optimal performance of the manufacturing plant, and (ii) reduced set output and APACC synthesis feding into the APACC linear logic system.

The apparatus may be one which includes a synchronization assembly which optimises the input signal to the output signals and which comprises cascaded diophantine frequency synthesis (DFS) means which predicts future stabilization output parameters of the manufacturing plant.

The use of neural networks for system identification is a relatively new approach, as is fuzzy logic.

Embodiments of the invention will now be described solely by ay of example and with reference to the accompanying drawings in which: Figure lJs a schematic of APACC application in a manufacturing plant; Figure 2 shouts Intel® Xeon® processor high performance computing for adaptive cognised control (APACC) in a manufacturing plant; and Figure 3 is a schematic for artificial precognition using adaptive cognised control (APACC).

Referring to the drawings, non-linear plant dynamics can be characteristically fuzzy' with a high degree of non-linearity. APACC feeds the instantaneous linearization of a nonlinear model with the cognized' output of a fuzzy logic circuit (fuizyifier in Figure 3) in each sampling instant. It is similar to GPC in most aspects except that the instantanecjus linearization of the fuzzy logic circuit output yields an adaptive linear regression model.

A key benefit of fuzzy logic is that it lets the designer describe the desired system behaviour with simple if-then' relations. In many applications this gets a simpler solution in less design time. In addition, the designer can use all available engineering know-how to optimise the system performance directly. While this is certainly the beauty of fuzzy logic, it has also been a major limitation. In many applications, knowledge that describes desired system behaviour is contained in data sets. Here the designer has had to derive the if-then' rules from the data sets manually, which requires a major effort with large data sets. When data sets contain knowledge about the system to be designed, a neural net promises a solution because it can train itself from the data sets. The present invention utilises a fuzzy logic controller that automates rule derivation eliminating the need to perform this function manually to predict plant dynamics instantaneously While neural nets are at advantage by learning from data sets, these have inherent disadvantages; for instance, the cause for a particular behaviour cannot be interpreted, nor can a neural net be modified manually to change to a certain desired behaviour. Also, selection of the appropriate net model and setting the parameters of learning algorithm are difficult and require much experience. On the other hand, fuzzy logic solutions are easy to verify and optimise.

A key advantages of APACC are that it can be used for a broad range of applications including: -Flexibility to plant size, automated setup -Based on step responselimpulse response model -On the fly reconfiguration if plant is changing -Systematic handling of multi-rate measurements and missed measurement points.

Fuzzy control methodologies have emerged in recent years as promising ways to approach nonlinear control problems. Fuzzy control, in particular, has had an impact in the control community because of the simple approach it provides to use heuristic control knowledge for nonlinear control problems. In very complicated situations, where,the plant parameters are subject to perturbations or when the dynamics of the systems are very complex, adaptive schemes have to be used online to gather data and adjust the control parameters automatically. However, no stability conditions have been provided so far for these adaptive approaches. APACC introduces two components into, its adaptive fuzzy-cdntrol scheme. One is a fuzzy logic system for coarse tuning. The other is the instantaneous linearization of the ifuzz'y logic circuit output which yields an adaptive linear model. This acts as a kind of robust compensator, such as supervisory, control, sliding-mode control, for the fine-tuning.

Recently, several stable adaptive fuzzy control schemes have been developed for multiple-input-multiple-output (MIMO) nonlinear systems.

However, these adaptive control techniques are only limited to the MIMO nonlinear systems whose states are assumed to be available for measurement. In many practical situations, state variables are often unavailable in nonlinear systems. Thus, the output feedback or APACC adaptive fuzzy control is required for such complicated applications. The fuzzy control system controls the MIMO system and maintains the system stability. The coarse and fine tuning improves system, performance by reducing the impact of external perturbations guaranteeing closed-loop stability.

The apparatus design is such that there are three levels of schematic.

APACC algorithms provide "optimal" performance with respect to large-but-low-frequency perturbations, such as predictable equipment failures and emergency maintenance. APACC is the basis for quality control assurance, prediction of time to failure, raw material needs and inventory control. This includes effective interaction with for example, fabrication facilities, with bi-directional interfaces for all controllers, schedulers, and release generators.

Figure 1 illustrates the integration of APACC with distributed*paralel processing systems*, for information support, the datacenter and the core manufacturing execution systems. There are four levels in the hierarchs& The lowest level (level 1) comprises machine process controllers; next (!evel 2) APACC; third (level 3) resource schedulers, and finally (level 4) the product release generators.

Figure 2 shows how APACC is used in manufacturing plant predictive control. The design in controlling plant operation, performance and maintenance requires sensors, actuators and algorithms to command the actuators based on (1) sensor measurements of prevailing plant performance and (2) specification of desired plant performance.

The plant has an output corresponding to automatic plant operation and performance as a consequence of the application of APACC. The essential components for artificial precognition are embedded as subsystems within a high performance computer hardware core system datacenter implementation I shown in Figure 3. The detailed implementation of the core system is shown in Figure 3. The inputs include perturbations to the plant as discussed above. The inputs will include a great number of variables from simple to very complex, linear and nonlinear, as well as fixed or variable dynamics. Input/output is managed by a multivariable multi-input multi-output (MIMO) subsystem 2 as shown in Figure 3. The sensor deltas are measured by sensors through a sequential control process which comprises feedback measurement (feedback elements) and a comparator that compares the differential between the input and output signals. The MIMO subsystem 2 is driven to different 9perating states using. set points which are added to the input signal at the comparator. The actuaLexcitation signal amplitude has to be as large as possible to ensure maximal excitation around each set point.

AP.ACC instantaneously compensates the error between the actual and -V cognized (possible and probable) outputs. This error is shown in the MIMO subsystem 2 as being fed into a controller which provides control outputs to actuators. The sensor actuator applies the mechanical action needed to change vehicle direction or speed.

The coarse tuning is provided by the adaptive fuzzy logic controller 3 shown in Figure 3. It has been developed for nonlinear MIMO systems involving external perturbations using fuzzy descriptions to model plant operation. The adaptive fuzzy logic controller is a Type-2 fuzzy logic system involving the operations of fuzzification, inference, and output processing. The sensor actuators output control demands to the fuzzyfler which fuzzyfies' the signal. There are variables (mentioned above) within the inputs that are closely correlated, and they will have high mutual information.

However, there are other pairs of variables that are related that will have low correlation, but high mutual information. The coarse tuning radically speeds up the cognition process making it possible to make virtually instantaneous dynamic correction possible. The Inference engine identifies these relationships using the rule base and outputs these as fuzzy sets' to the type reducer. The adaptive fuzzy logic controller provides "output processing" and comprises both type reduction and defuzzification.

Type reduction (reduced set) captures more information about rule uncertainties than does the defuzzified value (a crisp number); however, it is computationally intensive. The advantage is that it can cognize unpredicted perturbations-data uncertainties The adaptive fuzzy controller can perform successful control and guarantee that the global stability of the resulting closed-loop system and the tracking performanc can be achieved.

The adaptivefuzzy logic controller output processing is fed into a non-linear dynamic linearized regression controller 4 shown in Figure 3, and which provides the fine tuning. The crisp output signal feeds into APACC synthesis.

The reduced set output and the APACC synthesis feed into the APACC linear logic system. The linearized MIMO regression model that is extracted from the fuzzy logic controller at each time step is used to provide the fine' tuning parameters. Non-linear dynamic linearized regression controller computes the optimal future guidance, navigation and control sequence according to the objective function. Two precognition horizons' determine how many future samples the objective function considers for minimization and the length of the control sequence that is computed. As is common in most MPC methods, a receding horizon strategy is used and thus only the first control signal that is computed is actually applied to the vehicle systems to achieve loop closure.

Synchronization optimization of the input signal to the output signals is achieved using cascaded diophantine frequency synthesis (DFS) implemented using two or more phase lock loops (PLL). The DFS is DFS 5 in Figure 3. The DFS 5 is used to predict future outputs of continuous-time, infinite-dimensional, time-varying and non-linear systems. Its primary function is to parameterize vehicle stabilization factors and lock them into a continuous feedback loop. -The APACC for plant operations and performance requires the virtually instantaneous analysis of enormous data volumes.. Id achieve this, the convey high performance computing (HPC) architecture from Intel® was selected for APACC in manufacturing plants. The convey computers approach provides very fast access to random access to memory, and is very useful for the complex functions used in APACC.

The architecture is based on Intel® Xeon® processor as shown in the schematic of Figure 2. The architecture features a highly parallel memory subsystem to further increase performance. Programmable "on the fly," FPGAs are a way to achieve hardware-based, application-specific performance. Particular APACC-algorithms, for example, are optimized and translated into code that is loaded onto the FPGAs at runtime.

Large-but-low-frequency perturbations such as predictable equipment failures and emergency maintenance, present inputs to APACC in the manufacturing plant.

APACC plant performance optimization and maintenan comprises sensors, human resources, actuators and algorithms to command the actuators based on (1) sensor measurements of prevailing plant performance and (2) specification of desired plant performance.

APACC algorithms provide "optimal" performance with respect to large-but-low-frequency perturbations, such as predictable equipment failures anøemergency maintenance.

APACC supports quality control assurance, prediction of time to failure, raw material needs and inventory control.

The multivariable, multi-input multi-output (MIMO) subsystem comprises a sensor assembly essentially consisting of a sequential control process measuring feedback and a comparator comparing the differential l5etween the input and output signals. The MIMO subsystem 2 operates as follows.

a. The MIMO subsystem 2 is driven to different operating states using set points added to the input signal at the comparator b. To ensure maximal excitation around each set point, the excitation signal amplitude is maximized.

c. APACC instantaneously compensates the error between the actual and cogriized (possible and probable) plant outputs of the core manufacturing execution systems including fabrication facilities, schedulers, and release generators.

d. The error is fed into the controller providing control outputs to sensor actuators and human resources providing the actions needed to change plant performance The high performance computing (HPC) architecture providing very fast access to random access merriory for virtually instantaneous analysis of enormous data volumes needed to optimise plant performance It is to be appreciated that the embodiments of the invention described above with reference to the accompanying xlrawings have been given by way of example only and that modifications may be effected. Individual pomponents shown in the drawings are.:not limited to use in their drawings and they may be used in other drawings and in all aspects of the invention.

Claims (3)

1. CLAlMS Apparatus for controlling a manufacturing plant, which apparatus comprises a coarse tuning assembly and a fine tuning assembly.the coarse tuning assembly being such that it comprises: a. sensor interfaces measuring large-but-low-frequency perturbations, including predictable equipment failures and emergency maintenance, b. fuzzy performance descriptions for human dynamics, machine processes controllers resource schedulers, product release generators, c. an adaptive fuzzy logic controller for nonlinear MIMO systems with subsystems which comprise fuzzification, inference, and output processing, which comprise both type reduction and defuzzification, and which provide stability of a resulting closed-loop system, the adaptive fuzzy logic controller including: (i) inference engine identifying relationships using a rule base and outputs as fuzzy sets to a type reducer, and (ii) output control demands including commands which are for human resources and actuators including mechanical and electronic to the fuzzyfier and which are for fuzzyfying' the signal; and the fine tuning assembly being such that it comprises: a. inputs from the coarse tuning assembly, b. precognition horizons determining how many future samples the objective function considers for minimization and the length of the control sequence computed, c. a linearized MIMO regression model extracted from the adaptive fuzzy logic controller at each time step providing the fine' tuning parameters, and d. a non-linear dynamic linearized regression controller providing: a crisp output signal.feeding into APACC synthesis computing optimal performance of the manufacturing :,.: plantand.(ii) reduced set output and APACC synthesis feeding into the APACC linear logic system.
2. 2. Apparatus according to claim 1 and including a synchronization assembly which optimizes the input signal to the output signals, and which comprise cascaded diophantine frequency synthesis (DES) means which predicts future stabilization output parameters of the manufacturing plant..
3. 3. Apparatus for controlling a manufacturing plant, substantially as herein described with reference to the accompanying drawing&